© Jason Bradley Fice, 2019 NEUROMECHANICS OF NECK MUSCLES: IMPLICATIONS FOR WHIPLASH INJURY by Jason Bradley Fice B.A.Sc., University of Waterloo, 2008 M.A.Sc., University of Waterloo, 2010 A DISSERTATION SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Kinesiology) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) April 2019 ii The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled: Neuromechanics of Neck Muscles: Implications for Whiplash Injury submitted by Jason Bradley Fice in partial fulfilment of the requirements for the degree of Doctor of Philosophy in The Faculty of Graduate and Postdoctoral Studies (Kinesiology) Examining Committee: Jean-Sébastien Blouin, Kinesiology Co-supervisor Gunter P. Siegmund, Kinesiology Co-supervisor Peter A. Cripton, Mechanical Engineering Supervisory Committee Member Thomas R. Oxland, Mechanical Engineering University Examiner Peter R.E. Crocker, Kinesiology University Examiner iii Abstract Occupants involved in automobile collisions can adopt various postures preceding impact. Epidemiologic data have shown increased rates of whiplash associated disorders for drivers who have a rotated head posture or braced themselves before impact. In this dissertation we investigated how neck muscle activity and head/neck kinematics of volunteers are influenced by rotated non-neutral head postures and by bracing against the steering-wheel. We also provided important biomechanical and neuromuscular data to improve computational human body models of the neck (HBMs). In experiment one, 9 males performed neck maximum voluntary isometric contractions (MVIC) in 17 three-dimensional directions. We discovered how neck MVICs scale in directions not aligned with the principal axes and discuss how this scaling relationship can be used to validate an HBM’s off-axis strength, an important step before simulating rotated head postures. In experiment two, the biomechanical lines of action of three key neck muscles, as determined via electrical stimulation, were compared to their preferred activation directions, as determined via muscle activity during 15% MVICs in 26 3D directions (8M volunteers). We showed that neck muscles’ biomechanical lines of action are not an accurate predictor of their function, a finding which will help guide the development of realistic neck muscle controllers in HBMs. In experiment three, we quantified the non-neutral head postures that 20(14M, 6F) drivers adopted while driving on public roads. These data were then used in experiment four, where 12(5F, 7M) volunteers were exposed to low-speed rear impacts on a perturbation sled while in a neutral posture or four common non-neutral postures. In the final experiment, 11(3F, 8M) volunteers were exposed to low-speed frontal and rear impacts with their hands on the steering-wheel while either relaxed or braced by pushing with their arms. We found that non-neutral postures and bracing increased pre-impact muscle activity, but generally did not alter peak muscle activity during impact. Further, sagittal plane kinematic changes suggest a stiffer neck in both non-neutral postures and bracing, but non-neutral postures resulted in motions beyond the sagittal plane. The results of these experiments will help inform injury prevention methods, improve HBMs, and ultimately lead to safer automobiles. iv Lay Summary Neck injuries are the most common injury in automobile collisions. These injuries occur more often when drivers’ heads are turned and/or when drivers brace themselves before a collision. The goal of this dissertation was to measure human volunteers’ neck muscle activity and head/neck movements during applied body motions similar to low-speed vehicle collisions. Before impact, the volunteers either relaxed and looked straight-ahead, turned their head at angles common while driving, or braced by pushing against the steering wheel. Compared to relaxed and looking straight ahead, turning the head led to more left/right head and neck movements, and bracing led to less head movement but potentially higher neck forces. Neck muscle activity before impact increased when turning your head or bracing, but the level of muscle activity during impact was similar in all three conditions. These data will help improve computer models of humans and lead to safer vehicles. v Preface The work presented in this dissertation was conducted at the Sensorimotor Physiology Laboratory at the University of British Columbia, Point Grey Campus or MEA Forensic Engineers & Scientists, Richmond, British Columbia. All the experiments were approved by the University of British Columbia’s Research Ethics Board. The ethical certificate numbers were H11-02323 for Chapters 2 and 3, H16-01864 for Chapter 4, and H16-00378 for Chapters 5 and 6. A version of Chapter 2 has been published as Fice, J. B., Siegmund, G. P., & Blouin, J. S. (2014). Prediction of three dimensional maximum isometric neck strength. Ann Biomed Eng., 42(9), 1846-1852. Chapter 3, or a version of it, was published as Fice, J. B., Siegmund, G. P., & Blouin, J. S. (2018). Neck muscle biomechanics and neural control. J Neurophysiol., 120(1), 361–371. A form of Chapter 4 has been published as Fice, J. B., Blouin, J. S., & Siegmund, G. P. (2018). Head postures during naturalistic driving. Traffic Inj Prev., 19(6), 637–643. In Chapters 2 through 4, all authors contributed to the concept and design of the experiment, data interpretation, and manuscript editing. I was wholly responsible for the data collection, data analysis, and writing of the manuscript. Chapter 5 is in preparation for publishing with the following authors and title: Fice, J. B., Mang, D.W., Ólafsdóttir, J.M., Brolin, K., Cripton, P.A., Blouin, J. S., & Siegmund, G. P. Head/neck kinematics and muscle responses in volunteers with non-neutral initial head postures during low-speed rear impacts. All authors contributed to the concept and design of the experiment. I led the collection of the data with assistance from Mang and Ólafsdóttir. Data interpretation was a collaboration between Blouin, Siegmund and me. I was wholly responsible for data analysis and writing the manuscript. Manuscript editing was performed by Cripton, Blouin, Siegmund, and me. Chapter 6 is also in preparation for publishing with the following authors and title: Fice, J. B., Mang, D.W., Cripton, P.A., Blouin, J. S., & Siegmund, G. P. Neck muscle and head/neck kinematic responses while bracing against the steering wheel during frontal and rear impacts. All authors contributed to the concept and design of the experiment. I led the collection of the data with assistance from Mang. I was wholly responsible for data analysis and writing the manuscript. Data interpretation was a collaboration between Blouin, Siegmund and me. Manuscript editing was performed by Cripton, Blouin, Siegmund, and me. vi Table of Contents Abstract ........................................................................................................................................................ iii Lay Summary ................................................................................................................................................ iv Preface .......................................................................................................................................................... v Table of Contents ......................................................................................................................................... vi List of Tables ................................................................................................................................................ ix List of Figures ............................................................................................................................................... xi Acknowledgements .....................................................................................................................................xiii Dedication ................................................................................................................................................... xiv Chapter 1. Introduction .......................................................................................................................... 1 1.1 Motivation for Research ............................................................................................................... 1 1.1 Background ................................................................................................................................... 1 1.1.1 Musculoskeletal anatomy of the neck .................................................................................. 1 1.1.2 Neck neuromuscular physiology ........................................................................................... 3 1.1.3 Whiplash injury ..................................................................................................................... 5 1.1.4 Volunteer impact research ................................................................................................... 8 1.1.5 Muscle activation in human body models .......................................................................... 10 1.2 Objective & Goals ........................................................................................................................ 12 Chapter 2. Prediction of three-dimensional maximum isometric neck strength ................................. 14 2.1 Preamble ..................................................................................................................................... 14 2.2 Introduction ................................................................................................................................ 14 2.3 Methods ...................................................................................................................................... 15 2.3.1 Subjects ............................................................................................................................... 15 2.3.2 Procedures and Instrumentation ........................................................................................ 16 2.3.3 Data Analysis ....................................................................................................................... 18 2.4 Results ......................................................................................................................................... 19 2.5 Discussion .................................................................................................................................... 22 Chapter 3. Neck muscle biomechanics and neural control .................................................................. 26 3.1 Preamble ..................................................................................................................................... 26 3.2 Introduction ................................................................................................................................ 26 3.3 Methods ...................................................................................................................................... 28 3.3.1 Subjects ............................................................................................................................... 28 3.3.2 Spatial Tuning Curve Procedures ........................................................................................ 29 3.3.3 Neck Muscle Stimulation Procedures ................................................................................. 30 3.3.4 Data Analysis ....................................................................................................................... 31 vii 3.4 Results ......................................................................................................................................... 33 3.5 Discussion .................................................................................................................................... 42 Chapter 4. Head postures during naturalistic driving ........................................................................... 47 4.1 Preamble ..................................................................................................................................... 47 4.2 Introduction ................................................................................................................................ 47 4.3 Methods ...................................................................................................................................... 48 4.3.1 Subjects ............................................................................................................................... 48 4.3.2 Instrumentation .................................................................................................................. 48 4.3.3 Procedures .......................................................................................................................... 49 4.3.4 Data Analysis ....................................................................................................................... 49 4.4 Results ......................................................................................................................................... 52 4.5 Discussion .................................................................................................................................... 57 Chapter 5. Head/neck kinematics and muscle responses in volunteers with non-neutral initial head postures during low-speed rear impacts .................................................................................................... 60 5.1 Preamble ..................................................................................................................................... 60 5.2 Introduction ................................................................................................................................ 60 5.3 Methods ...................................................................................................................................... 62 5.3.1 Subjects ............................................................................................................................... 62 5.3.2 Instrumentation .................................................................................................................. 62 5.3.3 Protocol ............................................................................................................................... 64 5.3.4 Data Analysis ....................................................................................................................... 65 5.3.5 Statistics .............................................................................................................................. 68 5.4 Results ......................................................................................................................................... 69 5.5 Discussion .................................................................................................................................... 79 5.6 Data Availability .......................................................................................................................... 82 Chapter 6. Neck muscle and head/neck kinematic responses while bracing against the steering wheel during frontal and rear impacts .................................................................................................................. 83 6.1 Preamble ..................................................................................................................................... 83 6.2 Introduction ................................................................................................................................ 83 6.3 Methods ...................................................................................................................................... 84 6.3.1 Subjects ............................................................................................................................... 84 6.3.2 Instrumentation .................................................................................................................. 85 6.3.3 Protocol ............................................................................................................................... 87 6.3.4 Data Analysis ....................................................................................................................... 88 6.3.5 Statistics .............................................................................................................................. 90 viii 6.4 Results ......................................................................................................................................... 91 6.5 Discussion .................................................................................................................................. 103 6.6 Data Availability ........................................................................................................................ 107 Chapter 7. General Discussion and Conclusions ................................................................................. 108 7.1 Implications for modelling research ......................................................................................... 111 7.2 Aetiology of whiplash ................................................................................................................ 114 7.3 General limitations .................................................................................................................... 116 7.4 Future Research ........................................................................................................................ 117 7.4.1 Future volunteer experimental work ................................................................................ 117 7.4.2 Future computational work .............................................................................................. 118 7.5 Conclusions ............................................................................................................................... 119 Bibliography .............................................................................................................................................. 120 Appendix ................................................................................................................................................... 144 ix List of Tables Table 2.1 - Anthropomorphic data of the male subjects in this study. ...................................................... 15 Table 2.2 - The direction cosines for the 17 predefined MVIC directions. ................................................. 17 Table 2.3 - Confidence intervals for the normalized moment, experimental isometric neck strength, and resultant moments predicted using the principal axis moments. .............................................................. 20 Table 2.4 - Isometric neck strength reported in moments about the C7-T1 for studies that included male subjects in a neutral posture. ..................................................................................................................... 25 Table 3.1 - Anthropomorphic data of the male subjects in this study. ...................................................... 29 Table 3.2 - Voluntary preferred and electrically stimulated directions. ..................................................... 36 Table 3.3 - The difference between voluntary preferred and electrically stimulated directions. .............. 36 Table 4.1 - Anthropomorphic data of subjects in this study. ...................................................................... 48 Table 4.2 - Percentage of time the vehicle was stationary (bottom row) compared to the percentage of movements for each movement type while the vehicle was stationary. ................................................... 53 Table 4.3 - The group averages for yaw, pitch and roll at the instant of peak yaw angle for each head movement. .................................................................................................................................................. 53 Table 5.1 - Anthropomorphic data of participants in this study. ................................................................ 62 Table 5.2 - The target and actual head orientation of subject before the experimental conditions. ........ 71 Table 5.3 - Initial position and orientation of the head before impact in each of the conditions tested. . 71 Table 5.4 - The median (1st, 3rd quartile) pre-impact and peak normalized RMS EMG activity for the muscles in the neutral, left shoulder check, left mirror check, rear-view mirror check, and look-at-passenger experimental trials. .................................................................................................................... 72 Table 5.5 - The median (1st, 3rd quartile) onset of EMG activity for the muscles in the neutral, left shoulder check, left mirror check, rear-view mirror check, and look-at-passenger experimental trials. .. 73 Table 5.6 - The median (1st, 3rd quartile) peak kinematic variables in the neutral, left shoulder check, left mirror check, rear-view mirror check, and look-at-passenger experimental trials. ............................ 74 Table 5.7 - The median (1st, 3rd quartile) onset of select kinematic variables in the neutral, left shoulder check, left mirror check, rear-view mirror check, and look-at-passenger experimental trials. ................. 75 Table 6.1 - Anthropomorphic data of participants in this study. ................................................................ 85 Table 6.2 - The median (1st, 3rd Quartile) pre-impact normalized RMS EMG activity for the muscles tested in both rear and frontal impact. ................................................................................................................. 94 x Table 6.3 - The median (1st, 3rd Quartile) peak normalized RMS EMG activity for the muscles tested in both rear and frontal impact. ..................................................................................................................... 95 Table 6.4 - The median (1st, 3rd quartile) onset of RMS EMG activity and select kinematic measures in both rear and frontal impact. ..................................................................................................................... 96 Table 6.5 - The median (1st, 3rd quartile) peak kinematic variables in both rear and frontal impact. ........ 97 Table 6.6 - The median (1st, 3rd quartile) timing of peak kinematic variables in both rear and frontal impact. ........................................................................................................................................................ 98 Table 6.7 - Summary of hypothesis testing for braced versus relaxed trials in both rear and frontal impact. ...................................................................................................................................................... 104 Table A.1 - Chapter 2: Magnitude of the resultant moment produced by each subject for all of the directions tested. ...................................................................................................................................... 144 Table A.2 - Chapter 2: Normalized 3D moments produced by each subject for all of the directions tested. .................................................................................................................................................................. 145 xi List of Figures Figure 1.1 - Frontal-lateral view of the cervical spine. ................................................................................. 2 Figure 1.2 - Neck MRI transverse plane cross-section .................................................................................. 3 Figure 1.3 - A schematic of the progression of head/neck kinematics during an automotive rear impact. 9 Figure 1.4 - The Hill muscle model. ............................................................................................................. 11 Figure 2.1 - The experimental set-up (A) showing the subject’s head fixed to a 6-axis load cell with a modified helmet, and the visual feedback shown to the subjects during a flexion + right axial rotation MVIC task (B). .............................................................................................................................................. 16 Figure 2.2 - Normalized 3D MVICs in an isometric view shown with a unit sphere (A), and cross sections taken in the flexion / lateral bending plane (B), the axial rotation / lateral bending plane (C) , and the axial rotation / flexion plane (D). ................................................................................................................ 21 Figure 2.3 - Predicted 3D MVICs compared to experimentally measured values for each subject and non-principal axis direction. ............................................................................................................................... 22 Figure 3.1 - The experimental set-up showing the subject seated with their torso constrained and their head fixed to a six-axis load cell through a tightly fitting modified helmet (A). ......................................... 37 Figure 3.2 - Exemplar filtered EMG for one repetition in 26 target directions at 15% MVIC for the sternocleidomastoid (SCM), splenius capitis (SPL) and semispinalis capitis (SSC) muscles. ...................... 38 Figure 3.3 - Spatial tuning plots for the 15% MVIC contraction of the SCM, SPL, and SSC muscles showing the voluntary preferred directions (black) and the electrically stimulated directions (grey) (n=8 subjects). .................................................................................................................................................................... 39 Figure 3.4 - Exemplar data from one subject showing the moments generated by electrical stimulation in the sternocleidomastoid (top), splenius capitis (middle), semispinalis capitis (bottom). .......................... 40 Figure 3.5 - Intra-subject variability of the electrically stimulated direction and voluntary preferred direction as shown by the standard deviation ellipse from a Kent distribution (A). .................................. 41 Figure 4.1 - Custom MATLAB user interface used to manually extract movements from the head position data. ............................................................................................................................................................ 54 Figure 4.2 - The number of movements recorded for each subject and movement type for when the vehicle was stationary (triangle), moving (square), and both (circle). ....................................................... 55 Figure 4.3 - Mean peak head angles in yaw, pitch, and roll for each subject and movement type. .......... 56 Figure 4.4 - Mean peak head angles in yaw for a moving vs stationary vehicle. ........................................ 57 Figure 5.1 - Experimental setup with a subject seated in a 2005 Volvo S40 drivers seat mounted to a feedback-controlled sled. ........................................................................................................................... 64 xii Figure 5.2 - Exemplar traces for electromyographic and triaxial kinematic data from a single subject during rear impact with their head in the neutral posture or postures mimicking a left shoulder check, left mirror check, rear-view mirror check, or talking to their passenger (right side). ................................ 76 Figure 5.3 - Median normalized pre-impact and peak RMS EMG of the eight muscles studied while participant’s heads were in the neutral posture or postures mimicking a left shoulder check, left mirror check, rear-view mirror check, or looking at the passenger (right side) during rear impact. .................... 77 Figure 5.4 - Median peak values of kinematic measures during rear impact while participant’s heads were in the neutral posture or postures mimicking a left shoulder check, left mirror check, rear-view mirror check, or looking at the passenger (right side). ............................................................................... 78 Figure 6.1 - Experimental setup with subject seated in a 2005 Volvo S40 drivers seat mounted to a feedback-controlled sled. ........................................................................................................................... 86 Figure 6.2 - Exemplar traces for EMG and kinematic data from a single subject in rear impact for hands in lap, relaxed and braced experimental conditions. ..................................................................................... 99 Figure 6.3 - Exemplar traces for EMG and kinematic data from a single subject in frontal impact for hands in lap, relaxed and braced experimental conditions. ..................................................................... 100 Figure 6.4 - Normalized RMS EMG pre-impact and peak during impact for both rear and frontal impacts of the eight muscles studied. .................................................................................................................... 101 Figure 6.5 - Peak and the timing of peaks for kinematic measures in both rear and frontal impact. ...... 102 Figure A.1 - Chapter 5 X-axis mean kinematic subject data shown with ± standard deviation response corridors for neutral, left shoulder check, left mirror check, rear-view mirror check, and looking at passenger (right side) experimental conditions. ...................................................................................... 146 Figure A.2 - Chapter 5 Y-axis mean kinematic subject data shown with ± standard deviation response corridors for neutral, left shoulder check, left mirror check, rear-view mirror check, and looking at passenger (right side) experimental conditions. ...................................................................................... 147 Figure A.3 - Chapter 5 Z-axis mean kinematic subject data shown with ± standard deviation response corridors for neutral, left shoulder check, left mirror check, rear-view mirror check, and looking at passenger (right side) experimental conditions. ...................................................................................... 148 Figure A.4 - Chapter 6 rear impact kinematic subject data shown with response corridors for control, relaxed and braced experimental conditions. .......................................................................................... 149 Figure A.5 - Chapter 6 frontal impact kinematic subject data shown with response corridors for hands in lap, relaxed and braced experimental conditions. ................................................................................... 150 xiii Acknowledgements I would like to thank a number of individuals and organizations for their support during my doctoral research. Firstly, thank you to my supervisors Jean-Sébastien Blouin and Gunter Siegmund. These men have both a steadfast commitment to the scientific method and an awe-inspiring work ethic that I will aspire to for the duration of my career. Jean-Sébastien and Gunter provided me with endless guidance and were committed to teaching me about neurophysiology and cultivating my interest in the human body beyond biomechanics. I also want to thank Peter Cripton for rounding out my PhD committee and asking difficult questions which led to a depth of understanding that I will always be grateful for. Vital help and support for this work was provided by MEA Forensic Engineers and Scientists in the form of matching contribution to my MITACS grant plus lab space and equipment. I want to thank MEA employees Jeff Nickel and Mircea Oala-Florescu for their guidance and assistance in designing and building the electrical and mechanical components required for the experiments of this dissertation. Funding for this work was provided by the Canadian Graduate Scholarship from NSERC, a four-year fellowship from the University of British Columbia, and a MITACS accelerate research internship. Further, travel was supported by a Killam fellowship travel and research allowance. I am thankful for the faith that was put in me and the opportunities that faith provided. Thanks to all my lab mates for the camaraderie and guidance during my doctorate research. In particular, I want to thank Daniel Mang for his friendship and assistance in using the perturbation sled, Optotrak system, and related instrumentation. xiv Dedication I dedicate this dissertation to my loving wife Amanda. Without her patience, emotional support, and critical mind, I could not have completed this work. Thank you. 1 Chapter 1. Introduction 1.1 Motivation for Research Whiplash associated disorder (WAD) or soft tissue cervical spine sprains/strains are the most common injury in motor vehicle collisions, representing 21 to 28% of all injuries (Quinlan et al., 2004; Styrke et al., 2012). WAD can reduce the victim’s quality of life for a long time, as 24% of whiplash victims experience symptoms one year after the accident, and 18% after two years (Radanov et al., 1995). Developing tools for the design of automotive safety systems to reduce the potential of WAD are computational human body models (Fice et al., 2011; Östh et al., 2015; Schap et al., 2019). Contemporary modelling efforts in the neck involve active musculature, but simplifications are required because detailed data of neck muscle activation during automobile collisions are not available or are unsuitable for use in models. In this dissertation, five experiments are presented that will improve our understanding of volunteers’ neck muscle activation during isometric contractions and whiplash-like sled perturbations that mimic automotive collisions. These experiments will also provide novel data for injury prevention and neck muscle modelling. 1.1 Background 1.1.1 Musculoskeletal anatomy of the neck A background discussion of musculoskeletal anatomy of the neck is helpful for the section on whiplash injuries later in this document. It will also highlight the complexity of the neck and lend some appreciation to the task faced by modellers who want to simulate neck injury with the aim of prevention in the future. The neck has an important role in the human body. It is responsible for stabilizing and moving the head, in which four of our five senses are located, in addition to our ‘sense’ of balance. The neck has seven cervical vertebrae (Figure 1.1), and in the lower and mid cervical spine (c-spine) intervertebral discs form a fibrocartilaginous joint between vertebral bodies, whose motion is guided and constrained by at least five ligaments per spinal level and by the facet joints. The upper cervical spine is formed by two unique vertebrae (Figure 1.1), the atlas, which is shaped like a ring with two lateral masses connected by an anterior and posterior arch, and the axis, which is shaped similar to lower c-spine vertebrae, but the superior surface of the vertebral body is extended with the odontoid process or dens. A large portion of flexion/extension of the head comes from the joints between the occipital condyles of the skull and the 2 superior surface of the atlas’ lateral masses and the majority of axial rotation of the head comes from pivoting of the atlas around the odontoid process of the axis (Anderst et al., 2015; Bogduk & Mercer, 2000). Both of these joints are stabilized by a unique set of ligaments and by the sub-occipital neck muscles. Figure 1.1 - Frontal-lateral view of the cervical spine. Reproduced with added labels from work by DrJanaOfficial, CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0)], from Wikimedia Commons. The musculoskeletal anatomy of the neck is complex: over twenty five pairs of muscles (Knaub & Myers, 1998; Figure 1.2) are arranged in several layers, are mostly symmetric about the mid-sagittal plane, have multi-segmented insertions, and have lines of action that generally do not lie in the principal anatomical planes (Vasavada et al., 1998). In the interest of brevity, only the eight muscle pairs that were the focus of this dissertation will be discussed. Starting with the sternohyoid muscle, which is a thin band that originates on the sternum and inserts on the hyoid bone. Despite its relatively small physiological cross sectional area (PCSA) of 0.6cm2 (Knaub & Myers, 1998), it has a large moment arm from the centre of intervertebral joint rotations and thus has been thought to contribute to the flexion strength of the neck (Knaub & Myers, 1998). Next is the sternocleidomastoid which is composed of two sub-volumes and it is the largest neck flexor with a PCSA of 4.9cm2 (Knaub & Myers, 1998). The cleideomastoid sub-volume originates on the clavicle and the sternomastoid sub-volume originates at the sternum, with both inserting on the mastoid process of the skull. The levator scapula originates at the scapula and inserts on the spinous process of the upper cervical spine, and, as the name suggests, is primarily thought to contribute to elevation of the scapula. The multifidus is a deep muscle which originates on the articular process of seventh through fourth vertebrae and inserts on the spinous process of the vertebra one or Odontoid Process Atlas Axis Vertebral Body of 3rd Cervical Vertebrae Intervertebral Disc Facet Joint 3 two levels up, terminating at the axis. As will be discussed further in later sections, the multifidus muscle has also been shown to have direct attachments on the capsular ligament (J. S. Anderson et al., 2005), which opens the possibility it has a role in WAD (Mang et al., 2015; Siegmund, Blouin, et al., 2008). The semispinalis cervicis originate in the upper thoracic spine and insert on the cervical vertebra, terminating at the third cervical vertebra. The semispinalis capitis, and the splenius capitis muscles are large posterior neck muscles which originate in the upper thoracic spine and insert at every vertebral level up to the occipital bone. Finally, the trapezius muscle, which is large and has a flat trapezoidal shape, originates at the clavicle and scapula and inserts onto the occipital bone and the spinous process of most vertebral levels in the thoracic and cervical spine. Figure 1.2 - Neck MRI transverse plane cross-section 1.1.2 Neck neuromuscular physiology Several reflexes have the potential to influence neck muscle activity during an automotive impact including the cervicocollic, vestibulocollic, optocollic, and the startle reflexes. In this section, neck muscle physiology as it relates to reflexive activity and its potential role in WAD will be discussed. Muscle spindles in the neck, which code for length and change of length, are more numerous compared to other parts of the body, and are arranged in complex and dense conjunctive arrays (Boyd-Clark et al., 2002; Chan et al., 1987; Richmond & Abrahams, 1975). Despite a multitude of stretch receptors in the neck, different researchers have either been in support of the presence of a monosynaptic stretch reflex Sternocleidomastoid Trapezius Semispinalis Capitis Multifidus Semispinalis Cervicis Levator Scapulae Splenius Capitis Sternohyoid Vertebral Body or Intervertebral Disc Facet Joint Spinal Cord Scalene Group 4 (Alexander & Harrison, 2002; M. E. Anderson, 1977; Rapoport, 1979) or against (Abrahams, 1977). Muscle spindles play a role in the cervicocollic reflex (CCR), which activates muscles to resist them being stretched and keeps the head fixed to the body. During an anterior perturbation of the body (i.e. rear-impact) the head rotates in extension stretching the anterior neck muscle spindles and potentially eliciting the CCR to activate flexor muscles to resist this stretching. The vestibulocollic reflex (VCR) activates neck muscles in order to keep the head fixed in space (Goldberg & Cullen, 2011). The vestibular system, located in the inner ear contains three approximately orthogonal semicircular canals with hair cell afferents that fire when head angular acceleration initiates fluid movement and two otolith organs that respond to linear acceleration of the head and the head’s orientation with respect to gravity (Fitzpatrick & Day, 2004). The CCR has been isolated from the VCR by either rotating the head and body together (VCR) or rotating the body while the head is fixed (CCR) and it was found that both had similar neck muscle activity gains suggesting similar strengths of the two reflexes (Peterson et al., 1985). In situations where a volunteer’s body was rotated under them with sinusoidal motions while their head was free, they likely acted to keep their head fixed in space and this resulted in a diminished role of the CCR (Keshner et al., 1999; Keshner & Peterson, 1995; Peng et al., 1996). In linear translations of the body (i.e. automotive impact) however, both the VCR and CCR would act to right the head in space. While the VCR may play a role in modulating neck muscle responses during an impact, it may not initiate the response. During perturbations of seated volunteers with either leg up rotations or anterior translations, neck muscle activity was elicited before head motion and the neck muscle activity patterns were consistent despite opposite direction head rotations, suggesting VCR did not initiate the neck muscle response (Forssberg & Hirschfeld, 1994). Forssberg & Hirchfeld (1994) suggested instead that somatosensory signals derived from the pelvis may trigger the neck muscle response. Another reflex that could contribute to stabilization of the head in automotive impact is the optocollic reflex (OCR), which activates neck muscles in response to visual stimuli to control head position and stabilize gaze during voluntary or perturbed body movements (Maurice et al., 2006). The presence of OCR in humans has been demonstrated in gaze shift experiments that have showed changes in the neck muscles responses as a result of changes in the visual stimulus (Gresty, 1974; Roucoux & Crommelinck, 1988). More specifically, the proportion of a gaze shift achieved with head vs. eye movement was significantly greater when elicited with a briefly flashed visual stimulus compared to a continuously 5 illuminated visual stimulus, suggesting the OCR plays a role in stabilizing gaze (Gresty, 1974). More work is needed to understand how this reflex potentially contributes in automotive collisions. Finally, the startle reflex is a whole body reaction to unexpected visual, auditory, or somatosensory stimuli that is strongly elicited in neck muscles (Brown et al., 1991). Auditory, vestibular, and somatosensory stimuli converge on large rapidly-conducting neurons in the nucleus reticularis pontis caudalis of the reticular formation before travelling to motor neurons (Shemmell, 2015; Yeomans & Frankland, 1995). Further, it has been suggested that during startle, EMG drive in the 10-20Hz band is associated with activity in the reticular formation (Grosse & Brown, 2003). The startle reflex is able to accelerate reaction times in voluntary head movement in both flexion and axial rotation (Nijhuis et al., 2007; Siegmund et al., 2001). It is recently becoming clear that startle responses are not stereotyped EMG patterns but are context dependent in a manner similar to voluntary responses (Shemmell, 2015; Valls-Solé et al., 2008). Startle has also been shown to be a part of the postural response to a simulated automotive rear impact (Blouin et al., 2006b). Mang et al. (2015, 2012) also showed that the startle response is a potentially harmful reaction of the body to an impact and may increase the risk of WAD. Further discussion on the aetiology of whiplash injuries will be continued in the next section. 1.1.3 Whiplash injury Many areas of the neck and cervical spine have been implicated in neck pain or WAD including the cervical facet joints, spinal ligaments, intervertebral discs, vertebral arteries, nerve roots, and finally neck muscles (reviewed in both Curatolo et al., 2011; and Siegmund et al., 2009). Using joint neural blocks and double blind study designs, the cervical facet joints have been shown to be the source of pain in 54 to 60% of chronic whiplash patients (Barnsley et al., 1995; Lord et al., 1996) and 60% of the general population with chronic neck pain (Manchikanti et al., 2002). Pinching of the synovial fold and excessive capsular ligament strain have both been suggested as a source of facet joint pain. Pinching of the synovial fold may be caused by a dynamic change of the vertebrae’s normal axis of rotation during extension (Kaneoka et al., 1999). This change of rotation axis leads to contact between the posterior edge of the superior facet and the inferior facet surface, which may trap the synovial fold (Yoganandan et al., 2002). The synovial fold contains nociceptive nerve endings (Inami et al., 2001), but it is not clear if the fold is loose enough to be pinched by the facet surfaces. Intervertebral flexion or extension can cause excessive strain in the capsular ligaments, and shear can magnify this effect (Siegmund, Myers, et al., 2001; Winkelstein et al., 2000). Animal models have shown that the sub-catastrophic failure strains in the capsular ligaments during whiplash have been linked to sustained pain (Cavanaugh, 2000; Lee et 6 al., 2006, 2004; Lu et al., 2005; Quinn & Winkelstein, 2007). Cadaveric cervical spine studies have suggested that sub-catastrophic tissue damage to the capsular ligament is possible in humans during low speed rear impacts (B. Deng et al., 2000; Ivancic et al., 2008; Panjabi et al., 1998; Pearson et al., 2004). Whiplash-like loads applied to isolated cadaveric cervical spines resulted in reduced ligament tensile load at failure when compared to control spines, suggesting that the ligaments sustained damage (Tominaga, Ndu, et al., 2006). Tominaga et al., (2006) grouped the results from the anterior longitudinal ligaments, the middle third disc, posterior longitudinal ligament, capsular ligament, ligamentum flavum, and the interstitial ligament for their whiplash vs. control ligament comparison and hence did not comment on which ligaments specifically were injured. Similarly, the inter-segmental flexibility of cadaveric cervical spines increased in their whiplash vs. control specimens, which signifies altered mechanical properties of the intervertebral discs, capsular ligaments, and/or other ligaments, after exposure to frontal impact accelerations of 8-10g to the T1 vertebra (Pearson et al., 2005). During these impacts, it was shown that impacts as low as 4g would cause physiologic strains on the ligament to be exceeded, which is the most conservative threshold for ligament injury available (Panjabi, Pearson, et al., 2004). In the upper cervical spine, MRI studies have suggested that high intensity signal variations in the alar and transverse ligament were associated with WAD (Kaale et al., 2005a, 2005b; Krakenes & Kaale, 2006). Although, more recent meta-analysis of case-control studies suggests that MRI signal fluctuations in the alar and transverse ligaments are not correlated with WAD (Li et al., 2013). Some evidence has shown a high prevalence of intervertebral disc pathology in WAD patients, including herniated discs (Pettersson et al., 1997). However, more recent evidence in a 10 year follow up MRI study suggests that WAD patients and controls had similar levels of intervertebral disc pathology, despite the WAD group being more likely to experience neck pain (Matsumoto et al., 2010). Similarly, some studies have failed to identify MRI signal fluctuations consistent with disc pathology in WAD patients (Ulbrich et al., 2014). Isolated cervical spines exposed to frontal and rear-impact accelerations have shown that the disc may be a location of injury during car crashes because strains exceeded the physiological limit (Ito et al., 2005; Panjabi, Ito, et al., 2004). They also showed that the annular tissue of the disk is potentially injured at a lower impact levels for rear impact compared to frontal impact (3.5g vs. 6g) and that the spinal level with the highest risk for frontal impact is C2-C3, compared to C5-C6 for rear impact. Clinical findings in WAD patients during a 2-year follow-up found that 8 out of 12 disc 7 herniations occurred at the C4-C5 and C5-C6 levels, but only 1 herniation at C2-C3 (Pettersson et al., 1997). Occlusion of the vertebral artery may occur during whiplash due to non-physiological elongation or pinching, and can cause symptoms such as headache, dizziness, and vertigo (Reddy et al., 2002; Šerić et al., 2000). Using a cadaveric cervical spine model, no vertebral artery elongation was seen beyond physiological limits in neutral posture frontal and rear impacts, but in side impact starting at 6.5g and a head-turned rear impact at 5g physiologic elongation was exceeded (Carlson et al., 2007; Ivancic et al., 2006). Thus, based on these cadaveric studies, the vertebral artery is potentially exposed to injurious elongation during certain automotive impacts. Damage to the dorsal nerve roots or dorsal root ganglion would cause radicular pain that would be felt not only in the neck, but also other areas of the body such as the head, shoulders, or back (Bogduk, 2002). Two mechanisms for damage to cervical nerve tissue during whiplash include pressure gradients caused by rapid deformations of the spinal canal and narrowing of the intervertebral foramen that the nerve roots pass through. Pressures that were shown to cause histopathological changes in the nerve tissue in pigs were exceeded in cadaver rear impacts, suggesting that dorsal root ganglion damage might occur during whiplash (Eichberger et al., 2000; Svensson et al., 2000; Yao et al., 2018). In humans, cadaveric cervical spines exposed to whiplash loads were shown to have a maximum vertebral foramen narrowing of 1.8-2.7mm during rear impacts and the lower cervical spine was at the highest risk of injury, but no effort was made to determine what level of narrowing is injurious (Ivancic, 2012; Panjabi, Maak, et al., 2006; Tominaga, Maak, et al., 2006). Muscle tissue can be injured when undergoing eccentric-contraction, which is muscle lengthening during activation (Garrett, 1996; Proske & Allen, 2005). The sternocleidomastoid may be exposed to eccentric contraction during a rear impact as it has been shown to be active while the neck was going into extension during a rear-end impact (Brault et al., 2000). Human kinematic data from sled rear-impacts has been imposed on a biomechanical model of the head and neck to show that the sternocleidomastoid, splenius capitis, semispinalis capitis, and trapezius all experienced potentially injurious eccentric contractions during different phases of the impact (Vasavada et al., 2007). However, measurements of serum creatine kinase levels, a marker for muscle injury, in WAD patients showed that only two of twenty-five subjects had increased levels within 24 hours after impact and neither of these subjects had symptoms beyond three months (S. Scott & Sanderson, 2002). The lack of evidence of long-8 term muscle related symptoms suggest that, although direct muscle injury is possible in whiplash, the muscle heals quickly and may not be a direct cause of chronic symptoms. Neck muscles may still play an important role in WAD by altering the magnitude and distribution of forces in the cervical spine and through post-injury altered activation patterns that may prolong chronic pain. Activation of the multifidus muscle during an impact has been theorized to increase facet capsule strain, a known source of whiplash pain (Lord et al., 1996), because of its insertion directly onto the capsular ligaments (J. S. Anderson et al., 2005; Winkelstein et al., 2001) and its activity during rear impact (Siegmund, Blouin, et al., 2008). In addition, altered neck muscle activation after injury may lead to prolonged pain as activation patterns have been shown to differ between chronic WAD patients and non-injured controls (Daenen et al., 2013; Falla et al., 2004; Juul-Kristensen et al., 2013). Further, muscle atrophy as measured by fatty degeneration of neck muscles observed during MRI scans, has been shown in deep neck extensors in WAD patients at much higher levels when compared to controls (Abbott et al., 2015; Elliott, 2011; Elliott et al., 2006). However, it remains unclear, whether altered muscle activation is a cause or effect of neck pain. 1.1.4 Volunteer impact research Rear-impact perturbations (i.e. forward translations) on human volunteers are an important tool to study WAD mechanisms. These studies have revealed the kinematics of the occupant’s head/neck and intervertebral kinematics during a rear-impact. In the early stage of the impact, the seat back engages the subject’s torso and the first thoracic vertebrae (T1) begins to move upward and rotate in extension, which is potentially due to straightening of the kyphotic curvature of the thoracic spine or ramping of the torso up the seat back (Davidsson et al., 2001; Ono et al., 1997; van den Kroonenberg et al., 1998). In one test series, the upward movement of the T1 before significant head motion resulted in a calculated compressive upper neck axial force of approximately 50N (van den Kroonenberg et al., 1998). In the next stage the torso is accelerated forward, and the inertia of the head causes it to lag behind without rotating, which is referred to as retraction. Retraction forces the cervical spine into an S-shape configuration (Figure 1.3), with extension in the lower cervical spine and flexion in the upper cervical spine (Grauer et al., 1997; Ono et al., 1997). Next, the head is rotated into extension by forces transferred through the neck to the base of the skull, placing the cervical spine into extension at all levels (C-shape spine; Figure 1.3). Head restraint contact generally occurs somewhere between the S-shape and C-shape phases of the head/neck motion. As energy absorbed by the seat during the impact is transferred back to the occupant, their trunk, head and neck will rotate into flexion, which is called 9 the rebound phase. Volunteer impact studies ruled out the popular theory that the mechanism for whiplash injury was hyperextension of the spine (Ono et al., 1997), as the highest intervertebral rotations occurred early in the impact before peak head extension. Figure 1.3 - A schematic of the progression of head/neck kinematics during an automotive rear impact. As the T1 accelerates forward, inertia causes the head to leg behind leading to flexion in the upper cervical spine and extension in the lower spine which is known as the S-shape phase (shown around 50ms). As the head begins to rotate the whole spine goes into extension which is know as the C-shape phase (shown around 100ms). This schematic shows computational modeling results from Fice et al., 2011. Early studies suggested that neck muscle activation could not influence head kinematics unless the occupant pre-activated their muscle before the impact began, because peak muscle activity could not be generated until ~200ms after impact (Ono et al., 1997). This perspective that muscle activity does not influence head kinematics neglects the possibility that muscle activity can influence head kinematics before reaching maximum force, and that muscle activity in the SCM occurs as early as 50ms after impact (Blouin et al., 2006a; Brault et al., 2000; Szabo & Welcher, 1996). Additionally, reducing the peak neck muscle response of volunteers, through the use of a pre-impact loud tone to attenuate the impact-related startle response, has been shown to reduce retraction, linear and angular head accelerations (Mang et al., 2014, 2015, 2012). Thus, establishing a link between muscle activation, kinematics, and potentially whiplash injury risk. Modelling efforts also consistently show that muscle activation can influence head kinematics during impacts (Brolin et al., 2005; Fice et al., 2011). It is possible that early studies under-estimated neck muscle contribution during impacts because subjects were exposed to repeated perturbations and habituation led to diminishing responses, as has been shown to occur in the SCM and neck paraspinal muscles in as little as three trials (Blouin et al., 2003; Siegmund et al., 2003b). Further, being surprised by an impact compared to receiving a countdown has been shown in males to increase cervical paraspinal muscle activity by 260% and angular head acceleration 180% (Siegmund et al., 2003a). In the volunteer sled impact studies discussed thus far, the volunteers are generally looking straight ahead, are relaxed, and have their hands in their lap at the time of impact. Epidemiologic evidence has 10 shown increased rates of whiplash associated disorders for drivers that displayed real-world pre-impact behaviours such as a rotated head posture or bracing themselves before impact (Jakobsson et al., 2008, 2004). Due to the amount of time in the pre-impact phase (often ~1s is used), muscle activity and even voluntary contractions (in addition to reflexive ones) are important to predict occupant response in these scenarios. For example, it has been shown that when occupants tense (i.e. co-contract) their muscles before an impact their head and torso excursions are reduced (Ejima et al., 2012, 2008; Ono et al., 1997). Drivers have also been shown to brace themselves by pushing against the steering wheel and straightening their arms (Choi et al., 2005; Hault-Dubrulle et al., 2011). Bracing in this manner has been shown to increase peak torso accelerations (Kemper et al., 2014), and a model has suggested that injury risk changes from the torso to the extremities (Iwamoto et al., 2012). When occupants turned their head by 45 deg before a low speed rear impact, muscle activity in the contralateral sternocleidomastoid muscle increased (Kumar et al., 2005). 1.1.5 Muscle activation in human body models Computational models of the human body commonly use the Hill muscle model (Hill, 1938) to represent both the passive and active properties of human skeletal muscle. While not a perfect representation of muscle (Herzog, 2013), it has been a valuable tool due to its computational efficiency and ability to capture important features of force generation in human muscle. The Hill muscle model consists of two parallel spring elements that represent the passive (PE) and the contractile components (CE) of muscle tissue (Figure 1.4). A third spring element is sometimes placed in series with the contractile element to represent elastic properties of the tendon (SE), but this is often incorporated into the properties of the contractile element. The active force generated in the contractile element is a function of the maximum force that can be generated in the muscle (Fmax) which is then scaled by the activation state (a(t)), force-length relationship (fAL), and force-velocity relationship (fAV) of the active muscle (Equation 1.1). The physical constants required to build the hill muscle model have been experimentally measured and are available to model many muscles in the body including the neck (Fung, 1993; Winters, 1995; Winters & Stark, 1988; Winters & Woo, 1990). The activation state as a function of time is important to predict the forces generated by muscles during automotive impact, but this variable is context dependent and difficult to quantify. 11 Figure 1.4 - The Hill muscle model. Reproduced without modification from work by Rudolf.hellmuth [CC BY-SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0) or GFDL (http://www.gnu.org/copyleft/fdl.html)], from Wikimedia Commons. Equation 1.1 - Equation for the force of the contractile element (FCE) in the Hill muscle model. The maximum force generated by a muscle (Fmax) is scaled by the activation level as a function of time (a(t)), the force-length relationship (fAL) which is a function of current muscle length (LM), and the force-velocity relationship (fAV) which is a function of current muscle lengthening/shortening velocity (vM). a(t) has been difficult to define accurately in human body models. 𝐹𝐶𝐸 = 𝐹𝑚𝑎𝑥𝒂(𝒕)𝑓𝐴𝐿(𝐿𝑀)𝑓𝐴𝑉(𝑣𝑀) In the earliest computational models of the human body, muscle activation was generally ignored (De Jager et al., 1994; Y.-C. Deng & Goldsmith, 1987). In the first attempts to simulate muscle activity (Brolin et al., 2005; Stemper et al., 2006), the models used a(t) curves that were based on the principles of neural delays and the time course of a neural input turning into muscle force (Happee, 1994), but with a parametric study used to define many features (i.e. onset, deactivation, & shape). Brolin et al., (2005) found that muscle activity was important to reproduce the kinematics of volunteers in lateral impacts. In subsequent models (Choi et al., 2005; Cronin, 2014; Fice et al., 2011; Panzer et al., 2011), authors continued to base a(t) curves on the principles of neural delays and the time course of a neural input turning into muscle force (Happee, 1994), but attempted to extract additional variables from volunteer automobile crash-like perturbations. For example, in one model (Fice et al., 2011) muscle onset was derived from volunteer studies (Siegmund et al., 2003a) and the muscles were grouped into flexors and extensors based on presumed function and their relative activation was based on volunteer data (Brault et al., 2000; Siegmund et al., 2003a). Using the model in Fice et al., 2011, it was found that not only is muscle activation important to predict the head kinematics of volunteers during rear impact, but also the strains in the capsular ligament were influenced by muscle activity. Another approach is to optimize a(t) in human body models to match the kinematics of a volunteer experiment (Bose et al., 2010; Chancey et al., 2003; Dibb et al., 2013; Iwamoto et al., 2012). A unique 12 aspect of this approach was that no muscle groupings were needed. As with previous modelling efforts, this approach has shown how important muscle activation is to predict volunteer experiments, but this time for children (Dibb et al., 2013). Recent developments have led researchers away from using feed-forward predefined muscle a(t) curves and towards muscle controllers that close the feedback loop and control a(t) based on joint angles or muscle length feedback (Iwamoto & Nakahira, 2015; Meijer et al., 2013; Östh et al., 2015, 2012). Feedback control is an important part of human neuromuscular control, and thus far the efforts to model feedback control have focused on the proportional, integral and derivative (PID) controller (Östh et al., 2012). In this approach, the activation levels used in different muscle groups were determined using a PID controller that acted based on the current angle of the head or neck. The feedback loop contained neural delays and a model of activation dynamics. Controller gains were fitted to the experimental results using either an iterative approach (Östh et al., 2012), or an optimization approach (Östh et al., 2015). Currently, several simplifications are utilized to model neck muscles including splitting muscles into coarse groups (i.e. flexor/extensors) and optimizing muscle activation timing and magnitudes to volunteer kinematics (Dibb et al., 2013; Fice et al., 2011; Östh et al., 2015). Splitting muscles into course groups may not accurately reflect the contributions of particular muscles, and the groupings are not trivial when models are used in situations that lead to kinematics outside the sagittal plane. Optimizing muscle activation characteristics to a specific human kinematic dataset potentially limits the predictive capability of the model in novel situations. Further, without validating to human muscle activation data, muscles can potentially mask other deficits in the model. Part of the motivation of this work was to collect data will help researchers developing human body models to understand neck muscle activation during automobile collisions and avoid over simplifying their models. More specifically, human volunteer data is lacking to validate or define neck muscle activation schemes for drivers that are in rotated head postures or bracing themselves before a collision. These load cases have shown to be important from an injury risk perspective (Jakobsson et al., 2008, 2004). 1.2 Objective & Goals The overall goal of this dissertation was to understand how occupant behaviours before a collision, including non-neutral head postures and bracing against the steering wheel, influenced neck muscle activity and head/neck kinematics during impact. An extension of this goal was to provide data and 13 knowledge that can be applied to computational models of the head and neck to prevent neck injuries in automobile collisions. The goals of this dissertation were explored in a series of five experiments which asked the following questions: i) How does the strength of the neck measured in the principal axis directions get scaled when producing off-axis moments? ii) Can the biomechanical line of action of a given neck muscle be used to predict how it will be activated? iii) What non-neutral head/neck postures do automobile drivers adopt during naturalistic driving? iv) How does being in a rotated non-neutral head/neck posture alter the muscle activity and head/neck kinematics of volunteers during sled tests that simulate automotive collisions? v) How are neck muscle responses and head/neck kinematics altered when volunteers braced against the steering wheel during sled tests that simulate automotive collisions? 14 Chapter 2. Prediction of three-dimensional maximum isometric neck strength 2.1 Preamble1 Currently data for validating the maximum strength of head-neck models is limited to the principal axis directions. It is important that neck models reproduce off-axis strength well to be able to simulate real-life situations such as omni-directional impacts and occupants in non-neutral postures. In this experiment, we measured the 3D neck strength of volunteers in 17 directions including the principal axes and off-axis directions. The non-neutral head postures that drivers adopt in naturalistic driving and the neck muscle responses of occupants in rear impacts while in these non-neutral postures will be presented in Chapters 4 and 5 of this dissertation. 2.2 Introduction The musculoskeletal anatomy of the neck is complex: over twenty pairs of muscles are arranged in several layers, have multi-segmented insertions, and have lines of action that generally do not lie in the principal anatomical planes (Vasavada et al., 2001). Since characterising maximum voluntary isometric contractions (MVIC) for individual neck muscles is not possible because volunteers cannot selectively activate individual muscles, researchers have instead quantified the combined maximum isometric strength of all neck muscles about the principal axes, i.e., flexion/extension about the mediolateral axis, lateral bending about the anteroposterior axis, and axial rotation about the inferosuperior or vertical axis. The largest neck moments about the first thoracic vertebrae are generated in extension, the smallest moments are generated in axial rotation, and intermediate moments are generated in flexion and lateral bending (Cagnie et al., 2007; Jordan et al., 1999; Mayoux-Benhamou et al., 1989; Portero et al., 2001; Queisser et al., 1994; Vasavada et al., 1998). Maximum neck moments in combined principal directions in the horizontal plane, i.e., two-dimensional (2D) moments, are smaller than the vector sum of the maximum moments generated in pure flexion/extension and lateral bending (Gabriel et al., 2004; Kumar et al., 2001). These observations suggest that the generation of 2D moments in the horizontal- 1 This chapter is a post-peer-review, pre-copyedit version of an article published in Annals of Biomedical Engineering. The final authenticated version is available online at: http://dx.doi.org/10.1007/s10439-014-1046-0. The full citation is as follows: Fice, J. B., Siegmund, G. P., & Blouin, J. S. (2014). Prediction of three-dimensional maximum isometric neck strength. Ann Biomed Eng., 42(9), 1846-1852. 15 plane relies on the neck muscles functioning together as a single system rather than distinct muscle groups functioning separately in each principal direction. To date, it remains unclear whether a similar pattern of maximum neck strength extends to the generation of three-dimensional (3D) neck moments. If this pattern remains consistent across subjects, it may be possible to predict 3D neck strength in any direction using only neck strength measurements about the principal axes. In the present study we sought to characterize how the magnitudes of maximum neck moments vary in 3D. Based on the pattern of 2D moments observed in prior studies (Gabriel et al., 2004; Kumar et al., 2001) we postulated that the maximum moment that can be generated in an arbitrary 3D direction would be a direction-dependent weighted sum of the maximum principal-axis moments. More specifically, we hypothesized that for a MVIC in any 3D direction, the resultant vector of the three moment components divided by (i.e., normalized by) their corresponding principal axis MVIC would have a length of one. If this hypothesis is correct, or another consistent pattern of MVICs across subjects emerges, then maximum neck strength could be predicted in any 3D direction using only neck strength data about the principal axes. 2.3 Methods 2.3.1 Subjects Nine male subjects (Table 2.1) with no history of whiplash injury, neck/back pain, frequent/severe headaches, or neuromuscular injury participated in the study. Subjects provided written informed consent prior to participating in the study, which was approved by the UBC Clinical Research Ethics Board and conformed to the Declaration of Helsinki. Table 2.1 - Anthropomorphic data of the male subjects in this study. Volunteer Age (years) Height (cm) Weight (kg) 1 27 178 76 2 32 171 62 3 21 173 62 4 21 170 70 5 24 183 84 6 38 165 76 7 25 183 74 8 27 180 86 9 49 184 77 Mean (SD) 29.3 (9.1) 176.3 (6.8) 74.1 (8.4) 16 2.3.2 Procedures and Instrumentation Subjects sat upright against a foam-covered backboard with their shoulders firmly strapped against the board and their hips and knees flexed (Figure 2.1A). Their head was fixed to a 6-axis load cell (Model 45E15, JR3 Incorporated, Woodland, CA) using a tight-fitting helmet modified to include two rubber pads that pressed firmly against the zygomatic arches bilaterally. Subjects were instructed to relax and look straight ahead before the helmet position was fixed for the duration of the experiment. Figure 2.1 - The experimental set-up (A) showing the subject’s head fixed to a 6-axis load cell with a modified helmet, and the visual feedback shown to the subjects during a flexion + right axial rotation MVIC task (B). Note that the dial needle rotated to indicate the subject’s current axial rotation moment, and the origin of the needle moved to indicate their horizontal plane moments (positive values for flexion and right lateral bending). The white wedge indicated the acceptable range of axial moments and the grey shaded area the tolerance for horizontal plane moments. Neck moments were resolved to the joint axis of the seventh cervical and first thoracic vertebrae (C7-T1) by calculating the reaction moments needed at the C7-T1 joint axis to balance the forces and moments measured at the load cell. The C7-T1 joint axis was defined as the midpoint between the sternal notch and the C7 spinous process (Queisser et al., 1994), and this was digitized relative to the load cell with a three dimensional localizer (Polaris Vicra, Northern Digital Inc., Waterloo, ON, Canada). Load cell data were collected at 2000Hz with a data acquisition system (PXI-6289, National Instruments, Austin, TX) and custom LabVIEW software (National Instruments, Austin, TX). Subjects performed MVICs in 17 predefined directions that included contractions about the principal axes, i.e., flexion, extension, right lateral bending, left/right axial rotation, and various 2D and 3D combinations of these directions (Table 2.2). Subjects performed two trials in each direction (34 trials in total). Based on an assumption of symmetry, directions with a left lateral bending component were not 17 investigated to reduce the number of trials for each subject. Not investigating left lateral bending reduced the number of trials from 52 (eight horizontal plane directions, at three levels of axial rotation, plus left/right axial rotation, with two repeats) down to 34 (five horizontal directions, at three levels of axial rotation, plus left/right axial rotation, with two repeats). When combining moments in the horizontal plane (2D moments), subjects exerted a net moment in a direction that had equal components in both axes. For directions that included axial rotation (3D moments), the subjects generated an axial rotation moment that was 30% of their moment in the horizontal plane, which was based on the ratio of MVIC moments in axial rotation (15 Nm) and extension (50 Nm) reported previously (Vasavada et al., 2001). The directions were presented in random order, each trial took ~5-15 s with a short duration (~1-3 s) at maximum effort, and subjects rested two minutes between trials. Table 2.2 - The direction cosines for the 17 predefined MVIC directions. Note the direction cosines are reported as (flexion/extension, lateral bending, axial rotation). Positive values denote extension, right lateral bending, and right axial rotation. Components in the horizontal plane were always equal, and the axial rotation was 30% of the resultant in the horizontal plane. Horizontal Plane Moments Left Axial Rotation Zero Axial Rotation Right Axial Rotation Flexion (-0.958, 0, -0.287) (-1, 0, 0) (-0.958, 0, 0.287) Flexion + Right Lateral Bending (-0.677, 0.677, -0.287) (-0.707, 0.707, 0) (-0.677, 0.677, 0.287) Right Lateral Bending (0, 0.958, -0.287) (0, 1, 0) (0, 0.958, 0.287) Extension + Right Lateral Bending (0.677, 0.677, -0.287) (0.707, 0.707, 0) (0.677, 0.677, 0.287) Extension (0.958, 0, -0.287) (1, 0, 0) (0.958, 0, 0.287) No Horizontal Plane Moment (0, 0, -1) N/A (0, 0, 1) Real-time visual feedback was given to subjects of their moments in three dimensions (Figure 2.1B). Subjects were required to keep their moments within ±5 Nm of the prescribed direction in the horizontal plane and ±1.5 Nm in the axial rotation direction (see corridors in Figure 2.1B). Both tolerances values represent 10% of the expected MVIC values for extension and axial rotation (Vasavada et al., 2001). Subjects were presented with a green light when their moment direction was valid, i.e., each component was within tolerance. Verbal encouragement was not given during trials to avoid distraction from the visual feedback; however, subjects were reminded before each trial to produce their maximum voluntary effort. Subjects were familiarized with the experimental set-up, visual feedback, fitted to the helmet, and allowed to practice sub MVIC moments (~40-60% MVIC) in each of the 17 directions at least one day before the experiment. 18 2.3.3 Data Analysis The maximum valid resultant moment was determined for each pair of trials in each direction. For all directions within each subject, a paired t-test was used to determine whether there was a trial effect indicative of fatigue or learning. Due to the potential challenge of producing off-axis MVICs, a paired t-test was also performed for the 12 off-axis directions within each subject to determine if there was a trial effect in these directions. To check our assumption of symmetry, peak moments in pure axial rotation were compared across all subjects using a two-tailed paired t-test. A significance level of p<0.05 was used for both analyses. To evaluate our hypothesis, we performed two analyses. In the first analysis, we assessed whether the normalized moment was similar in all directions across all subjects. For each subject and direction (i), we calculated the normalized moment (Ni), i.e., the magnitude of the vector sum of each component (𝑀𝑓𝑙𝑒𝑥𝑖𝑜𝑛𝑖 or 𝑀𝑒𝑥𝑡𝑒𝑛𝑠𝑖𝑜𝑛𝑖 , 𝑀𝑙𝑎𝑡𝑒𝑟𝑎𝑙 𝑏𝑒𝑛𝑑𝑖𝑛𝑔𝑖 , 𝑀𝑎𝑥𝑖𝑎𝑙 𝑟𝑜𝑡𝑎𝑡𝑖𝑜𝑛𝑖 ) normalized by its corresponding maximum moment for that axis alone (𝑀𝑓𝑙𝑒𝑥𝑖𝑜𝑛𝑚𝑎𝑥 or 𝑀𝑒𝑥𝑡𝑒𝑛𝑠𝑖𝑜𝑛𝑚𝑎𝑥 , 𝑀𝑙𝑎𝑡𝑒𝑟𝑎𝑙 𝑏𝑒𝑛𝑑𝑖𝑛𝑔𝑚𝑎𝑥 , 𝑀𝑦𝑎𝑤𝑚𝑎𝑥) (Equation 2.1). This normalization is not a typical vector normalization, but rather an element-wise normalization performed prior to calculating the resultant magnitude of the normalized elements. The 95% confidence interval of the normalized moment in each direction were calculated by finding Ni for each subject in each direction and then finding the confidence intervals across subjects in each direction. These confidence intervals were examined to determine if they included one to test the hypothesis that Ni =1 for all directions. Equation 2.1 𝑁𝑖 = √(𝑀𝑓𝑙𝑒𝑥𝑖𝑜𝑛𝑖𝑀𝑓𝑙𝑒𝑥𝑖𝑜𝑛𝑚𝑎𝑥 𝑜𝑟 𝑀𝑒𝑥𝑡𝑒𝑛𝑠𝑖𝑜𝑛𝑖𝑀𝑒𝑥𝑡𝑒𝑛𝑠𝑖𝑜𝑛𝑚𝑎𝑥 )2+ (𝑀𝑙𝑎𝑡𝑒𝑟𝑎𝑙 𝑏𝑒𝑛𝑑𝑖𝑛𝑔𝑖𝑀𝑙𝑎𝑡𝑒𝑟𝑎𝑙 𝑏𝑒𝑛𝑑𝑖𝑛𝑔𝑚𝑎𝑥 )2+ (𝑀𝑎𝑥𝑖𝑎𝑙 𝑟𝑜𝑡𝑎𝑡𝑖𝑜𝑛𝑖𝑀𝑎𝑥𝑖𝑎𝑙 𝑟𝑜𝑡𝑎𝑡𝑖𝑜𝑛𝑚𝑎𝑥 )2 In the second analysis, we assessed whether the normalized moment calculated above could be used to predict each subject’s maximum 3D moments. To evaluate the predictive capability of Equation 2.1, we predicted the magnitude of moment each subject could produce in the non-principal-axis directions using their principal axis strength and compared these values to the experimentally measured values. To do so, the direction cosines of an arbitrary direction (xi, yi, zi) times the predicted moment (𝑀𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛𝑖 ) for that direction (i) were inserted into Equation 2.1 for the three moment components (Mi), and the equation was rearranged to solve for the predicted moment (Equation 2.2). The normalized moment was set to unity (Ni=1) for directions when unity was found within the 95% confidence interval previously. For directions in which the confidence interval did not include one, the moment prediction 19 for each subject used a normalized moment (Ni) that was set to the average computed for the other eight subjects. The principal axis strengths of each subject were used for Mmax, and the direction cosines (xi, yi, zi) were set according to the nominal directions requested for each contraction (Table 2.2). The coefficient of determination (r2) was calculated between the predicted and the measured non-principal-axis strengths to assess the quality of the prediction. Equation 2.2 𝑀𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛𝑖 =𝑁𝑖√(𝑥𝑖𝑀𝑓𝑙𝑒𝑥𝑖𝑜𝑛 𝑜𝑟 𝑒𝑥𝑡𝑒𝑛𝑠𝑖𝑜𝑛𝑚𝑎𝑥 )2+(𝑦𝑖𝑀𝑙𝑎𝑡𝑒𝑟𝑎𝑙 𝑏𝑒𝑛𝑑𝑖𝑛𝑔𝑚𝑎𝑥 )2+(𝑧𝑖𝑀𝑎𝑥𝑖𝑎𝑙 𝑟𝑜𝑡𝑎𝑡𝑖𝑜𝑛𝑚𝑎𝑥 )2 2.4 Results Peak resultant moments were generated in 47% of first trials and 53% of second trials across all subjects and directions. When only the off-axis directions were considered across all subjects, 48% of first trials and 52% of second trials produced the largest moment. When all directions were considered, eight of nine subjects showed no significant difference between their first and second trials, but one subject generated lower moments in their second trials (% difference -4.7%, t16=1.836, p=0.04). When only the off-axis directions were considered, seven of nine subjects had no difference between their first and second trials, but the same subject as above had lower second trials (% difference -8.7%, t11=3.54, p=0.002) and an additional subject was shown to exhibit higher moments during their second trials (% difference 9.7%, t11=-2.13, p=0.03). About the principal axes, the maximum moments about C7-T1 were 30±6 Nm for flexion, 32±9 Nm for lateral bending, and 51±11 Nm in extension (Table 2.3; Individual subject results in Appendix Table A.1 and Table A.2). There was no difference between maximum moments in left and right axial rotation (13±5 Nm, t8=0.0378, p=0.97). The 95% confidence intervals for the normalized moments contained one for all but the three directions that combined both right lateral bending and right axial rotation (Table 2.3; Figure 2.2). The normalized moment of these three directions all exceeded one (averages of 1.15 to 1.49). The predicted versus actual resultant non-principal-axis moments for each subject had a coefficient of determination (r2) equal to 0.88 (Figure 2.3). 20 Table 2.3 - Confidence intervals for the normalized moment, experimental isometric neck strength, and resultant moments predicted using the principal axis moments. * denotes values that differed from unity. Data reported in Nm, mean (S.D.). Horizontal Plane Moment Direction Axial Rotation Direction 95% Confidence Interval for Normalized Moment Measured Resultant Moment Predicted Resultant Moments Flexion Left 0.88 - 1.10 24 (5) 24 (6) None n/a 30 (6) n/a Right 0.85 - 1.02 23 (5) 24 (6) Flexion + R. Lateral Bending Left 0.90 - 1.09 24 (5) 24 (7) None 0.92 - 1.06 30 (6) 31 (7) Right 1.02 - 1.28* 29 (7) 28 (8) R. Lateral Bending Left 0.78 - 1.10 25 (8) 26 (9) None n/a 32 (9) n/a Right 1.20 - 1.74* 36 (11) 38 (13) Extension + R. Lateral Bending Left 0.78 - 1.05 28 (9) 28 (10) None 0.87 - 1.14 39 (11) 38 (10) Right 1.30 - 1.68* 41 (10) 42 (15) Extension Left 0.88 - 1.05 33 (11) 31 (11) None n/a 51 (11) n/a Right 0.87 - 1.14 34 (11) 31 (11) Pure Axial Rotation n/a 13 (5) n/a 21 Figure 2.2 - Normalized 3D MVICs in an isometric view shown with a unit sphere (A), and cross sections taken in the flexion / lateral bending plane (B), the axial rotation / lateral bending plane (C) , and the axial rotation / flexion plane (D). Gray shaded areas are the upper bound of the 95% confidence interval. Note that moments with a left lateral bending component were mirrored from the right lateral bending data. Lat Bend = lateral bending, R = right, L = left. The data showed a generally spherical normalized moment space that is elongated when ipsilateral axial rotation and lateral bending are combined. 22 Figure 2.3 - Predicted 3D MVICs compared to experimentally measured values for each subject and non-principal axis direction. A line showing a 1:1 relationship is shown for reference. Coefficient of determination (r2) was equal to 0.88. 2.5 Discussion Our goal was to determine whether maximum 3D neck moments could be predicted from measurements of maximum neck strength about the principal axes only. Due to the complex multi-layer architecture, insertions and lines of action of neck muscles, we hypothesised that the principal components of maximum 3D moments, if normalized by the corresponding maximum principal-axis moments, would sum to one in all directions (Equation 2.1). Across the 17 MVIC directions investigated here, the three directions combining right lateral bending and right axial rotation exhibited a normalized moment greater than unity. Based on these findings, we reject our hypothesis. Our findings indicate that adding ipsilateral axial rotation to lateral bending (or vice versa) does not attenuate the capacity of neck muscles to generate an ipsilateral lateral bending moment (or the ipsilateral axial rotation moment if applied in the reverse order). This was demonstrated by the normalized right lateral bending and right axial rotation, both of which were individually close to one when producing a MVIC in this combined direction (Figure 2.2C). The fact that when combined both components simultaneously were close to their principle axis MVICs suggests a degree of independence between muscles generating maximum moments in the ipsilateral axial rotation and lateral bending directions. An explanation of this apparent independence may lie in our observation that all subjects on their first attempt to produce a pure lateral bending moment also generated an ipsilateral axial rotation moment. This observation suggests that producing a pure lateral bending moment required subjects to activate neck muscles that did not contribute to the lateral bending moment but rather compensated 23 for the ipsilateral axial rotation moment. Thus, the addition of ipsilateral axial rotation to lateral bending did not require additional activation of the muscles generating an ipsilateral axial rotation moment, but rather a decrease in activation of the muscles generating the contralateral axial rotation moment. This pattern of muscle activation would result in a normalized moment greater than unity, and potentially explains our findings (Figure 2.2). Electromyography (EMG) was not recorded in this study because we did not want the presence of multiple indwelling electrodes to interfere with a subject’s maximum effort. Based on the lines of action of neck muscles, it is nevertheless possible to speculate that splenius capitis, levator scapulae, longis capitis, semispinalis capitis, oblique capitis, and posterior rectus muscles contribute to generating an ipsilateral axial rotation moment during unrestrained lateral bending with the head upright. On the other hand, only trapezius, sternocleidomastoid, and multifidus muscles likely contribute to contralateral axial rotation during lateral bending, again with the head upright. Therefore, activation of the neck muscles generating a lateral bending moment would likely lead to a net ipsilateral axial rotation moment. Further work including EMG is needed to explore the specific muscle contributions further. Although our hypothesis was invalidated for three of the twelve combined moment conditions we tested, it was valid in the remaining nine combined moment conditions. This finding suggests that, for most 2D and 3D moment directions, the neck muscles interact more simply than described in the two previous paragraphs. For instance, the transition from pure flexion, to combined flexion and axial rotation, to pure axial rotation can be explained by continuous changes in activation levels between different muscles. Activating the left and right flexor muscles generates pure flexion moments, whereas activating the ipsilateral flexor and contralateral extensor muscles generates pure axial rotation moments. During the transition, deactivating the contralateral flexor while activating the contralateral extensor decreases the flexor moment as the axial rotation moment increases. This continuous variation between pure flexion and axial moments explains why the normalized moment remains equal to unity. A similar continuous variation can be used to explain why the normalized moments equal one for the other combined directions. Had our hypothesis been true, it would have yielded a three-dimensional plot of normalized moments that was spherical. Instead, our data suggest an ellipsoid shape elongated in the plane combining ipsilateral axial rotation and lateral bending (Figure 2.2). Though not spherical, there was a consistent pattern of 3D moments across subjects for every direction. Therefore, our data can be used to predict 24 maximum 3D moments from a subject’s maximum moments about the principal axes (Figure 2.3). When the principal-axis moments were used to predict 3D neck strength, they compared well to the measured strengths (r2 = 0.88; Figure 2.3 & Table 2.3). This predictive ability is valuable to biomechanical modellers of the neck because it potentially allows them to validate their model’s force output across a large range of directions using only existing principal axis MVIC data. To be more useful for prediction, additional work is needed to better define the shape where lateral bending and ipsilateral axial rotation combine. The principal axis moments measured in this study compared well with the published literature (Table 2.4). Aside from extension strength (36 ± 8 Nm) measured in 48 males by Cagnie et al. (2007), our principal axis moments were within one standard deviation of those reported previously. Other studies of non-principal-axis MVICs in the horizontal plane reported their results as forces rather than moments (Gabriel et al., 2004; Kumar et al., 2001), and cannot be directly compared to our results. By assuming left and right symmetry, we reduced the number of trials from 52 to 34 for each subject. Given the possible evidence of fatigue in one subject, we believe this approach was sound. Our symmetry assumption was partially validated by our finding of no difference between pure left and right axial rotation moments. Others have also found neck strength to be symmetrical (Portero et al., 2001; Vasavada et al., 2001-males; Ylinen et al., 2004), although this finding is not universal (Gabriel et al., 2004; Vasavada et al., 1998-females). A symmetry assumption allowed our subjects to perform each MVIC twice while maintaining a reasonable total number of overall trials. Since peak moments were generated roughly equally between the first and second trials (47 / 53% respectively), we believe this repetition was worthwhile. A possible limitation of this study is that there could have been a non-plateaued learning effect in the more difficult off-axis directions, but we showed that only one subject had significantly larger second trial moments in these directions, so this is unlikely. Another potential limitation of this study was not giving the subjects verbal encouragement during the MVIC tasks, instead allowing them to focus on the visual feedback. Comparison of our principal axis MVIC data with previous reported literature suggests this did not result in reduced MVICs (Table 2.4). 25 Table 2.4 - Isometric neck strength reported in moments about the C7-T1 for studies that included male subjects in a neutral posture. Data reported in Nm, mean (S.D.) Subjects Flexion Lateral Bending Extension Axial Rotation Cagnie et al., 2007 48M 24 (6) 36 (8) Portero et al., 2001 7M 34 (4) Jordan et al., 1999 50M 30 (9) 55 (14) Mayoux-Benhamou et al., 1989 5M, 10F 53 (12) Queisser et al., 1994 12M 60 (9) Vasavada et al., 2002 11M 30 (5) 36 (8) 52 (11) 15 (4) Current study 9M 30 (6) 32 (9) 51 (11) 13 (5) In summary, we collected 3D neck strength data and found that for 9 of the 12 direction combinations measured, the moment normalized by the principal axis moments was equal to one. In the directions that combined ipsilateral axial rotation and lateral bending, the normalized moment exceeded one, but was consistent across subjects. This consistent pattern of normalized moments allows for the prediction of 3D neck strength in many directions using principal axis strength data only. The predictive nature of these MVIC data will be valuable to biomechanical modellers of the neck because it will allow them to validate the maximum strength of their models across a wide range of directions using only existing principal axis MVIC data. 26 Chapter 3. Neck muscle biomechanics and neural control 3.1 Preamble2 As neck muscle modeling advances towards a physiologically based muscle controller, it will be tempting to make assumptions such as the activation of a muscle can be predicted by its line of action. In this study we tested if such an assumption is valid. In Chapter 2, we defined neck MVICs in three dimensions as a function of principal axis MVICs, which provided a framework for describing 3D sub-MVIC contractions that were used to define the 3D spatial tuning task in the present chapter. Further, the relationships quantified here between neck muscle line of action and muscle activation help to inform the muscle activity results in the whiplash-like perturbation studies presented in Chapters 5 and 6. 3.2 Introduction The mechanics, morphometry, and geometry of our joints, segments and muscles are fundamental biomechanical properties intrinsic to human neural control. This concept applies equally to simple and complex musculoskeletal systems within the body, but in the neck, which includes over twenty-five muscle pairs with multi-joint insertions and complex lines of action, the relationship between biomechanics and neural control has been difficult to unravel. The question we pose here is whether the biomechanics (e.g. line of action and variability) of individual neck muscles are a useful indicator of how we voluntarily activate them to generate forces/moments in different directions. Spatial tuning curves have been used to explore how humans control and activate neck muscles. These tuning curves define a muscle’s preferred direction during voluntary isometric tasks (Blouin, Siegmund, Carpenter, et al., 2007; Keshner et al., 1989; Vasavada et al., 2002) and reflexive activity during seated perturbations (Ólafsdóttir et al., 2015). Apart from the splenius capitis muscle, consistent preferred directions were observed among volunteers for each muscle studied (Blouin, Siegmund, Carpenter, et al., 2007; Keshner et al., 1989). This consistency is perhaps surprising given that there are more actuators, i.e. neck muscles, than there are degrees of freedom of head movement, possibly making the head-neck a redundant system (Bernstein, 1967). The previously reported consistency in preferred 2 A version of this chapter has been previously published as Fice, J. B., Siegmund, G. P., & Blouin, J. S. (2018). Neck muscle biomechanics and neural control. J Neurophysiol., 120(1), 361–371. Available at https://doi.org/10.1152/jn.00512.2017. 27 directions suggests a common (or similar) neural control strategy amongst subjects for the active generation of neck forces and moments, which in turn suggests that voluntary and reflexive control of the neck muscles are based, at least in part, on biomechanical constraints. In human volunteers, the biomechanical function of individual neck muscles is difficult to characterize because humans cannot voluntarily activate isolated neck muscles. Instead, muscle morphometric measurements in cadavers (J. S. Anderson et al., 2005; Kamibayashi & Richmond, 1998) have been combined with a computational model of the head and neck (Vasavada et al., 1998) to infer the biomechanical function of neck muscles. A comparison of the biomechanical lines of action predicted by this computational model to the preferred directions derived from the spatial tuning curves of human subjects revealed differences of at least 45° for the sternocleidomastoid, splenius capitis, and semispinalis capitis muscles, and 15° for the trapezius muscle (Vasavada et al., 2002). These differences imply that individual neck muscles are not activated according to their biomechanics alone; however, the lines of action used for this analysis were inferred from a computational model and have not been validated in human volunteers. Muscle-induced intersegmental dynamics have proven very complex and difficult to unravel even with detailed computational models of the neck (Cox et al., 2014; Suderman et al., 2012). Differences between a muscle’s line of action and its neural control have also been reported in appendicular muscles (Buchanan et al., 1989; Hoffman & Strick, 1999; Kurtzer et al., 2006; Nozaki et al., 2005; van Zuylen et al., 1988), but these studies relied upon textbook definitions of a muscle’s predicted line of action and did not attempt to measure the biomechanics of human muscles. One way to overcome these limitations and characterize experimentally the biomechanics of individual human neck muscles is to measure the neck moment produced during intramuscular electrical stimulation. Intramuscular stimulation can selectively activate individual muscles (Popovic et al., 1991) and stimulation of individual muscles has been used to validate biomechanical models in the lower limbs (Hunter et al., 2009; Riener et al., 1996). Moreover, using the same electrodes to stimulate a muscle to establish its moment direction and to measure a muscle’s activity to establish its preferred directions provides an opportunity to determine how the biomechanics of a muscle shapes its neural control. The goal of our study was to investigate if the biomechanics of individual neck muscles can predict the neural control of these muscles when activated voluntarily. To achieve this goal, we first compared the neck moment direction generated by electrically stimulating a volunteer’s individual neck muscle, which measures the underlying biomechanics of the muscle, to the preferred direction derived from that volunteer’s spatial tuning curve, which represents the neural control of the muscle. Based on previous 28 findings in the neck (Vasavada et al., 2002), we hypothesized that the electrically stimulated directions and voluntary preferred directions for each muscle would not align. Second, we compared the intra-subject variability for both the electrically stimulated responses and voluntary preferred directions. Jones et al., (2002) argued that motor variability was mostly of central origin because it increased with the magnitude of voluntary activation, but not with the amplitude of muscle electrical activation. Accordingly, we hypothesized that variability in the voluntary preferred direction would be larger than variability of the electrically activated moment direction. This secondary analysis was possible because we measured the biomechanical action of each subject’s muscles rather than assuming or computing its action. The results of this study will help further our understanding of the complexity of the neural control of the human neck muscle system and provide valuable data for the design of neck muscle controllers for use in computational neck models. 3.3 Methods 3.3.1 Subjects Eight male subjects (Table 3.1) with no history of whiplash injury, neck/back pain, frequent/severe headaches, or neuromuscular injury participated in the study. Subjects provided written informed consent prior to participating in the study, which was approved by the UBC Clinical Research Ethics Board and conformed to the Declaration of Helsinki. Each subject had indwelling electrodes (0.076mm 304 SS, California Fine Wire Company, Grover Beach, CA, USA) inserted into the right sternocleidomastoid (SCM), splenius capitis (SPL), and semispinalis capitis (SSCAP) muscles under ultrasound guidance (MicroMaXX, Sonosite Inc., Bothell , WA) (Figure 3.1B). Two single wires were inserted into each muscle to achieve a ~20mm electrode spacing along the muscle fiber direction. This spacing was used to increase the volume of muscle that was activated during electrical stimulation and recorded during voluntary activation of the muscles. Each wire had ~2mm of insulation removed at the hooked tip and the electrode pairs were placed near the centre of each muscle’s horizontal cross section nominally spanning the C3-C5 levels (Figure 3.1B). In the SCM, the wires always remained superficial to the readily identifiable cleidomastoid subvolume (Kamibayashi & Richmond, 1998). 29 Table 3.1 - Anthropomorphic data of the male subjects in this study. Volunteer Age (years) Height (cm) Weight (kg) 1 27 178 76 3 27 180 86 4 32 171 62 5 25 180 82 6 27 178 89 7 39 185 85 8 27 178 90 9 30 185 88 Mean (SD) 29.3 (4.5) 179.4 (4.5) 82.3 (9.3) 3.3.2 Spatial Tuning Curve Procedures Subjects sat with their torso constrained to a vertical seatback and their head fixed to a six-axis load cell (Model 45E15, JR3 Incorporated, Woodland, CA, USA) through a tightly fitting helmet (Figure 3.1A). At least a day before the main experiment, subjects were familiarized with the set-up and performed maximum voluntary isometric contractions (MVIC) with visual feedback in the six principal directions (flexion/extension, left/right lateral bending, and left/right axial rotation) with two trials per direction (Fice et al., 2014). The MVIC magnitudes for each subject were defined as the peak moment measured in each direction over the two trials. The MVICs were done beforehand to avoid fatigue during the main experiment and they were done without the wire electrodes. For the main experiment, the spatial tuning task consisted of isometric contractions in twenty-six 3D moment directions while neck muscle EMG and head reaction forces and moments were recorded at 2 kHz (PXI-6221, National Instruments, Austin, TX, USA). EMG signals were amplified (× 2000) and filtered using a Neurolog system (Digitimer, Welwyn Garden City, Hertfordshire, UK). The 26 target directions were defined in a normalized moment space (Figure 3.1C; Fice et al., 2014), and then personalized for each subject by multiplying the components for each normalized direction by the subject’s corresponding principal-direction MVICs to calculate their target resultant moments (Equation 3.1). Note that we define our moment directions by the direction of the axis about which a moment is generated; e.g. flexion has a moment direction towards the left ear, which is considered a moment in the horizontal plane. Eight of the target directions were in the horizontal plane (flexion, extension, left and right lateral bending, and combinations of each, all with no axial moment) and then repeated with both a left and right axial moment to yield 24 target directions. Pure left and right axial moments brought the total to 26 target directions. Contractions in all target directions were performed at two magnitudes (7.5% or 15% MVIC). Subjects were given visual feedback of their 3D moments on a computer screen, which showed their axial rotation moment with a rotating needle indicator whose 30 origin moved to indicate their horizontal plane moments (Figure 3.1D). A contraction was considered successful if the resultant magnitude of the subject’s moment error (3D target moment minus 3D actual moment) remained less than 10% of the target resultant moment magnitude for one second. Contractions at each target direction were performed three times at both magnitudes, and the presentation order was randomized for direction within blocks of the same magnitude. Each block contained 26 trials (one in each direction), and block order was randomized between subjects. Equation 3.1 ?⃗⃗? target= (𝑁𝑇𝐹𝑙𝑒𝑥 𝐸𝑥𝑡⁄ ∙ 𝑀𝑉𝐼𝐶𝐹𝑙𝑒𝑥 𝐸𝑥𝑡⁄ )𝑖 + (𝑁𝑇𝐿𝑎𝑡𝐵𝑒𝑛𝑑 ∙ 𝑀𝑉𝐼𝐶𝐿𝑎𝑡𝐵𝑒𝑛𝑑)𝑗 + (𝑁𝑇𝐴𝑥𝑙𝑅𝑜𝑡 ∙ 𝑀𝑉𝐼𝐶𝐴𝑥𝑙𝑅𝑜𝑡)?⃗? where: ?⃗⃗? target = personalized target moment vector NTFlex/Ext, NTLat Bend, NTAxl Rot, are the components of the normalized target moment directions (Figure 3.1C), MVICFlex or Ext = maximum flexion or extension moment, MVICLat Bend = maximum left or right lateral bending moment, MVICAxl Rot = maximum left or right axial moment, 𝑖 , 𝑗 , ?⃗? unit vectors in the flexion/extension, lateral bending, and axial rotation directions respectively. 3.3.3 Neck Muscle Stimulation Procedures The indwelling electrodes were then connected to a constant current stimulator (Model DS5, Digitimer, Welwyn Garden City, Hertfordshire, UK) to electrically activate the neck muscles one at a time. The stimulation waveform consisted of 20 trains, with a random inter-train interval of 2-5s. Each train consisted of three square-wave pulses with duration of 0.5 ms and inter-pulse interval of 10 ms. This waveform was selected to deliver a substantial contraction while reducing discomfort for the subject and was based on pilot work and prior studies (Binder-Macleod et al., 1995, 1998; Crago et al., 1980; Popovic et al., 1991). The neck moments induced by electrical stimulation are produced by the activation of a single muscle rather than by multiple muscles for the voluntary tuning task. Since we expected the moment generated by an individual muscle during electrical stimulation to be less than the net moment generated by multiple neck muscles during voluntary activation, we chose two levels of electrical stimulation that generated moments that were about half those used to quantify the voluntary preferred directions. The high stimulation amplitude for each subject was set to generate a target moment of 7.5% MVIC and was compared to the muscle’s preferred direction for the voluntary task at 31 15% MVIC. In some cases, the subject found this stimulation level too uncomfortable and their maximum tolerable amplitude was utilized as the high-level stimulation. The actual moment was less than 80% of the target moment in five subjects’ SPL muscle and one subject’s SSCAP muscle, with lowest moment being 44% of the target moment. The low stimulation amplitude for each subject was set to generate a target moment of about 3.5-4% MVIC. Across all subjects, the high level electrical current amplitude varied from 2 to 5mA for SCM, 7 to 18mA for SPL, and 7 to 19mA for SSCAP, and the low level current amplitude was half of these values. Subjects were instructed to remain relaxed while the stimulation was delivered. The 3D moments generated by each stimulated neck muscle were recorded by the load cell. All moments were resolved to the C7-T1 joint axis by calculating the reaction moment required at C7-T1 to balance the moments and forces measured at the load cell. The location of the C7-T1 joint axis in each subject was determined as the midpoint of the line joining the sternal notch and the C7 spinous process (Queisser et al., 1994) measured using a 3D localizer (Polaris Vicra, Northern Digital Inc., Waterloo, ON, Canada). 3.3.4 Data Analysis To calculate the voluntary preferred directions from the spatial tuning task, we selected from each trial the 500ms window with the smallest standard deviation in the resultant moment from within the one-second period when the subject was generating a moment within 10% of the target moment. The corresponding EMG data for each muscle were digitally filtered to remove movement artifacts (4th order, dual-pass, Butterworth, 50Hz high-pass) and then the root mean squared (RMS) magnitude was calculated over the whole 500ms window to define the amplitude of the muscle activity for each trial (Figure 3.2). These amplitudes were averaged across the three repetitions for each direction and magnitude, and then the average was normalized by the largest average RMS EMG for each muscle across all directions within each contraction level, i.e., normalized RMS EMG varied between 0 and 1. Separate 3D spatial tuning curves were then created for each muscle and contraction magnitude for each subject using spherical coordinates, with each of the 26 contraction directions represented by a radial vector whose length was equal to the average normalized RMS across the three repetitions in that direction and whose 3D direction was defined by the vectorial average of the moment directions across the same three repetitions (Fisher et al., 1987). Each muscle’s preferred direction was then calculated from the vectorial sum of the 26 radial vectors and its focus was calculated as the magnitude of the vectorial sum of the 26 radial vectors divided by the arithmetic sum of the magnitudes of the radial 32 vectors. The spatial tuning curves were visually inspected for uni-modality and tested to determine if they were statistically different from a uniform distribution using a Rayleigh test on the focus (Batschelet, 1981; N=26, p<0.05). We defined the moment axis of the voluntary preferred direction in spherical coordinates using an azimuth angle (φ) in the horizontal plane (zero degrees at extension, positive 90° at right lateral bending) and an elevation angle (θ) above the horizontal plane (positive angles denote left axial rotation). To determine the electrically stimulated directions for both the high and low level, the C7-T1 3D moment components from the electrical stimulation trials of each muscle were aligned to the stimulus onset and averaged over the 20 stimulation trains. Before averaging the moment components, the bias was removed by subtracting the mean of each component over the 25ms before each electrical stimulation train. The resultant moment was then calculated from the averaged components and the direction of the moment was calculated at the time of the peak resultant moment. To test our first hypothesis, we performed a pairwise, i.e., within-subject, comparison of each muscle’s voluntary preferred direction and electrically stimulated direction for the high and low levels. We performed this comparison by first rotating both direction vectors of each muscle about an axis in the horizontal plane so that the electrically stimulated direction vector aligned with the pole (θ =90°). These rotations maintained the 3D angular difference between each subject’s voluntary preferred direction and their electrically stimulated direction but aligned the electrically stimulated directions of all subjects with the pole (θ =90°). A 95th percentile confidence ellipse was then calculated for the rotated voluntary preferred directions of each muscle across all subjects (Fisher et al., 1987; Kent, 1982; Leong & Carlile, 1998). If the 95th percentile confidence ellipse included the pole, then we concluded that there was no significant difference between the voluntary preferred direction and the electrically stimulated direction for that muscle. The mean difference between the voluntary preferred and electrically stimulated directions was reported as the angle between the centroid of the 95th percentile confidence ellipse and the pole (θ =90°). To give the reader a sense of the variability of these differences, the major and minor axes of the 95th percentile confidence ellipse were also reported. To visualize the mean response across subjects, voluntary preferred directions and electrically stimulated directions were separately vectorially averaged. The mean spatial tuning curve for each muscle was calculated by taking the vectorial average of each direction from each subject’s spatial tuning curves (Fisher et al., 1987). To estimate the variability across subjects, we fit a Kent distribution 33 to both the voluntary preferred directions and electrically stimulated directions, and then calculated the standard deviation ellipse (Fisher et al., 1987; Leong & Carlile, 1998). Intrasubject variability of the voluntary preferred direction was estimated by resampling the spatial tuning curves for each subject’s muscle 10 000 times by randomly selecting the RMS EMG from one of three repetitions for each of the 26 directions that made up the spatial tuning curve. The preferred direction from each tuning curve was then calculated using the same method described previously. A Kent distribution (Kent, 1982; Leong & Carlile, 1998) was then fit to the 10 000 preferred directions. Finally, the radius of a circle with the same area as the standard deviation ellipse was used as a univariate estimate of each muscle’s variability. This equivalent radius (expressed in degrees) was calculated separately for both the 7.5% MVIC and 15% MVIC contraction levels for all three muscles and for all subjects. The intrasubject variability of the electrically stimulated directions was estimated by fitting a Kent distribution (Kent, 1982; Leong & Carlile, 1998) to the 20 moment vectors measured for the 20 stimulations trains performed for each stimulation intensity, muscle and subject. Similarly, an equivalent radius was calculated for each standard deviation ellipse representing the intrasubject variability for each stimulation intensity, muscle and subject. To test our second hypothesis, i.e. to determine if intrasubject variability is larger for the voluntary preferred directions compared to the electrically stimulated directions, we performed a repeated-measures two-way ANOVA on the equivalent radius (voluntary vs. electrical activation; low vs. high intensity). This analysis was performed separately for each muscle studied. Because the equivalent radii were not normally distributed (Shapiro-Wilk test; p>0.05), a Box-Cox transformation was applied (λ = -0.65). No outliers were found in the transformed data as determined by the studentized residuals being within ±3. Any interaction observed between terms in the ANOVA was decomposed using post-hoc Tukey HSD test. All data processing and statistical analyses were performed in Matlab (Mathworks, Natick, MA, United States) except for the repeated measures ANOVA and its related Box-Cox transformation and post-hoc testing, which were performed using Statistica (TIBCO Software Inc., Palo Alto, CA, USA). Statistical significance was set to p<0.05. 3.4 Results The raw and RMS EMG data for the voluntary contractions revealed stereotypical patterns of activity for each muscle and contraction direction (Figure 3.2). In the SCM muscle, the voluntary activation levels 34 were larger for contraction directions that included flexion, ipsilateral bending and contralateral axial rotation. In SPL, the voluntary activation levels were larger for contraction directions that included extension and ipsilateral axial rotation, whereas in SSCAP, the voluntary activation levels were larger for contraction directions that included an extension moment but no axial rotation. The 24 spatial tuning curves (8 subjects × 3 muscles) generated from the voluntary contraction data had better-defined preferred directions at the 15% MVIC activation level than at the 7.5% MVIC activation level. At 15% MVIC, 21 of the 24 spatial tuning curves were significantly different from a uniform distribution (focus = 0.35 to 0.71; p = 0.00 to 0.04). The three distributions that were not different from uniform were in SSCAP for two subjects and SCM for a third subject (focus = 0.31 to 0.33; p = 0.06 to 0.08). At 7.5% MVIC, 16 of the 24 spatial tuning curves were significantly different from a uniform distribution (focus = 0.34 to 0.70; p = 0.00 to 0.05). The eight distributions that were not different from uniform included SCM & SPL in two subjects, SPL & SSCAP in another subject, and finally SCM and SPL in separate subjects (focus = 0.18 to 0.33; p = 0.06 to 0.42). When the voluntary contraction data from all subjects were combined, the average spatial tuning curves for the three muscles were unimodal and significantly different from a uniform distribution at the 15% MVIC level (focus = 0.44, 0.39, 0.52; p = 0.01, 0.02, 0.00 for SCM, SPL, and SSCAP), but not at the 7.5% MVIC level (focus = 0.30, 0.30, 0.51; p = 0.09, 0.09, 0.00 for SCM, SPL, and SSCAP; Figure 3.3). Based on these findings, we limited the analyses of the voluntary preferred directions to the 15% MVIC data. At the 15% MVIC contraction level, the average voluntary preferred direction for SCM consisted primarily of flexion and ipsilateral bending with a component of contralateral axial rotation (Figure 3.3), and this was consistent across subjects. SPL’s average voluntary preferred direction was predominantly extension with a small component of ipsilateral axial rotation and variable amounts of lateral bending: three subjects had a contralateral bending component, three had a minimal lateral bending component, and two had an ipsilateral bending component. SSCAP’s average voluntary preferred direction was almost entirely extension, although one subject had some contralateral bending, and another had a large ipsilateral bending component. Electrical stimulation generated well-defined moments for each muscle in all subjects (see exemplar data from one subject: Figure 3.4). For the SCM muscle, the grouped high level electrically stimulated direction was primarily lateral bending with components in flexion and contralateral axial rotation (Figure 3.3). In SPL, the high level electrically stimulated direction was primarily ipsilateral bending and extension with a component of ipsilateral axial rotation; whereas in SSCAP, the electrically stimulated 35 direction was predominantly extension with smaller components of ipsilateral axial rotation and lateral bending (Figure 3.3). The voluntary preferred directions and electrically stimulated directions were significantly different for all three muscles at the high level (95% confidence ellipse did not include the pole; Table 3.2; Table 3.3; Figure 3.3). The high level electrically stimulated direction for the SCM included more lateral bending and less axial rotation than its voluntary preferred direction. For SPL, the high level electrically stimulated direction included more lateral bending and slightly less axial rotation, whereas for SSCAP the high level electrically stimulated direction included more lateral bending and axial rotation (Figure 3.3; Table 3.2). The differences in average directions ranged from 21° for SSCAP to 39° for SPL. For all muscles, the intrasubject variability (estimated with the equivalent radius) was largest for the voluntary preferred direction at the low activation intensity (Figure 3.5). These observations were confirmed by the results of the repeated measures ANOVA: for the SPL and SSCAP, main effects revealed larger intrasubject variability for the voluntary activation than for electrical stimulation (SPL: 2.43°, F1,7= 23.8, p=0.0018; SSCAP: 0.78°, F1,7= 6.2, p=0.042) and larger intrasubject variability for low activation intensity than for high activation intensity (SPL: 0.86°, F1,7= 10.0, p=0.016; SSCAP: 0.62°, F1,7= 20.9, p=0.0026). For the SCM, a significant interaction between activation type (electrical/voluntary) and intensity (F1,7= 9.8, p=0.017) was observed. Post-hoc testing of the interaction revealed that the intrasubject variability of the voluntary preferred directions was larger than the intrasubject variability of the electrically stimulated directions (0.55° - 3.66°, multiple p, all < 0.016). Intrasubject variability also decreased from low to high voluntary activation intensity (3.1°, p=0.0007), but the effect of activation intensity was not significant for the electrical activation of the SCM (0.37°, p=0.06). 36 Table 3.2 - Voluntary preferred and electrically stimulated directions. The average azimuth and elevation angles for the mean voluntary preferred directions at 7.5% and 15% MVIC and the electrically stimulated directions for the right SCM, SPL and SSC muscles across all eight subjects. The azimuth angle (φ) was defined in the horizontal plane, with zero degrees at extension, positive 90° denoting right lateral bending. The elevation angle (θ) denotes axial rotation, with positive angles to the left and negative angles to the right. Voluntary Preferred Direction Electrically Stimulated Direction Muscle Azimuth φ (°) Elevation θ (°) Focus Muscle Azimuth φ (°) Elevation θ (°) 7.5% MVIC Low Stim. SCM 113.0 24.7 0.30 (p = 0.09) SCM 101.1 13.4 SPL -3.5 -41.3 0.30 (p = 0.09) SPL 41.5 -35.0 SSC 8.1 -4.1 0.51 (p = 0.00) * SSC 23.3 -9.2 15% MVIC High Stim. SCM 130.8 20.4 0.44 (p = 0.01) * SCM 105.1 9.8 SPL -3.3 -35.8 0.39 (p = 0.02) * SPL 42.2 -31.1 SSC 6.4 -3.4 0.52 (p = 0.00) * SSC 26.0 -12.6 * The pooled across subjects spatial tuning curve for the voluntary preferred direction was significantly different from a uniform distribution (p<0.05). Tested using the focus calculated over the 26 contraction directions and a Rayleigh test, see methods for details. Table 3.3 - The difference between voluntary preferred and electrically stimulated directions. The difference between the voluntary preferred directions at 15% MVIC and the high level electrically stimulated directions (n=8 subjects). Also included are the major and minor axes of the 95 percentile confidence ellipses of the differences. At p<0.05 there was a significant difference in the angular difference for all three muscles. Muscle Difference Δ (°) 95% C.I. Major (°) 95% C.I. Minor (°) 15% MVIC SCM 23.4 † 9.3 6.9 SPL 38.6 † 21.3 8.9 SSC 21.3 † 18.2 4.1 † The voluntary preferred direction was significantly different from the electrically stimulated direction (p<0.05). Tested by checking if the 95% confidence ellipse included the pole, see methods for details. 37 Figure 3.1 - The experimental set-up showing the subject seated with their torso constrained and their head fixed to a six-axis load cell through a tightly fitting modified helmet (A). Approximate locations of indwelling electrodes that were inserted with ultrasound guidance in the right sternocleidomastoid (upper left), splenius capitis (upper right), and semispinalis capitis (bottom) (B; Adapted from Gray’s Anatomy (Gray, 1918)). Spatial tuning task contraction directions and direction cosines for one quadrant of the 26 contraction directions performed (C; M. - moment, R. - right, Axl. - axial, Ext. - extension, Lat. Bend. -lateral bending). Visual feedback provided to the subjects during the spatial tuning task (D). Subjects generated horizontal plane moment to move the dot into the circle and generated axial moments to rotate the dial into the wedge. 38 Figure 3.2 - Exemplar filtered EMG for one repetition in 26 target directions at 15% MVIC for the sternocleidomastoid (SCM), splenius capitis (SPL) and semispinalis capitis (SSC) muscles. The shaded band spanning all three muscles shows the 500ms window that was selected based on minimal moment variability. Data beyond this band are not shown. The relative height of the dark columns within each shaded band represent the magnitude of the RMS EMG within that window. Positive moments are depicted as curved arrows about the moment axis (straight arrows). 39 Figure 3.3 - Spatial tuning plots for the 15% MVIC contraction of the SCM, SPL, and SSC muscles showing the voluntary preferred directions (black) and the electrically stimulated directions (grey) (n=8 subjects). Three different views are shown for each muscle: the left column shows the mean directions and standard deviation ellipses projected into the horizontal plane, the middle column shows the mean directions and standard deviation ellipses projected in the sagittal plane, and the right column shows a 3D view with the standard deviation ellipses (found by fitting a Kent distribution) and individual subject’s voluntary preferred directions and the electrically stimulated directions. The left and middle columns also show the tuning plots (black dashed line) plus one standard deviation (gray shading) for the directions that lie within the horizontal and sagittal planes respectively. The mean voluntary preferred direction shown here is the vector sum of all points on the spatial tuning curve of all subjects. The electrically stimulated direction shown here is the direction of moment generation when the muscle was electrically stimulated. Both the mean voluntary preferred and electrically stimulated directions are shown with arcs to represent the standard deviation elliptical cone across subjects, and note the arcs appear asymmetrical because of the way a 3D elliptical cone projects on a 2D plane as an arch. 40 Figure 3.4 - Exemplar data from one subject showing the moments generated by electrical stimulation in the sternocleidomastoid (top), splenius capitis (middle), semispinalis capitis (bottom). The shaded area is the mean moment response ±1 standard deviation. Positive values denote flexion, right lateral bending, and right axial moments. 41 Figure 3.5 - Intra-subject variability of the electrically stimulated direction and voluntary preferred direction as shown by the standard deviation ellipse from a Kent distribution (A). The left side shows the low activation/stimulation level, the right the high level and the rows show the three different muscles. The radius with the equivalent area of the standard deviation ellipses for each subject and muscle is also shown (B). Note the means and standard deviations were calculated on box cox transformed data (lambda = -0.65) which was then transformed back. 42 3.5 Discussion Our goal was to determine if the biomechanical actions of individual neck muscles are useful indicators of their neural control. To achieve this goal, we compared a volunteer’s preferred activation direction of three neck muscles during voluntary isometric contractions (representing their neural control) to the direction of the moment produced by the same muscles when electrically activated (representing their biomechanical action). We found that the average directions of voluntary activation and electrically stimulation differed by a minimum of 21° (SSCAP) to a maximum of 39° (SPL) at the 15% MVIC level. We also found that the intra-subject variability decreased for electrical vs. voluntary activation of the neck muscles as well as for larger activation intensity. These differences between the electrically stimulated direction and the voluntary preferred direction show that a neck muscle’s activation is not based solely on its underlying biomechanics, and that the neural system that controls neck muscles combines the biomechanics of numerous muscles to achieve its movement and stability. In all three muscles, we observed larger components in the right lateral bending direction during electrical stimulation than during voluntary activation. This finding is perhaps not surprising given the isolated, unilateral nature of the stimulus. A more unexpected finding was the reduced left axial rotation component for SCM during electrical stimulation compared to voluntary activation. The geometry of the right SCM suggests that the mastoid process would be pulled forward during isolated activation of this muscle (Kamibayashi & Richmond, 1998), and thus we expected the electrical activation to have a larger left axial rotation component compared to voluntary activation. Another unexpected result was a right axial rotation component that increased more for SSCAP than SPL during the electrical stimulation compared to the voluntary activation. The line of action of SPL is more lateral and less vertical than the line of action of SSCAP (Kamibayashi & Richmond, 1998), and therefore we expected isolated electrical activation to increase the right axial rotation component more in SPL than in SSCAP. Our expectations were based on an assumption that the neural control of the neck muscle system would attenuate rather than amplify the underlying biomechanics of individual muscles. However, these findings show our assumption was incorrect, and provide compelling evidence that the neural control can both attenuate and amplify, sometimes simultaneously, different components of a muscle’s underlying biomechanics. The 21° to 39° differences between the voluntary preferred directions and the measured biomechanical directions are smaller than the 45° to 65° differences reported by Vasavada et al. (2002) for the same muscles. Vasavada et al. (2002) compared the voluntary preferred directions of various neck muscles to a computational model that predicted the direction of the muscle’s moment arm. The 3D directions of 43 our electrically stimulated muscles closely matched the lines of action predicted by their computational model, and thus the primary differences between our findings and their findings reside in the voluntary preferred directions. The voluntary preferred directions in their work had larger axial rotation components for all three muscles, a larger flexion component for SCM, and smaller lateral bending components for SPL and SSCAP. Methodological differences in the voluntary contraction procedures, particularly our use of moments normalized to the volunteer’s MVIC components compared to their use of fixed moments that ranged from 11% of the maximum extension moment to 80% of the maximum axial rotation moment, could explain these different results. The differences in preferred direction that we observed between the average voluntary and electrically stimulated directions in human neck muscles are similar to those observed for human appendicular muscles, where the preferred activation direction does not align with assumed biomechanical function based on musculoskeletal anatomy (Buchanan et al., 1989; Hoffman & Strick, 1999; Kurtzer et al., 2006; Nozaki et al., 2005; van Zuylen et al., 1988). Our findings also agree with observations in primate neck muscles, where natural head movements are generated by neck muscle activity that does not accord solely with the muscle’s assumed biomechanical function (Farshadmanesh, Byrne, Keith, et al., 2012; Farshadmanesh, Byrne, Wang, et al., 2012). A major distinction with previous appendicular and cervical work, however, is that we measured the biomechanical line of action for each muscle using electrical activation instead of relying on the muscle’s line of action reported in anatomical textbooks or computational models. Relying on anatomical descriptions is particularly problematic in the neck because over twenty-five pairs of muscles act over several joints with multiple degrees of freedom, with some muscles not even attaching to the cervical vertebrae. All these factors render assumptions regarding the line of action of a neck muscle based on its insertion/attachment points and fiber orientation difficult. Measuring the biomechanical action of a muscle also allowed us to determine that the intra-subject variability was 1) larger for the voluntary preferred direction than the electrically stimulated direction and 2) decreased as the level of activation increased. The generally larger intra-subject variability for voluntary preferred directions supports the notion that the nervous system is optimizing something other than individual muscle biomechanics. This “other” optimization strategy may be due to the neck’s large mobility, which constrains the size and line of action of the neck muscles, meaning that not enough muscles may be optimally aligned to perform a given task. For example, there are no large neck muscles devoted entirely to generating axial moments; muscles like the right SPL and left SCM will act 44 agonistically to generate right axial rotation, but the moments they generate in other directions need to be cancelled by each other and/or other muscles. Indeed, many predominantly vertical muscles may be used in non-optimal ways to generate axial moment (Ackland et al., 2011; Peterson et al., 2001; Vasavada et al., 1998) and could explain our findings that the voluntary preferred directions of SCM and SPL were more aligned with axial rotation than the electrically stimulated directions. For the voluntary activations, the decreased intra-subject variability we observed with increased activation level may appear to contradict the concept of signal dependent noise. This latter concept stipulates that, within a single muscle, the variability of force output increases as the voluntary activation increases (Harris & Wolpert, 1998; Jones et al., 2002; Schmidt et al., 1979). However, to produce our voluntary preferred directions we did not measure the variability of a single neck muscle; rather we measured the variability of a group of neck muscles working synergistically. For multiple muscles acting across a joint, there is only one combination of muscle activation that can achieve the MVIC effort in any given direction (Valero-Cuevas, 2000). As activation decreases (i.e. submaximal activation of the muscles), multiple combinations of muscle activation can achieve the target force direction, similar to the redundancy problem where more muscles can be activated across a joint than there are degrees of freedom for that joint (Bernstein, 1967). Hence, the multiple combinations of neck muscles available to produce a submaximal effort in a given direction could explain the larger intra-subject variability in the preferred direction for lower activation levels. Potentially, the biomechanical constraints in the neck musculoskeletal system increasingly dictate the activation of muscles contributing to the net neck moment, resulting in lower intra-subject variability in preferred direction as activation level increases. It is also possible that subtle differences in posture and stability requirements of the unstable neck system between repetitions contribute to the variability in voluntary muscle activation to produce the target 3D moment. Our findings have potentially important implications on musculoskeletal modeling of the human neck. The latest neck modeling efforts use feedback controllers for muscle activation (Meijer et al., 2013; Östh et al., 2015), but the neural motor controller cannot be a simple feedback controller because the neural delays and sensorimotor noise would make the system (i.e. our body) unstable (Franklin & Wolpert, 2011). Thus, as we move towards more advanced physiologically-based neck muscle controllers, it may be tempting to make simplifying assumptions about the neural control of neck muscles and its variability. One such assumption is that the neural controller will activate a neck muscle maximally when generating or resisting a load on the head that is aligned with that muscle’s biomechanical line of action. 45 This assumption would simplify a physiologically-based muscle controller, although the results we present show that such a simplification is not valid. In the future, optimal feedback control (reviewed in S. H. Scott, 2012) could potentially be used to generate muscles activation schemes in novel situations that would improve the biofidelity of neck models, and data from the results we present here could be used to build the muscle activity versus moment relationships. A limitation of our work is the number of contraction directions we examined. We attempted to balance the number of directions and repetitions against the risk of subject fatigue. Future work could eliminate the lower activation level and increase the number of contraction directions at a single activation level to better map the entire contraction direction space. Because the voluntary preferred direction data at 7.5% MVIC were not significantly different from a uniform distribution, we only compared voluntary data at 15% MVIC to electrically stimulated data that attempted to match the 7.5% MVIC level. The similarity in preferred directions between the non-significant 7.5% MVIC and significant 15% MVIC distributions justifies this decision. It further suggests the uniform distributions observed at the 7.5% MVIC level were likely due to sample size. Also, our intra-subject variability analysis was based on electrical stimulation data at 3.75% and ~7.5% MVIC versus voluntary data at 7.5% and 15% in order to match the expected decrement in moment from activating a single muscle during the electrical stimulation relative to activating multiple muscles during the voluntary tuning task. This 50% reduction may not accurately reflect the actual contribution of the other muscles during the voluntary task; however, given that variability increased with lower voluntary activation levels, we would expect the reported differences between voluntary and electrical activation to be larger if we had matched the voluntary activation and electrically stimulated levels of muscle activation. We also stimulated the muscles using a single pair of indwelling electrodes spaced about 20 mm apart to increase the number of muscle fibers that were electrically activated. It is possible that the moment generated with this electrode spacing may not represent the biomechanics of the whole muscle, but rather only the muscle fibers that were activated by the stimulation electrodes. Differences between local and whole muscle biomechanics could stem from functional compartmentalization in the neck muscles (Richmond et al., 1985; Wilson et al., 1983), although a preliminary study did not identify obvious compartments in SPL (Siegmund et al., 2011). Finally, our results are limited to three muscles tested during an isometric task in a single neck posture. There are more than 20 other muscle pairs in the neck, and further work is needed to examine their neural control in different postures and during different tasks. 46 In summary, the voluntary preferred directions of the sternocleidomastoid, splenius capitis, and semispinalis capitis muscles derived from an isometric spatial tuning task in the neutral head/neck posture did not align with the moment produced when electrically stimulating those muscles in the same posture. The intra-subject variability of the voluntary preferred direction was larger than the electrically stimulated direction and decreased as the level of activation increased. These findings show that the neural control of neck muscles is not solely optimized for their biomechanical function but, as activation increases for the levels tested here, biomechanical constraints in part dictate the activation of synergistic neck muscles contributing the net moments in a target direction. The concept of biomechanical constraints dictating synergistic muscle activation with larger net forces/moment requires further validation in other multi-muscle multi-segment musculoskeletal systems. 47 Chapter 4. Head postures during naturalistic driving 4.1 Preamble3 The risk of whiplash injury is higher when occupants have their head turned before rear impact. In Chapter 5, the head/neck kinematics and neck muscle responses of subjects in non-neutral head posture before whiplash-like perturbations will be presented. To ensure the perturbation data were relevant to real-world situations, the postures drivers are likely to adopt while driving needed to be quantified. Therefore, the goal of this chapter was to quantify common head postures during naturalistic driving. 4.2 Introduction A rotated head posture at the time of a rear-end impact is associated with a higher risk of acute and chronic whiplash injury. The risk of acute whiplash injury in rear-end impacts increases from 23% when looking straight ahead to 36% when the head is turned (Jakobsson et al., 2008). Moreover, chronic whiplash patients were in rotated and/or inclined head postures at the time of the impact more frequently (46-57% vs 24-28%) than asymptomatic patients 1-2 years after injury (Radanov et al., 1995; Sturzenegger et al., 1995). These field data provide compelling reasons to study the effect of rotated head postures on whiplash kinematic and neuromuscular responses in the laboratory. To do so, however, we must first quantify normal rotated head postures while driving. Thus, the overall goal of this study was to quantify the head posture of drivers during naturalistic driving. Prior measurements of the rotation angle between a driver’s head and torso have shown that drivers spend about 13% of their time in non-neutral head postures (head-torso yaw angles greater than ±15°; Shugg et al., 2011). Non-neutral head postures were more common during starts, stops, lane-changes, and while driving on residential streets compared to driving on high-speed city streets or highways. Because most patients (54-90%) with neck pain following a rear-end impact report that their vehicle was stopped at impact (Deans et al., 1987; Gibson et al., 2000; Norris & Watt, 1983; Ryan et al., 1994; Sturzenegger et al., 1995), the more frequent and potentially larger head rotation angles while stopped 3 This chapter is an Author’s Original Manuscript of an article published by Taylor & Francis Group in Traffic Injury Prevention on September 27th, 2018, available online: https://doi.org/10.1080/15389588.2018.1493582. The full citation is as follows Fice, J. B., Blouin, J. S., & Siegmund, G. P. (2018). Head postures during naturalistic driving. Traffic Inj Prev., 19(6), 637–643. 48 may be particularly relevant to whiplash injury. Shugg et al (2011), however, did not quantify head postures during typical driving tasks (e.g., shoulder check, rear-view mirror check, etc.) and therefore additional work is needed to quantify non-neutral head postures while driving and stopped. The goal of this study was to quantify driver head orientation during real-world driving. Specifically, we measured the head orientation of drivers during six common movements: bilateral shoulder and side mirror checks, rear-view mirror checks and looking at the front seat passenger. We hypothesized that drivers spend proportionally more time with their head rotated out of the neutral posture and adopt head orientations that are further from the neutral posture when the vehicle is stationary compared to when the vehicle is moving. These data will be useful for investigating occupant responses and injury risk in common rotated head postures for human volunteer testing and computational modeling. 4.3 Methods 4.3.1 Subjects Twenty subjects (Table 4.1) with at least five years of driving experience and free from self-reported neck or back pain participated in this study. Subjects provided written informed consent prior to participating in the study, which was approved by the UBC Clinical Research Ethics Board and conformed to the Declaration of Helsinki. Table 4.1 - Anthropomorphic data of subjects in this study. Volunteer Age (years) Height (cm) Weight (kg) Volunteer Age (years) Height (cm) Weight (kg) Females Males 1 31 165 59 1 31 175 83 2 23 168 55 2 31 180 86 3 24 163 57 3 25 183 80 4 24 173 70 4 24 174 72 5 25 152 50 5 48 189 117 6 36 178 70 6 52 178 72 Mean (SD) 27.2 (5.2) 166.5 (8.8) 60.3 (8.2) 7 28 192 93 8 55 184 77 9 28 180 81 10 32 175 81 11 26 183 80 12 53 186 86 13 28 170 70 14 42 165 76 Mean (SD) 35.9 (11.5) 179.6 (7.3) 82.3 (11.8) 4.3.2 Instrumentation Each subject and the vehicle were instrumented with a 9 degree of freedom (DOF) inertial measurement unit (IMU; Razor, Sparkfun, Niwot, CO, USA) that consisted of tri-axial accelerometers, magnetometers, 49 and gyroscopes. The subject’s IMU was secured to their forehead with a tight-fitting headband and the vehicle’s IMU was mechanically fastened to its centre console about 46cm behind the vehicle’s centre of mass. IMU data were recorded at 102.4Hz on a laptop computer using LabVIEW software (National Instruments, Austin, TX, USA). This recording frequency covered the range of frequencies expected for head (<15Hz; Carriot et al., 2017) and vehicle (<10Hz; Jang & Han, 1997) movements. One video camera recorded the subject’s head, face and upper torso, and a second video camera recorded the road ahead for the duration of the experiment. The video was synchronized to the IMU data using a laptop-generated audible chime recorded on the audio track. 4.3.3 Procedures Subjects drove a 2010 Subaru Impreza with the experimenter in the right passenger seat. They drove along a set route through Richmond and Vancouver (British Columbia, Canada) that incorporated highways and local streets. The route took nominally 74 minutes to complete (route: https://goo.gl/maps/jyJ1vJUAKS72; driven counter clockwise). From a distance perspective, the route was 80% city streets (40-60km/h limit) and 20% highway (80-90km/h limit). Subjects adjusted the seat, steering wheel, and mirrors prior to the start of the test. The passenger seat was always fully rearward. We focused our analyses on six common driving-related head movements: bilateral shoulder and side mirror checks, rear-view mirror checks and looking at the front seat passenger. To record and classify when a head movement of interest occurred, the experimenter pressed a unique key on the data-acquisition laptop to flag each movement type in the recorded data stream. Each head movement classification was later confirmed using video data. Subjects were instructed to drive using their normal style. The experimenter engaged in periodic conversation with the driver to mimic normal conditions and the only instructions given to subjects once on the road were navigational. 4.3.4 Data Analysis Head and vehicle angular orientation data as a function of time were calculated from the IMU data. We used a Direction Cosine Matrix (DCM) fusion algorithm that combined the integral of angular velocity to compute head orientation in space with the magnetometer and gravity vector from the accelerometers to correct for drift in the yaw and pitch/roll axes respectively (Baldwin et al., 2007; Euston et al., 2008; Mahony et al., 2006, 2008; Premerlani & Bizard, 2009). The IMU utilized the Attitude and Heading Reference System firmware that was based on the DCM fusion algorithm (Bartz, 2016). We modified this firmware by offloading the DCM fusion algorithm from the sensor’s Arduino chip to the LabVIEW code to achieve the desired sampling rate (102.4Hz). 50 Absolute head and vehicle angular orientations in space were calculated as Euler angles with a decomposition order of yaw, pitch, and roll. Head angles are positive for rightward yaw, upwards pitch, and rightwards roll. For both sensors, the yaw axis was aligned with gravity, and the initial orientation of the pitch and roll axes was to the right and forward respectively. The absolute Euler angles for the head and vehicle were then converted into rotation matrices to compute the rotation matrix between the head and vehicle orientations. This latter rotation matrix was then converted back to Euler angles that defined the orientation of the head relative to the vehicle. Calibration of the magnetometer was performed in the test vehicle to account for ferromagnetic disturbances caused by the car (Bartz, 2016). The IMU accuracy, DCM algorithm, and calibration were evaluated by comparing the computed output to an optical motion tracking system (Optotrak Certus, Northern Digital, Waterloo, ON, Canada). In pilot data on a single subject, four infrared light emitting diodes (IRLED) markers were affixed on the skin (two on the zygomatic arch and two on the mandible) while wearing the IMU sensor on the forehead. While in the test vehicle, the motion tracking system captured the IRLED markers through the open driver’s side window, and the subject performed 18 movements that included forward looking to left shoulder check (6), left mirror check (8), and rear-view mirror check (4) over 90s. Peak head angles from the IMU and Optotrak markers were calculated in the same manner as the on-road experimental data, apart from omitting the vehicle IMU because it was stationary. The median (maximum) of the absolute value of the differences in peak head angles across all movements was 0.34 (1.26)°, 1.25 (2.88)°, and 1.59 (3.95)° for yaw, pitch, and roll respectively. This validation does not consider the potential for drift over a long time (~1hr), but the next paragraph addresses our approach to remove bias as a result of long term drift. Head movements were manually extracted from the data using a custom MATLAB (Natick, MA, USA) program that streamed the head position data alongside the video recordings (Figure 4.1). A window about each movement was first created (left and middle vertical lines in Figure 4.1), and then a reference near each movement (within 15s) was selected to define the neutral head orientation (rightmost vertical line in Figure 4.1). Selecting a local neutral reference minimized drift in the sensor signals. Head movement start time was manually selected when head angle deviated from neutral, i.e., looking forward. For linked head movements (e.g., left mirror check followed by left shoulder check), start time was selected as the end of the plateau of head angle for the first movement. Movement stop time was selected as the time when head angle returned to near forward facing, or for linked movements the time when the second movement started. Some sequential movements used the same 51 local reference. Movement classification was confirmed during the data/video review, and movements missed while on the road were added (19.5% of all movements). Peak head angles were extracted by finding the maximum yaw angle within the window and then selecting the roll and pitch angles at the same time point. All angles for each movement were computed relative to the local neutral reference head posture for that movement to remove bias that remained after the DCM fusion algorithm’s drift correction. Distinguishing between the driver looking out a side window and looking at the side-view mirrors was difficult, so these movements were grouped together. To classify whether the vehicle was stationary or moving, we examined the root mean square (RMS) noise on the vehicle’s IMU in the fore-aft accelerometer channel. We isolated noise using a high-pass 4-pole Butterworth filter with a 5Hz cut off. By manually segmenting 1hr of forward facing video into clearly moving or stationary parts, we found that the root mean squared (RMS) noise calculated over 1s non-overlapping windows exceeding 0.008g discriminated vehicle movement 99.3% of the time. For individual head movements, we classified vehicle movement status by calculating the RMS noise over the head movement duration. From the post-processed data, we first tallied the number of movements for each subject and movement type. We repeated this calculation when the vehicle was stationary and moving. Head movement duration was calculated as the difference between the start and stop time. For each subject, we determined the proportion of time they spent in non-neutral postures by summing the duration of all head movements and then dividing by the total time of the drive. Using the same approach, we calculated the proportion of time in non-neutral postures when the vehicle was stationary and moving and compared these values to test our first hypothesis. The comparison was done using a Wilcoxon matched pairs test because the proportion data are bounded between 0-1, which violates an assumption of continuous variables for parametric comparison methods. As an extension to the first hypothesis, we also compared the proportion of the movements while the vehicle was stationary against the proportion of time the vehicle was stationary using a separate Wilcoxon matched pairs test for each movement type. One subject was omitted from the look-at-passenger comparison because they did not perform this movement during their data collection. To determine the average head angles for each subject and movement type, the peak head Euler angles were converted to quaternions, averaged (Markley et al., 2007), and then converted back to Euler angles. Averages across all subjects were then calculated in the same manner. To test if the head angles were significantly different between the different movements we conducted a one-way ANOVA using 52 head yaw angle as the dependent variable. The data were normally distributed (Shapiro-Wilk Test) and there were no outliers (studentized residuals within ±3). The same individual referred to in the previous paragraph was omitted from this analysis again because they did not once look at the passenger during their data collection. Post-hoc Tukey comparisons were performed for significant main effects. To test our second hypothesis, i.e., that subjects adopt head orientations that are further from the neutral posture when the vehicle is stationary compared to when the vehicle is moving, we first used linear regressions to find the slope between the head yaw angles for stationary and moving vehicles across all head movements. A separate regression was performed for each subject to maintain a pairwise comparison, and the intercept was set to zero. We then averaged the slopes across subjects and tested if they were significantly different from unity using a one sample t-test. Statistical testing focussed on head yaw angle because the other angles were generally small (<10°) for all movement types. All statistical tests were performed with Statistica (TIBCO Software Inc., Palo Alto, CA, USA), and significance was set at p=0.05. Data are presented as mean ± standard deviation or median (range). 4.4 Results The driving time was 68±5mins, within which the vehicle was stationary for 13±2mins. The total number of recorded head movements per subject was 273.1±81.3, with 61.0±18.5 of these movements taking place while the vehicle was stationary (Figure 4.2). Overall, subjects spent 10.2(5.5-17.1)% of their drive times executing one of the six head movements. This proportion increased to 17.5(7.9-27.2)% of the time the vehicle was stationary and decreased to 8.2(4.4-15.0)% of the time the vehicle was moving. The larger proportion of time subjects spent in non-neutral head postures in a stationary versus moving vehicle confirmed our first hypothesis (Z=3.92, p<0.0001). The number of times subjects adopted the six postures over the entire drive and within the moving and stationary periods varied. Over the entire drive, subjects performed 15(5-39) left shoulder checks, 82.5(29-167) left mirror checks, 40.5(10-168) rear-view mirror checks, 27.5(3-113) right mirror checks, 60(0-185) passenger-looks, and 12.5(1-28) right shoulder checks (Figure 4.2). The proportion of movements in a stationary vehicle compared to the proportion of time the vehicle was stationary was greater for left mirror checks and passenger-looks and less for bilateral shoulder checks and rear-view mirror checks (Z=2.73-3.92, p=0.0001-0.006; Table 4.2). The proportion of right mirror checks in a stationary vehicle were similar to the proportion of time the vehicle spent stationary(Z=1.23, p=0.22). 53 Table 4.2 - Percentage of time the vehicle was stationary (bottom row) compared to the percentage of movements for each movement type while the vehicle was stationary. Significant differences (p < 0.05) are marked with an asterisk. Movement Type Percentage of movements while vehicle was stationary Median (range) Statistical difference between percentage of time the vehicle was stationary Left shoulder 6.5% (0.0-33.3%) Z = 3.17; p = 0.0015* Left mirror 23.1% (11.5-62.1%) Z = 2.73; p = 0.0064* Rear-view mirror 7.5% (0.0-17.8%) Z = 3.92; p < 0.0001* Right mirror 23.0% (0.0-44.4%) Z = 1.23; p = 0.22 Passenger-look 44.4% (23.9-90%) Z = 3.82; p = 0.0001* Right shoulder 0.0% (0.0-33.3%) Z = 3.58; p = 0.0003* Percentage of time the vehicle was stationary 19.3% (13.8 - 26.1%) Yaw was the biggest component of head movement for all movements (Table 4.3; Figure 4.3). The average pitch angle for all movements was negative (chin down), except for rear-view mirror checks, which had an average pitch angle of 1.9° (chin up). Roll was generally towards the same side as yaw for all movements, with the right shoulder check having the largest average roll angle (9.4°) towards the right. Subjects adopted different yaw angles for the six head movements (F5,90=1092.5, p<0.0001; Table 4.3). Decomposition of this main effect revealed that peak yaw angle differed between all movement types (post-hoc Tukey: p=0.0001-0.0002). Table 4.3 - The group averages for yaw, pitch and roll at the instant of peak yaw angle for each head movement. Separate averages were calculated for when the vehicle was stationary or moving. Movement Type Stationary Vehicle Moving Vehicle Combined Yaw Pitch Roll Yaw Pitch Roll Yaw Pitch Roll Left shoulder -91.7° -10.5° -6.6° -80.5° -7.9° -2.0° -81.5° -8.1° -2.4° Left mirror -39.5° -3.0° -2.0° -32.7° -3.1° -1.6° -34.3° -3.0° -1.7° Rear-view mirror 16.7° 1.8° 2.5° 16.1° 1.9° 1.4° 16.2° 1.9° 1.5° Right mirror 44.6° -1.3° 5.5° 41.5° 0.2° 3.7° 42.1° -0.1° 4.0° Passenger-look 59.8° -2.4° 7.6° 57.3° -2.9° 8.0° 58.2° -2.7° 7.9° Right shoulder 87.5° -3.2° 10.8° 84.0° -2.3° 9.3° 84.3° -2.4° 9.4° In all but two subjects, the driver’s head moved further from its neutral posture when the vehicle was stationary than when the vehicle was moving (see the slopes greater than unity in Figure 4.4). Overall, subjects rotated their heads further from neutral when the vehicle was stationary than when the vehicle was moving, a finding that confirmed our second hypothesis (m=1.11±0.073, t19=5.90, p<0.0001). Given the average slope of the stationary vs moving yaw head angle, we expect drivers to move their head 9.6° further from neutral for a 90° head movement in a stationary compared to a moving car. 54 Figure 4.1 - Custom MATLAB user interface used to manually extract movements from the head position data. The head orientation data (solid gray traces) and logged movements (black vertical bars) are streamed along side the participant video (note the forehead-mounted IMU). The black circle on each trace shows the head angle matching the displayed video frame. The height of the black bar (see right scale) corresponds to the movements logged during the experiment (1 = L. shoulder, 2 = L. mirror, 3 = rear-view mirror, 4 = R. mirror, 5 = passenger, and 6 = R. shoulder). To extract all movements, a researcher watched the video/data stream and identified movements using the shape of the data and the visible head movements. The (same) experimenter then manually selected a time point at the initiation of movement (leftmost vertical line), a time point at the end of the movement (middle vertical line), and a time point near the movement when the driver was looking straight ahead (rightmost vertical line). Each movement was re-categorized to confirm or correct the logged value. Movements missed on the road were added. The movement highlighted in this figure is a look at the passenger. 55 Figure 4.2 - The number of movements recorded for each subject and movement type for when the vehicle was stationary (triangle), moving (square), and both (circle). Individual subject data are shown with solid markers and group medians are also shown with hollow markers. Note the large variation between subjects for some of the movements. 56 Figure 4.3 - Mean peak head angles in yaw, pitch, and roll for each subject and movement type. Group means (hollow markers) are shown below the intra-subject means (solid markers). 57 Figure 4.4 - Mean peak head angles in yaw for a moving vs stationary vehicle. Intra-subject means (hollow markers) are shown with a linear regression fit (zero intercept; grey lines) for each subject, and a black dashed line representing a unity slope that would indicate no difference between moving and stationary vehicle data. Note that data points were omitted when a subject did not perform a movement when the vehicle was both stationary and moving. 4.5 Discussion The goal of this study was to quantify occupant head posture during naturalistic driving. To achieve this goal, we recorded the average 3D angular head position of drivers during six common movements and found that the peak yaw head angles averaged -81.5°, -34.3°, 16.2°, 42.1°, 58.2°, 84.3° for left shoulder check, left mirror check, rear-view mirror check, right mirror check, passenger-looking, and right shoulder check respectively. Drivers spent more time in the six postures we studied and adopted larger head yaw angles in these postures when the car was stationary than when the car was moving, confirming both of our hypotheses. Our results provide a possible link between seemingly unrelated observations in the whiplash injury epidemiology literature. Most patients (54-90%) with neck pain following a rear-end impact report that their vehicle was stationary at the time of impact (Deans et al., 1987; Gibson et al., 2000; Norris & Watt, 1983; Ryan et al., 1994; Sturzenegger et al., 1995). Further, symptomatic patients were significantly 58 more likely than asymptomatic patients to be in a stationary vehicle with their head rotated at the time of impact (Sturzenegger et al., 1995). Although a statistical link between injury and stationary vehicles has not been shown (Norris & Watt, 1983; Radanov et al., 1995; Sturzenegger et al., 1995), our results indicate that drivers are more likely to have their heads turned and turned by larger amounts when their vehicles are stopped than when their vehicles are moving. The reason non-neutral head postures increase whiplash injury risk remains unclear. Current hypotheses for increased injury risk include greater damage to the alar and transverse ligaments (Kaale et al., 2005a), greater elongation of the vertebral artery (Ivancic et al., 2006), greater nerve root damage due to intervertebral foramen narrowing (Tominaga, Maak, et al., 2006), and greater facet capsular ligament strain (Siegmund, Davis, et al., 2008). Further, modeling axially rotated postures during rear-end impacts has shown increased alar and capsular ligament strains when compared to neutral posture (Shateri & Cronin, 2015). The head postures characterized here for common driving tasks can now be used as initial conditions to further study occupant response and tissue injury risk for non-neutral postures during rear-end impacts. Our finding that drivers, on average, spend about 10.2% of their driving time in one of the six postures we studied is similar to prior work showing that drivers spend on average 13% of their driving time in non-neutral postures (head-torso yaw > 15°; Shugg et al., 2011). All six of our movements generated average yaw angles greater than ±15° (head relative to car), and therefore the extra 2.8% observed by Shugg et al. (2011) may be due to other head movements, different proportions of moving/stopped times, or different driving environments. When we post-process our data to match Shrugg et al. (2011), we found peak yaw angles that averaged -40.8° and 43.9° for left and right movements, which was similar to their findings of -35.7° and 42.5° respectively. Our averages may be slightly further from neutral because our values were relative to the vehicle rather than relative to the torso. A more thorough comparison of the head angles measured in Shugg et al. (2011) is not possible because they did not segment their data into different movement types. Although we were able to measure driving behaviour on open roads, our study was limited to a single compact car. Vehicles with different geometries would likely affect the absolute values of the head angles we measured. Nevertheless, we believe that the proportions of time spent in the different postures depends less on the specific vehicle, and that the main findings of the study are more widely applicable. Moreover, compact cars form one of the biggest segments of the vehicle fleet, and therefore our absolute angles may be representative of a large portion of the global vehicle fleet (Demandt, 2018). 59 We could not differentiate between side-view mirror checks and looks out the side window, which increased the variability of the side mirror check data. Although subjects were instructed to drive normally, they were nonetheless aware they were being studied, and we do not know whether they adapted their natural driving behavior to the experiment. In summary, we quantified the average head postures drivers adopt when performing six common head movements while driving. We showed that drivers spend a larger proportion of their time in these rotated head postures and rotate further from the neutral head posture when their vehicle is stationary compared to when it is moving. These findings provide a possible explanation for why drivers may be more likely to be injured when hit from behind while their vehicles are stationary. The head postures characterized here can be used as initial conditions in volunteer and computational studies to improve our understanding of why non-neutral head postures increase the risk of whiplash injury. 60 Chapter 5. Head/neck kinematics and muscle responses in volunteers with non-neutral initial head postures during low-speed rear impacts 5.1 Preamble Epidemiological and biomechanical evidence suggest that occupants are at higher risk for neck injury when their head is in a rotated non-neutral posture before impact. There are limited studies to define the neck muscle responses in these situations, and what is available is insufficient to validate head-neck models. In this experiment, we measured the neck muscle responses and head/neck kinematics of non-neutral posture volunteers during rear impact. The data from Chapter 4 were used to set non-neutral postures that are relevant to postures that drivers adopt during natural driving. The data from Chapter 2 (off-axis MVICs) can be combined with these data to improve neck modeling of occupants in non-neutral postures. 5.2 Introduction Whiplash associated disorders (WAD) are the most common injury in motor vehicle collisions, accounting for 21 to 28% of all injuries (Quinlan et al., 2004; Styrke et al., 2012). Epidemiological evidence from occupants involved in real world rear impacts has shown that head-turned postures increase the risk of minor neck injuries (36% injury risk for head turned compared to 23% for head straight; Jakobsson et al., 2008). The increased injury risk when the head is in non-neutral positions may result from ligament, arterial or nerve root damage (Ivancic et al., 2006; Siegmund, Davis, et al., 2008; Tominaga, Maak, et al., 2006), but the potential role of neck muscle activity has not been fully investigated. Knowledge of the role of neck muscles is important to model the head/neck responses under head-turned conditions and to understand the underlying injury mechanics. Several tissues in the cervical spine have been proposed as the source of increased whiplash injury risk in non-neutral postures. First, the alar and transverse ligaments have shown a higher incidence of damage in MRI scans when the occupants of automobile crashes report having their head turned at the time of impact (Kaale et al., 2005a). Results from modeling kinematic responses to rear impacts also suggest axial rotation increases the strains in the alar and capsular ligament of the upper cervical spine when compared to neutral posture (Shateri & Cronin, 2015). These findings, however, were questioned in a recent meta-analysis of upper cervical spine ligament MRI that suggested signal fluctuations attributed to ligament damage are not correlated with WAD symptoms (Li et al., 2013). Further, a cadaveric cervical spine test that showed no increase in alar, transverse, or apical ligament strain in head 61 turned rear impact up to 8g acceleration (Maak et al., 2006). In the same set of impacts and cadaveric cervical spines, Ivancic et al., (2006) found that axially rotated postures led to vertebral artery elongation beyond baseline values, which may occlude blood to the brain or spinal cord. Tominaga, Maak, et al., (2006) further observed increased narrowing of the intervertebral foramen, which may increase the risk of damage to nerve roots; and Panjabi, Ivancic, et al., (2006) reported multi-axis intervertebral flexibility increases, which suggest that ligaments or intervertebral discs were damaged. Finally, cervical spinal segments that had an axial torque applied before a quasi-static whiplash-like load was applied resulted in double the capsular ligament strain (17 to 34%), which potentially increases injury risk (Siegmund, Davis, et al., 2008). Experimental work on the neck muscle response of occupants in non-neutral postures is limited. Indeed, previous efforts to model occupants in non-neutral postures relied on a muscle activation scheme from head-forward rear impacts because of a lack of detailed neck muscle activation data (Shateri & Cronin, 2015). Some activation data are presented by Kumar et al., (2005), who reported increased peak muscle activity in the sternocleidomastoid on the contralateral side to the head-turned direction during rear impact. These authors examined axial head rotations of 45° in both directions, although recent natural driving data have shown that while their right axial rotation posture approximates a right mirror check (42.1°); their leftward rotation posture doesn’t match any commonly held postures while driving (Fice, Blouin, et al., 2018). In the present study, we investigated neck muscle responses of seated volunteers exposed to rear impacts with their heads in a neutral and multiple non-neutral postures. The non-neutral postures studied included a left shoulder check, left mirror check, rear-view mirror check and looking at a passenger with the head orientations for these postures taken from a prior naturalistic driving study (Fice, Blouin, et al., 2018). During the simulated impacts, neck muscle activity and head and torso kinematics were measured. We hypothesized that non-neutral head postures would increase pre-perturbation and impact-related peak neck muscle activity and alter the peak head/neck kinematic responses to simulated rear impacts. This experiment will provide data to improve human body models in non-neutral head postures and improve our understanding of injury risk of occupants in non-neutral head postures at the time of a rear impact. 62 5.3 Methods 5.3.1 Subjects Twelve subjects (Table 5.1) with no history of whiplash injury, neck/back pain, frequent/severe headaches, or neuromuscular injury participated in this study. Pre-existing neck pain was assessed using the Neck Disability Index, and we found a median (range) score of 1.5 (0-4) indicating the absence of disability due to neck pain amongst the subjects (Vernon & Mior, 1991). Subjects provided written informed consent prior to participating in the study, which was approved by the UBC Clinical Research Ethics Board and conformed to the Declaration of Helsinki. Table 5.1 - Anthropomorphic data of participants in this study. Volunteer Age (years) Height (cm) Weight (kg) Females 1 24 169 64 2 26 160 58 3 26 157 48 4 32 163 61 5 43 170 64 Mean (SD) 30.2 (7.8) 163.8 (5.6) 58.7 (6.6) Males 6 22 173 80 7 24 173 66 8 24 184 71 9 24 168 82 10 27 172 62 11 31 178 76 12 31 178 82 Mean (SD) 26.1 (3.6) 175.0 (5.3) 74.2 (8.1) 5.3.2 Instrumentation EMG activity was recorded with indwelling electrodes (0.05mm Stablohm 800A, California Fine Wire Company, Grover Beach, CA, USA) inserted bilaterally into the sternocleidomastoid (SCM), splenius capitis (SPL), semispinalis capitis (SSCAP), and C4 multifidus (MU) under ultrasound guidance (MicroMaXX, Sonosite Inc., Bothell, WA; Siegmund et al., 2007). Bipolar EMG recordings were performed with twisted electrode pairs that had a hooked tip with 2-4mm insulation removed. Electrodes were placed near the centre of each muscle’s horizontal cross section at approximately the C4-C5 level. In the SCM, the wires remained superficial to the readily identifiable cleidomastoid subvolume (Kamibayashi & Richmond, 1998). EMG signals were amplified (×100), notch filtered at 60Hz, and bandpass filtered between 50-2000Hz using a Neurolog system (Digitimer, Welwyn Garden City, Hertfordshire, UK). 63 Head acceleration was measured with a nine-accelerometer array (6 Kistler 8302B20S1; 20 g, Amherst, NY, USA; 1 Silicon Design 2220–010; 10 g, Issaquah, WA, USA; 2 Endevco, 7265A; 100g, Irvine, CA, USA) arranged in a 3-2-2-2 configuration (Padgaonkar et al., 1975) that measured both 3D angular and linear accelerations. These sensors were mounted to a headgear (Blouin, Siegmund, & Inglis, 2007) that was fitted tightly to the participant’s head. Torso kinematics were recorded using a triaxial accelerometer (Summit 34103A; 7.5g, Akron, OH) and a triaxial angular rate sensor (Y & Z axes DynaCube; 5730 deg/s, ATA Sensors, Albuquerque, NM; X-axis DTS ARS Pro-1500; 1500 deg/s, Seal Beach, CA, USA; Figure 5.1 for coordinate frame). These sensors were mounted to a plate positioned on the chest just below the sternal notch using tightly fitted shoulder straps and double-sided tape affixed to the skin. The relative motion overserved between the torso and head accelerometer mounts and the respective body parts they were mounted to was minimal during high speed video of a pilot subject. A motion capture system (Optotrak Certus, Northern Digital, Waterloo, ON, Canada) was used to measure head, torso, and sled displacements using infrared light emitting diode (IRLED) markers on each of the head accelerometer array, torso chest plate and car seat/sled platform. For the head accelerometer array, four markers were on a plate facing the camera when the subject looked straight ahead, and eight markers were arranged on two plates that aligned with the camera when the subjects rotated their head 45deg to the left or right (yaw). This arrangement of markers ensured that rigid body rotations of the head could be calculated for all head motions observed in this study. The location of the accelerometers and IRLED markers were digitized relative to anatomical landmarks so that the kinematics could be resolved to anatomically relevant locations (i.e. atlanto-occipital joint, head center of mass, and first thoracic vertebrae). Sled acceleration was measured with a uniaxial accelerometer (Silicon Design 2220–100; 100 g; Issaquah, WA, USA). All kinematic signals were filtered with a 500Hz low pass filter. EMG and kinematics were recorded simultaneously at 4,000 Hz using a PXI DAQ system and custom LabVIEW code (EMG: PXI-4495 24-bit; other signals: PXI-6289 18-bit, National Instruments, Austin, TX). Motion capture data were acquired at 200 Hz, and acquisition was triggered by the DAQ system to ensure synchronization. 64 Figure 5.1 - Experimental setup with a subject seated in a 2005 Volvo S40 drivers seat mounted to a feedback-controlled sled. The coordinate system is shown. Note the Y axis goes into the page, subject right. Green stars denote the locations used to calculate initial seatback angle. 5.3.3 Protocol Subjects were seated in a 2005 Volvo S40 driver seat that was mounted on a custom fabricated sled powered by feedback-controlled linear induction motors (Kollmorgen IC55-100A7, Kommack, NY; Siegmund et al., 2003a). The steering wheel was adjusted in the vertical and horizontal directions, as directed by each subject, to reflect how they would normally set their steering wheel position while driving, while ensuring that their elbows were bent in a relaxed position. The feet were supported by footplates that were angled 55° from horizontal and adjusted fore-aft to form a knee angle of 115°. A crash sound from an actual vehicle-to-barrier crash accompanied the perturbation (peak amplitude 109 dB, time-to peak 34 ms; Mang et al., 2012). The head restraint was removed, and no seat belts were used to simplify the boundary conditions for future modeling work. Subjects were exposed to rear impacts with an average speed change of 0.78 ± 0.004 m/s, peak acceleration of 2.1 ± 0.02 g, and duration of 64.4 ± 0.4 ms. The onset of the acceleration matched that of a vehicle-to-vehicle crash producing a speed change of 8 km/h (Siegmund et al., 1997). Before perturbations, subjects were instructed to hold the steering wheel and fully relax other than the muscle activity required to maintain the required head posture. Subject were unaware of the exact timing of the perturbations, although they were told it would occur within a 5-10s window. The four non-neutral postures studied in this experiment represent drivers (left-hand drive) performing a left shoulder check, left mirror check, rear-view mirror check, or looking at a passenger in the front 65 seat. The target head angles used to mimic these postures were determined experimentally in a prior naturalistic driving study (Table 5.2; Fice et al., 2018). Before the non-neutral posture trials, subjects were coached into the appropriate head posture using a custom Matlab (Mathworks, Natwick, MA, USA) program that plotted the 3D head angle of the subject calculated with the IRLEDs on the head accelerometer array. Head angle tracking was accomplished in real-time using an NDI C++ API (Northern Digital, Waterloo, ON, Canada). Due to the time required to configure the software, instruction was given for subjects to maintain this position for up to 30 s before perturbation. A small laser mounted to the head gear allowed subjects to approximately maintain their target head posture during this interval. Participants were exposed to nine whiplash-like perturbations: five neutral posture control trials that alternated with four non-neutral experimental trials. The non-neutral posture order was randomized. Habituation of muscle activity and kinematic responses to the sled perturbation was a potential confound in this experiment (Blouin et al., 2003; Siegmund et al., 2003b). We reduced the potential for habituation by giving participants at least five minutes between trials and not performing the same condition successively. Further, the experimental design with five exposures of the control trial bookending every non-neutral trial allowed the presence of habituation to be statistically tested and, if present, controlled for. After the perturbation trials, participants were seated with their torso strapped against a flat vertical surface in order to acquire maximum voluntary isometric contractions (MVIC) data. Participants performed one MVIC in ten directions: the six principal directions (flexion, extension, left/right lateral bending, and left/right axial rotation) and four directions comprising of equal combinations of flexion or extension and lateral bending (Fice et al., 2014). Participants wore a tightly fitting helmet (The Classic, Pro-Tec, Santa Fe Springs, CA, USA) which was fixed to a 6-axis load cell (Model 45E15, JR3 Incorporated, Woodland, CA) mounted above the subject. Visual feedback (Fice et al., 2014) and verbal encouragement were provided. Indwelling muscle activity was recorded in the same manner (filters, DAQ, sampling frequency) as done during the perturbation trials. The MVIC data were used to normalize the EMG recorded during the perturbations. 5.3.4 Data Analysis To remove sled electrical artifacts and movement artifacts, EMG data were bandpass filtered between 50-1000Hz with a zero-phase 8-pole Butterworth filter and a zero phase IIR comb filter with 200Hz increments and a Q factor of 50. All EMG data were normalized by the largest RMS (20ms moving window) calculated for each muscle across all MVIC directions. Pre-impact EMG was calculated as the 66 root-mean-square (RMS) of the EMG data over a 100ms window before the initiation of sled acceleration. Peak EMG was calculated as the maximum 20ms moving windows during the first 400ms after perturbation. Note that we did not remove pre-impact activity from the EMG signals, because we were interested in studying this activity, and as a result the peak EMG is an absolute measure rather than relative to pre-impact values. EMG onset was calculated using an approximate generalized log-likelihood ratio approach designed specifically for EMG signals (Staude, 2001; Staude & Wolf, 1999). These onset times were visually confirmed and manually adjusted 7% of the time. Onsets were omitted in trials where the normalized peak EMG was less than 5% MVIC above pre-impact levels, which occurred in <1% of trials. Due to EMG technical difficulties and artefacts, we removed the R SSCAP in the looking-at-passenger trial in one subject; the L MU in the looking-at-passenger trial plus the L MU and L SSCAP in three trials (left-mirror-check, control 3 & 4) for another subject; the L SSCAP in the left-shoulder-check trial for a third subject; and, the L MU and L SSCAP in four trials (left-shoulder-check, look-at-passenger, and control 3 & 4) for a fourth subject. Accelerometer signals in the head and torso were first filtered with an 8-pole zero phase 100Hz low pass filter. Next, gravity was removed from the head and torso accelerometers using the Optotrak data to calculate the sensors orientation as a function of time (Blouin, Siegmund, & Inglis, 2007). The bias of the accelerometers and angular rate sensors was then removed using the mean signals 100ms before impact. Then all the head and torso linear and angular kinematics were resolved in a lab fixed coordinate system with the Z-axis aligned with gravity (positive values downwards), the X-axis aligned with the sled axis (positive values forward), and the Y-axis with positive values to the right (Figure 5.1). Rigid body transformations were then used to resolve the head accelerations to the head centre of mass (CG) (Blouin, Siegmund, & Inglis, 2007), which was estimated to lie in the midsagittal plane, rostral to the interaural axis by 17% of the distance between the interaural axis and the vertex (NASA, 1978). Head angular acceleration was computed from the 9-accelerometer array (Padgaokar, 1975) and then integrated to calculate angular velocity. Assuming the torso deformations are minimal, we again used rigid body transformation to resolve torso accelerations to the C7-T1 joint, which was assumed to be midway between the sternal notch and C7 spinous process (Queisser et al., 1994). For this transformation, angular velocities were differentiated to calculate angular accelerations and then both were low-pass filtered at 30Hz with a zero-phase 8-pole Butterworth filter. Finally, only head accelerations were also resolved to a local coordinate system aligned and rotating with the Frankfort plane, with the X-axis forwards, Y-axis towards the right ear, and Z-axis downwards. 67 Tri-axial peak values and their timings were calculated for the kinematic variables (See Figure 5.2 for the peaks used). The tri-axial onsets of head and torso linear acceleration and angular velocity were calculated using a finite difference approach (Siegmund, 2001), before visual confirmation which resulted in adjustment in 7% of cases. This onset algorithm, designed for movement signals, eliminates the delayed bias that would be introduced with the use of a threshold value for onset detection and instead uses a threshold value on the phase-advanced signals determined with finite differentiation using an appropriately selected window size, see Siegmund (2001) for further details. Tri-axial peak head and torso angular displacements were calculated from the Optotrak data and were reported relative to their initial positions, defined as the average position over the 100ms before impact. The angles were calculated as intrinsic Euler angles with a yaw-pitch-roll (Z-Y’-X’’) decomposition order. The tri-axial peak head linear displacements were defined as the maximum distance between the C7-T1 joint and the atlanto-occipital joint (C0-C1) in the lab reference frame. C0-C1 was defined as 24mm posterior and 37mm inferior to the head CG (Siegmund et al., 2007). Initial head linear position was quantified as the average over 100ms before impact of the X distance between the head CG and the seat hinge and the average Y direction between the head CG and the centre of the steering wheel. Initial head angle was calculated as the component wise average over the 100ms before impact of the 3D Euler angles of the local head X-axis, which was defined as a vector in the mid sagittal plane aligned with the Frankfort plane (plane passing through external auditory meatus and inferior orbital rims) and pointing forward. Euler angles were defined as intrinsic angles with a yaw-pitch-roll (Z-Y’-X’’) decomposition order. Additionally, for the non-neutral conditions we calculated the initial head angles relative to neutral posture to facilitate a comparison with the target head orientations. This was done by calculating relative initial 3D Euler angles (intrinsic Z-Y’-X’’) of the Frankfort vector between the neutral looking-straight-ahead posture and the non-neutral postures recorded in subsequent trials. To aid reconstruction of the boundary conditions used in this experiment for computational human body models the arm and elbow angles were reported. Initial arm angles were calculated using points digitized on the shoulder (deltoid tuberosity of the humerus), elbow (lateral epicondyle), and hand (metacarpophalangeal joint of the middle finger). Initial upper arm angle was calculated as the angle between the Z axis and the vector connecting the shoulder and elbow points. Similarly, initial elbow angle was defined as the angle between the elbow/shoulder and elbow/hand vectors. Note that these initial arm angles were measured once before the first trial for each subject and potentially do not represent changes due to the experimental conditions or trial-to-trial variations. Initial seatback angle was calculated as the average over all subjects and trials of the mean angle in the 68 100ms preceding perturbation between vertical and the line connecting the seatback hinge and seatback IRLED markers in the X-Z plane (see Figure 5.1 for marker locations). 5.3.5 Statistics Habituation is a potential confound in this study (Blouin et al., 2003; Siegmund et al., 2003b), and thus we first tested for its presence in the following variables: peak EMG of each muscle and the 3D components of peak lab fixed head acceleration, torso acceleration, head & torso angular rate, head & torso rotation, and C0C1-T1 relative displacement. We compared the five control trials for each variable with a separate analysis using a linear mixed model with exposure number as a repeated-measure fixed effect and a heterogenous compound symmetric covariance matrix. This model was chosen to avoid losing statistical power due to missing data points that would result from a repeated-measures ANOVA. The covariance matrix was chosen to improve model fit (assessed with Akaike's Information Criterion) and avoid convergence failure. The residuals from these models were found to deviate from a normal distribution in 23 / 29 variables (8 muscles + 7 kinematic variables × 3 components) using the Kolmogorov-Smirnov test. We then performed a rank transform (Conover, 2012) to all the data, repeated the analysis, and found that the residuals were normally distributed in all cases. We found evidence of habituation (i.e. main effect P>0.05) in 3 / 29 variables, including the peak L SSCAP activity (F4,18.2=4.68, P=0.009), head X acceleration (F4,27.1=5.82, P=0.002), and head Z angular velocity (F4,26.6=3.17, P=0.030). When comparing the largest and smallest median response in these variables, the L SSCAP activity decreased by 46.3% MVIC (%difference: 85.5%) from the third control trial to the last control trial; head X acceleration decreased by 0.35g (%difference: 12.1%) from the second control trial to the last control trial; and head Z angular rate decreased by 24.0°/s (%difference: 29.3%) from the second control trial to the last control trial. Compared to the data reported in Blouin et al. (2003) and Siegmund et al. (2003b), the 3/29 variables showing habituation was considered minimal evidence of habituation and thus we continued the analysis assuming no habituation, i.e., control response was calculated as the average of the five control trials. We hypothesized that the pre-impact and peak muscle activity in our eight muscles would increase in the experimental vs. control trials. Further, we hypothesized that the peak head/neck kinematics would be altered in the experimental vs. control trials. The variables of interest for this latter hypothesis were the peak tri-axial components of lab fixed and local head acceleration, torso acceleration, head & torso angular rate, head & torso angular displacement, and C0C1-T1 relative displacement. Additional variables of interest included the onset of EMG and onset of head & torso acceleration and angular rate. 69 All dependent variables were tested with separate linear mixed models with condition as a repeated-measure fixed effect. The covariance matrices were set to heterogenous compound symmetric to avoid convergence failure and improve model fit, which was assessed with Akaike's Information Criterion. The residuals from these models were found to deviate from a normal distribution using the Kolmogorov-Smirnov test in 32 / 60 variables. We then performed a rank transform to the data, repeated the analysis, and found that the residuals were normally distributed in all cases. The significance criterion for the main effects of the kinematic 3D components was adjusted for the three comparisons using a Bonferroni correction. When a significant main effect was found, pairwise comparisons of the experimental vs control conditions were performed using Bonferroni adjustment for multiple comparisons. To test if the head angles relative to neutral posture before impact in each of our experimental conditions differed from the target angles, we performed separate one-sample t-tests against the target value (Fice et al., 2018) for each component of the 3D Euler angle for each non-neutral condition. Data analyses were performed with MATLAB 2017a (Mathworks, Natwick, MA, USA) and statistical analyses were performed with SPSS v25 (IBM, Armonk, NY, USA). Significance was set at P<0.05 before appropriate correction for multiple comparisons. 5.4 Results For most conditions and axes, the pre-impact head orientations were generally similar to the target directions (Table 5.2). Deviations from the target angles were observed for the left shoulder check (head yaw angle 2.7° less leftwards, head roll angle 8.7° further leftwards), left mirror check (head roll angle 2.2° further leftwards), rear-view mirror check (head yaw angle 1.4° further rightwards). The position of the head centre of gravity and orientation of the Frankfort plane before the impact varied between conditions (Table 5.3). The average elbow angle measured before the first perturbation in the neutral posture was 114.0 ± 9.5°, and the angle between the upper arm and vertical was 52.0 ± 7.0°. In the X-Z plane, the initial angle between vertical and the seatback was 27.5 ± 0.2° averaged over all trials and subjects. Pre-impact muscle activity in the neutral condition ranged from 1.2% (R SSCAP) to 2.4% MVIC (L & R SCM), and non-neutral postures increased this for several muscles (Figure 5.3, Table 5.4). Increases in pre-impact activity compared to the neutral condition for the left shoulder condition were observed in the R SCM (Δ7.5% MVIC), L SPL (Δ2.9% MVIC), R SSCAP (Δ0.9% MVIC), and R MU (Δ4.0% MVIC); the left 70 mirror condition in the R SCM (Δ2.4% MVIC) and L SPL (Δ2.9% MVIC); and the passenger condition for L SCM (Δ3.2% MVIC), R SPL (Δ1.0% MVIC). The L SCM in the left mirror condition was the only case that pre-impact muscle activity dropped when compared to the neutral case (Δ-0.4% MVIC). L SSCAP and L MU did not show any significant differences in pre-impact muscle activity for neutral vs. non-neutral conditions. For the perturbations in the neutral posture, peak muscle activity ranged from 35% MVIC (R SPL) to 101% MVIC (R SCM) (Figure 5.3, Table 5.4). The L MU during the left mirror condition was the only muscle to show a significant change from the neutral condition (12% MVIC less activity). Median muscle onset times in the neutral posture ranged from 65ms (L SCM) to 88ms (L SPL) (Table 5.5), and only in R SCM during the left shoulder condition did the onset times significantly change (13ms earlier). In the neutral posture, participants exhibited a typical rear impact kinematic response, with linear torso X acceleration leading linear head X acceleration followed by C0C1-T1 X displacement, head Y angular velocity and Y rotation (Figure 5.2, central panel). These motions were generally limited to the sagittal plane, but in non-neutral postures there were widespread significant deviations in head kinematics beyond the sagittal plane (Figure 5.4, Table 5.6). Comparing the left shoulder and neutral conditions, we observed more lab reference head leftward Y acceleration (Δ-1.8g), less C0C1-T1 X displacement (Δ7.8mm), more rightward X & leftward Z angular velocity (Δ225.0 & -287.6°/s), decreased Y angular velocity (Δ-88.7°/s), more leftward head X rotation (Δ-23.0°), and less backwards Y rotation (Δ-16.0°). In the passenger condition compared to neutral, the lab reference head Y acceleration increased leftwards (Δ-0.9g), C0C1-T1 X displacement decreased (Δ5.2mm), head X & Z angular velocities increased rightwards (Δ95.0 & Δ173.1°/s), and head rotation increased rightwards in X (Δ17.4°) and decreased backwards in Y (Δ-9.6). The head kinematics changes for left mirror and rear-view mirror conditions were generally in the same direction as the left shoulder and passenger conditions respectively, but with less widespread significant differences and smaller change magnitudes (Table 5.6). Peak torso kinematics remained mostly unchanged, except for the left shoulder condition which compared to the neutral condition saw increased leftward X rotation (Δ-3.4°), rightward X-axis angular velocity (Δ-75.8°/s), and leftward Z-axis angular velocity (Δ -64.1°/s). Compared to the neutral condition kinematic onsets occurred earlier in the left shoulder condition for torso Y acceleration (Δ 18ms) and X angular velocity (Δ 24ms), and in the passenger condition for head X acceleration (Δ 8ms) and torso X angular velocity (Δ 23ms) (Table 5.7). 71 Table 5.2 - The target and actual head orientation of subject before the experimental conditions. The mean (standard deviation) are presented, and orientation data are presented as intrinsic Euler angles (Z-Y’-X’’). The ‘target initial orientation’ is the angle we attempted to guide participants to ~20s before impact with respect to a neutral looking straight ahead posture; ‘Actual initial orientation’ is the average head orientation 100ms before impact again with respect to a neutral looking straight ahead posture. *Denotes a significant difference between the target and actual initial orientation in a one-sample t-test (P<0.05;). Condition Axis Target initial orientation (°) Actual initial orientation (°) One-sample t-test Results Left Shoulder Z -79 -76.3 (2.3) t11=3.93; P=0.002* Y -8 -8.1 (2.3) t11=-0.11; P=0.912 X -2 -10.7 (5.2) t11=-5.77; P=0.000* Left Mirror Z -35 -35.2 (1.9) t11=-0.34; P=0.744 Y -3 -2.8 (2.3) t11=0.37; P=0.722 X -2 -4.2 (2.6) t11=-2.96; P=0.013* Rear Mirror Z 15 16.4 (2.0) t11=2.43; P=0.033* Y 2 1.1 (2.8) t11=-1.05; P=0.316 X 2 1.7 (2.2) t11=-0.51; P=0.623 Passenger Z 55 54.4 (2.5) t11=-0.82; P=0.429 Y -1 -1.3 (2.9) t11=-0.41; P=0.693 X 7 7.7 (2.5) t11=0.97; P=0.351 Table 5.3 - Initial position and orientation of the head before impact in each of the conditions tested. The mean (standard deviation) are presented, and orientation data are presented as intrinsic Euler angles (Z-Y’-X’’). ‘Initial Position’ is the average location of the head centre of gravity from 100ms before impact in the lab coordinate system and is relative to a point aligned with the seat hinge in X & Z and the centre of the steering wheel in Y; ‘Local Head X-axis orientation’ is the orientation of a vector in the midsagittal plane of the head parallel to the Frankfort plane facing forward and with respect to the lab frame of reference. Condition Axis Initial Position (mm) Local Head X-axis orientation (°) Neutral X -62.5 (25.4) 2.8 (2.3) Y 0.4 (9.6) 12.7 (6.3) Z -668.2 (32.8) -0.3 (3.4) Left Shoulder X -73.1 (26.4) -8.1 (5.1) Y -59.4 (17.4) 4.4 (5.3) Z -661.6 (33.2) -79.8 (3.1) Left Mirror X -59.6 (25.7) -1.3 (3.4) Y -21.0 (14.3) 10.4 (6.5) Z -665.0 (34.1) -35.9 (3.8) Rear Mirror X -64.1 (27.7) 5.0 (3.0) Y 10.4 (15.7) 15.0 (5.5) Z -669.0 (33.7) 17.0 (3.0) Passenger X -66.1 (22.8) 10.5 (3.5) Y 32.7 (34.2) 10.3 (6.7) Z -664.8 (34.6) 56.1 (3.7) 72 Table 5.4 - The median (1st, 3rd quartile) pre-impact and peak normalized RMS EMG activity for the muscles in the neutral, left shoulder check, left mirror check, rear-view mirror check, and look-at-passenger experimental trials. L, left; R, right; SCM, sternocleidomastoid; SPL, splenius capitis; SSCAP, semispinalis capitis; MU, multifidus. *Denotes a significant (P<0.05) main effect. For the pairwise post-hoc tests, significance (P<0.05) is denoted by ‘#’ between the neutral and experimental cases. Muscle Neutral Left Shoulder Left Mirror Rear-Mirror Passenger Linear Mixed Model Results Pre-impact SCM L 0.024 (0.017, 0.028) 0.020 (0.018, 0.026) 0.020 (0.016, 0.022)# 0.023 (0.021, 0.031) 0.056 (0.048, 0.069)# F4,16=19.31; P=0.000* R 0.024 (0.019, 0.032) 0.099 (0.067, 0.127)# 0.048 (0.026, 0.055)# 0.026 (0.018, 0.037) 0.020 (0.014, 0.029) F4,22.2=18.09; P=0.000* SPL L 0.015 (0.013, 0.026) 0.044 (0.031, 0.088)# 0.021 (0.017, 0.033)# 0.014 (0.013, 0.030) 0.016 (0.013, 0.028) F4,18.4=26.75; P=0.000* R 0.013 (0.009, 0.016) 0.014 (0.009, 0.016) 0.014 (0.010, 0.016) 0.014 (0.011, 0.017) 0.023 (0.020, 0.044)# F4,23=9.72; P=0.000* SSCAP L 0.017 (0.010, 0.024) 0.016 (0.008, 0.034) 0.014 (0.008, 0.027) 0.022 (0.010, 0.027) 0.028 (0.010, 0.038) F4,31=1.37; P=0.268 R 0.012 (0.009, 0.015) 0.021 (0.012, 0.037)# 0.011 (0.009, 0.025) 0.013 (0.009, 0.015) 0.014 (0.011, 0.017) F4,32.6=3.26; P=0.023* MU L 0.017 (0.013, 0.025) 0.012 (0.009, 0.020) 0.014 (0.009, 0.017) 0.017 (0.012, 0.020) 0.023 (0.016, 0.041) F4,27.2=5.41; P=0.002* R 0.019 (0.011, 0.050) 0.059 (0.020, 0.076)# 0.036 (0.013, 0.068) 0.014 (0.010, 0.047) 0.015 (0.010, 0.042) F4,24.9=8.47; P=0.000* Peak SCM L 0.99 (0.82, 1.43) 1.15 (0.51, 1.89) 0.72 (0.57, 1.01) 0.97 (0.83, 1.11) 1.45 (0.90, 1.83) F4,20.2=3.37; P=0.029* R 1.01 (0.75, 1.35) 1.21 (0.66, 1.78) 1.17 (0.81, 1.43) 0.97 (0.67, 1.33) 0.77 (0.52, 1.17) F4,22.1=1.60; P=0.209 SPL L 0.59 (0.44, 0.94) 0.83 (0.55, 1.21) 0.45 (0.39, 1.03) 0.54 (0.40, 1.23) 0.95 (0.46, 1.33) F4,20.5=1.73; P=0.182 R 0.35 (0.16, 0.81) 0.42 (0.15, 0.84) 0.41 (0.17, 0.69) 0.38 (0.14, 0.82) 0.62 (0.32, 0.78) F4,21.7=1.06; P=0.402 SSCAP L 0.43 (0.22, 0.99) 0.59 (0.31, 1.16) 0.44 (0.31, 0.77) 0.46 (0.19, 1.08) 0.68 (0.34, 0.86) F4,22.2=0.93; P=0.466 R 0.52 (0.29, 0.70) 0.30 (0.24, 0.45) 0.37 (0.23, 0.51) 0.49 (0.39, 0.57) 0.54 (0.32, 0.63) F4,24.4=2.88; P=0.044* MU L 0.48 (0.36, 0.82) 0.44 (0.30, 0.80) 0.36 (0.22, 0.59)# 0.49 (0.25, 1.03) 0.55 (0.40, 1.06) F4,15.9=3.90; P=0.022* R 0.56 (0.39, 1.05) 0.62 (0.27, 1.04) 0.59 (0.31, 1.11) 0.43 (0.37, 0.80) 0.43 (0.25, 0.68) F4,28.3=0.73; P=0.577 73 Table 5.5 - The median (1st, 3rd quartile) onset of EMG activity for the muscles in the neutral, left shoulder check, left mirror check, rear-view mirror check, and look-at-passenger experimental trials. L, left; R, right; SCM, sternocleidomastoid; SPL, splenius capitis; SSCAP, semispinalis capitis; MU, multifidus. *Denotes a significant (P<0.05) main effect. For the pairwise post-hoc tests, significance (P<0.05) is denoted by ‘#’ between the neutral and experimental cases. Muscle Neutral Left Shoulder Left Mirror Rear-Mirror Passenger Linear Mixed Model Results SCM L 65 (56, 73) 67 (62, 78) 71 (66, 76) 61 (55, 72) 63 (50, 68) F4,16.6=2.72; P=0.066 R 66 (62, 73) 53 (50, 61)# 66 (61, 69) 73 (55, 76) 73 (59, 78) F4,17.3=5.11; P=0.007* SPL L 88 (58, 125) 54 (46, 58) 62 (52, 113) 57 (54, 132) 74 (55, 137) F4,18.4=2.60; P=0.070 R 68 (48, 108) 56 (49, 80) 61 (52, 77) 57 (49, 60) 51 (33, 67) F4,25.5=1.06; P=0.394 SSCAP L 68 (52, 131) 117 (51, 158) 67 (53, 128) 80 (59, 141) 88 (59, 129) F4,25.6=1.63; P=0.198 R 87 (71, 112) 59 (50, 134) 68 (51, 116) 77 (51, 139) 70 (54, 148) F4,18.2=0.41; P=0.800 MU L 72 (57, 101) 56 (43, 127) 69 (51, 153) 57 (52, 74) 62 (49, 126) F4,26.2=0.72; P=0.588 R 77 (57, 102) 80 (49, 177) 58 (47, 92) 79 (48, 174) 59 (46, 132) F4,17.9=0.85; P=0.513 74 Table 5.6 - The median (1st, 3rd quartile) peak kinematic variables in the neutral, left shoulder check, left mirror check, rear-view mirror check, and look-at-passenger experimental trials. CG, centre of gravity; C0C1, joint of occiput and first cervical vertebrae; T1, first thoracic vertebrae; a, linear acceleration; Δd, linear displacement; ω, angular velocity; Δθ angular displacement. *Denotes a significant main effect (P<0.0167, corrected for multiple comparisons). For the pairwise post-hoc tests, significance (P<0.0167; adjusted for multiple comparisons) is denoted by ‘#’ between the neutral and experimental cases. Kinematic Variable Neutral Left Shoulder Left Mirror Rear-Mirror Passenger ANOVA Results Head CG a (g) X 2.9 (2.5, 3.3) 3.3 (2.7, 3.6) 3.0 (2.4, 3.3) 3.0 (2.4, 3.5) 3.0 (2.7, 3.6) F4,20.4=1.25; P=0.323 Y 0.1 (-0.4, 0.5) -1.7 (-2.2, -1.1)# -0.9 (-1.1, -0.5)# -0.6 (-0.8, -0.5)# -0.8 (-1.1, -0.6)# F4,20.8=44.72; P=0.000* Z 1.4 (1.0, 1.8) 0.9 (0.7, 1.1)# 1.0 (0.9, 1.5) 1.2 (1.1, 1.7) 1.2 (1.0, 1.3) F4,27.3=5.82; P=0.002* Head CG Local a (g) X 2.4 (1.9, 2.6) 2.0 (1.4, 2.4)# 2.0 (1.5, 2.2)# 2.3 (1.8, 2.6) 1.6 (1.4, 1.6)# F4,18.1=29.11; P=0.000* Y 0.3 (-0.1, 0.5) 2.7 (2.3, 3.0)# 1.6 (1.3, 1.8)# -0.9 (-1.1, -0.7)# -2.3 (-2.6, -1.8)# F4,20.7=200.78; P=0.000* Z 2.1 (1.7, 2.4) 1.8 (1.6, 2.1) 1.8 (1.5, 2.1) 2.2 (1.7, 2.7) 1.9 (1.6, 2.3) F4,23=2.07; P=0.117 Torso T1 a (g) X 2.2 (1.8, 2.6) 2.1 (1.8, 2.5) 2.0 (1.8, 2.5) 2.4 (1.9, 2.8) 2.1 (1.8, 2.6) F4,28.4=3.99; P=0.011* Y 0.0 (-0.1, 0.4) -0.5 (-0.8, 0.1) 0.4 (-0.7, 0.4) 0.3 (-0.5, 0.6) -0.7 (-0.8, -0.6)# F4,21.2=14.24; P=0.000* Z 2.2 (1.7, 2.3) 1.7 (1.5, 2.0) 2.0 (1.7, 2.5) 2.2 (1.3, 2.8) 1.9 (1.5, 2.2) F4,22.9=3.06; P=0.037 C0C1-T1 Δd (mm) X -27.6 (-32.6, -22.9) -19.8 (-23.5, -16.0)# -24.4 (-30.4, -22.1) -27.3 (-32.4, -21.3) -22.4 (-26.5, -19.4)# F4,24.6=18.99; P=0.000* Y 0.7 (-4.0, 3.4) 8.0 (6.0, 9.7)# 4.9 (-6.8, 8.6) 0.3 (-8.1, 5.6) -11.6 (-13.3, 8.5) F4,26.4=3.67; P=0.017* Z 12.9 (7.1, 17.3) 9.6 (7.9, 12.6) 9.9 (7.5, 16.5) 11.7 (9.4, 15.0) 11.4 (8.7, 13.8) F4,26.6=1.63; P=0.197 Head ω (°/s) X -33.2 (-60.7, 5.2) 191.8 (160.5, 234.1)# 92.5 (51.0, 120.7)# -46.3 (-86.8, -36.7) 61.8 (47.0, 99.6)# F4,20.4=78.84; P=0.000* Y 411.4 (376.1, 486.0) 322.7 (283.4, 420.7)# 421.3 (330.0, 469.5) 399.5 (334.3, 478.6) 356.6 (316.7, 399.2)# F4,18.9=7.24; P=0.001* Z -73.8 (-107.2, -37.5) -361.4 (-414.8, -332.4)# -196.0 (-252.0, -128.6)# -59.9 (-72.6, -42.0) 99.3 (42.2, 149.4)# F4,23.8=92.30; P=0.000* Torso ω (°/s) X -0.8 (-30.2, 42.1) 76.6 (72.7, 119.4)# 46.2 (34.5, 71.0) -38.9 (-63.1, 41.3) -47.6 (-103.9, 69.0) F4,22.8=23.94; P=0.000* Y 117.0 (101.8, 158.2) 114.8 (86.9, 138.4) 98.8 (92.1, 143.2) 133.3 (90.4, 163.4) 105.6 (93.6, 125.4) F4,19.7=1.33; P=0.294 Z 1.0 (-36.1, 17.1) -64.1 (-80.9, -52.1) -49.6 (-85.2, -40.7)# 28.6 (-15.4, 51.5) 54.6 (-72.0, 66.4) F4,23.9=9.45; P=0.000* Head Δθ (°) X 2.3 (1.1, 3.7) -20.7 (-26.6, -19.3)# -11.6 (-16.9, -10.0)# 9.3 (7.3, 11.1)# 19.4 (17.2, 21.9)# F4,20.6=226.36; P=0.000* Y 22.8 (21.2, 28.1) 6.8 (5.1, 9.2)# 20.4 (17.3, 22.3)# 21.1 (19.6, 25.8) 13.4 (10.9, 15.4)# F4,13.9=101.85; P=0.000* Z 7.7 (5.7, 8.8) 8.6 (5.2, 9.8) 5.2 (4.0, 7.4) 6.3 (4.7, 9.0) -1.9 (-4.9, 6.4) F4,19.8=3.70; P=0.021 Torso Δθ (°) X 1.3 (-0.7, 1.8) -2.1 (-3.1, -1.9)# -0.9 (-1.7, 0.9) 1.4 (0.9, 1.9) 2.9 (1.6, 3.4) F4,20.8=30.87; P=0.000* Y 11.0 (8.8, 13.2) 8.1 (7.3, 11.1) 9.9 (8.6, 13.5) 10.3 (8.0, 12.5) 10.3 (8.0, 12.9) F4,21.8=1.59; P=0.213 Z -0.8 (-2.6, 2.6) 2.9 (-4.2, 4.2) -2.8 (-5.0, -1.2) 1.5 (-1.4, 3.8) 0.1 (-3.9, 3.5) F4,29.8=4.23; P=0.008* 75 Table 5.7 - The median (1st, 3rd quartile) onset of select kinematic variables in the neutral, left shoulder check, left mirror check, rear-view mirror check, and look-at-passenger experimental trials. CG, centre of gravity; T1, first thoracic vertebrae; a, linear acceleration; ω, angular velocity. *Denotes a significant main effect (P<0.0167, corrected for multiple comparisons). For the pairwise post-hoc tests, significance (P<0.05) is denoted by ‘#’ between the neutral and experimental cases. Kinematic Variable Neutral Left Shoulder Left Mirror Rear-Mirror Passenger ANOVA Results Head CG a X 38 (33, 52) 53 (41, 57) 44 (38, 63) 34 (28, 40) 30 (26, 32)# F4,19.6=21.46; P=0.000* Y 41 (38, 56) 39 (36, 46) 33 (29, 39) 46 (31, 50) 50 (38, 56) F4,19.7=2.32; P=0.093 Z 30 (25, 34) 28 (21, 36) 26 (23, 32) 30 (24, 33) 30 (27, 38) F4,19=1.16; P=0.360 Torso T1 a X 26 (24, 28) 31 (24, 35) 28 (25, 31) 26 (22, 28) 26 (23, 31) F4,19.1=1.13; P=0.370 Y 50 (44, 59) 32 (18, 40)# 55 (40, 62) 48 (39, 65) 36 (26, 60) F4,17.1=9.55; P=0.000* Z 33 (30, 37) 37 (35, 38) 37 (32, 42) 31 (26, 38) 36 (30, 41) F4,18.5=3.15; P=0.039 Head ω X 54 (50, 67) 46 (42, 53) 47 (39, 51) 69 (52, 80) 53 (50, 59) F4,20.1=4.96; P=0.006* Y 51 (45, 56) 46 (38, 52) 50 (47, 58) 49 (43, 58) 42 (31, 56) F4,18.4=0.73; P=0.581 Z 72 (52, 86) 55 (42, 71) 51 (48, 60) 62 (49, 80) 40 (31, 62) F4,19.1=3.02; P=0.044 Torso ω X 68 (62, 79) 46 (43, 52)# 67 (56, 95) 69 (55, 83) 45 (41, 70)# F4,20.1=12.23; P=0.000* Y 35 (32, 37) 34 (27, 37) 33 (31, 37) 37 (31, 43) 37 (33, 40) F4,17.7=1.25; P=0.327 Z 76 (68, 85) 57 (41, 86) 92 (59, 100) 81 (53, 93) 61 (40, 76) F4,20.9=2.77; P=0.054 76 Figure 5.2 - Exemplar traces for electromyographic and triaxial kinematic data from a single subject during rear impact with their head in the neutral posture or postures mimicking a left shoulder check, left mirror check, rear-view mirror check, or talking to their passenger (right side). Vertical scale bars are aligned with the onset of perturbation. L, left; R, right; SCM, sternocleidomastoid; SPL, splenius capitis; SSCAP, semispinalis capitis; MU, multifidus; a, linear acceleration; Δd, linear displacement; α, angular acceleration; ω, angular velocity; Δθ, angular displacement relative to initial position; CG, center of gravity; C0C1, atlanto-occipital joint; T1, first thoracic vertebrae. Kinematics are given in a coordinate system with X positive forwards, Y positive rightwards, and Z positive down (Figure 5.1). The coordinate system used is fixed to the lab for all kinematics except the Local Head CG acceleration which is aligned with the Frankfort plane and rotates with head motion. Head and torso rotations are presented as intrinsic Euler angles with a decomposition order of Z-Y’-X’’. 77 Figure 5.3 - Median normalized pre-impact and peak RMS EMG of the eight muscles studied while participant’s heads were in the neutral posture or postures mimicking a left shoulder check, left mirror check, rear-view mirror check, or looking at the passenger (right side) during rear impact. Error bars show the 25th and 75th quartiles of subject data. L, left; R, right; SCM, sternocleidomastoid; SPL, splenius capitis; SSCAP, semispinalis capitis; MU, multifidus. *Denotes significant differences (P<0.05) for main effects of head posture in the linear mixed model, and horizontal bars denote significant (P<0.05) post-hoc pairwise comparisons between the neutral and experimental conditions. 78 Figure 5.4 - Median peak values of kinematic measures during rear impact while participant’s heads were in the neutral posture or postures mimicking a left shoulder check, left mirror check, rear-view mirror check, or looking at the passenger (right side). Error bars show the 25th and 75th quartiles of subject data. a, linear acceleration; Δd, change in distance; α, angular acceleration; ω, angular velocity; Δθ, change in angular position relative to initial position; CG, center of gravity; C0C1, atlanto-occipital joint; T1, first thoracic vertebrae. Kinematics are given in a lab fixed coordinate system the X (positive forwards), Y (positive rightwards), and Z (positive down) dimensions (Figure 5.1), except for the Local Head CG acceleration which is aligned with the Frankfort plane and rotates with head motion. *Denotes significant differences (P<0.0167, corrected for multiple comparisons) for main effect of head posture in the linear mixed model, and horizontal bars denote significant (P<0.05) post-hoc pairwise comparisons between the neutral and experimental conditions. 79 5.5 Discussion The goal of this study was to examine how vehicle occupant head postures affected their head/neck kinematics and neck muscle responses to rear impact. Apart from L SSCAP and L MU, muscle activity before the impact increased by 0.4 to 7.5% of MVIC in non-neutral versus neutral postures. The increases in pre-impact muscle activity were generally largest for the left shoulder and passenger conditions. During impact, only the L MU showed a decrease of 12% of MVIC in the left mirror compared to neutral condition. The kinematic responses differed between the non-neutral vs neutral postures, with many of the changes in motions outside of the sagittal plane. Based on these observations, we partially accept our initial hypotheses: pre-impact muscle activity increased, and peak kinematic responses were altered by initial head posture, but peak muscle activity during impact was not significantly different. Taken together, our results suggest that changes in muscle activity during the impact may not contribute to kinematic changes or the increased injury risk associated with non-neutral posture rear impact (Jakobsson et al., 2008), but posture-related differences in pre-impact muscle activity may alter load paths and, along with altered initial postures, may modify head/neck kinematics during the impact under non-neutral-posture rear impacts. Our data can be used to design novel injury prevention methods or devices and to validate or improve computer simulations of human responses when in non-neutral postures before rear impact (See Appendix Figure A.1, Figure A.2, and Figure A.3 for kinematic response corridors). The magnitudes of the observed pre-impact muscle activity increases were limited to 0.4% to 7.5% of MVIC, and it is unclear if these modest levels of activity, alone, were sufficient to stiffen joints and alter kinematics. Although, this initial muscle-related force on the spine may potentially change joint angles and the resulting load path through the spine, contributing to the altered kinematics and the increased potential for injury during non-neutral posture rear impact (Jakobsson et al., 2008). The pre-impact muscle activity in SCM was generally largest for contralateral axially rotated postures, whereas SPL activity occurred in ipsilateral axially rotated postures. This set of findings is consistent with the recruitment of these muscles during isometric axial moment generation (Fice, Siegmund, et al., 2018; Vasavada et al., 2002). Further, the activity in the SSCAP during isometric axial moment generation has been shown to be minimal but ipsilateral (Fice, Siegmund, et al., 2018; Vasavada et al., 2002). This is consistent with our finding for L SSCAP activity that did not significantly increase in non-neutral vs neutral postures, but inconsistent with the R SSCAP activity increasing in the left shoulder condition. The inconsistency in the R SSCAP pre-impact activity in the left shoulder check condition may be due to 80 comparing neutral posture biomechanics to an axially rotated posture, but computational modelling approaches have found that the SSCAP axial moment capacity increases in ipsilateral axially rotated postures and decreased in contralateral postures (Vasavada et al., 1998). It is possible that R SSCAP is providing a stabilization role counter to its biomechanics, which is consistent with findings that a muscle’s recruitment is not always aligned with its apparent biomechanical role (Fice, Siegmund, et al., 2018). Recruitment patterns for MU in axial rotation are not available for comparison to our results. The lack of changes in peak muscle activity evoked during the perturbation between non-neutral vs neutral postures (except for a 12% MVIC decrease in L MU during the left mirror condition) may not rule out their role in altering strains and stresses on the cervical spine when occupants are in non-neutral postures. Consistent level of activation may not produce a consistent force due to muscle length / velocity and force relationships of active human muscle (Hill, 1938; Winters & Woo, 1990). Also, changes in the moment arms as a result of changed postures (Vasavada et al., 1998) may influence muscle-induced moments. Further, these potentially altered active muscle forces/moments will now be acting over joints that could be pre-strained due to intervertebral orientations resulting from non-neutral postures (Siegmund, Davis, et al., 2008). These potential effects could be explored further in computational models of the human body. The decrease in activity of the L MU during the left mirror condition was of particular interest because of its direct insertion on the capsular ligaments (J. S. Anderson et al., 2005; Winkelstein et al., 2001), which have been linked to the source of whiplash injury in 54-60% of cases (Barnsley et al., 1995; Lord et al., 1996). It has been proposed that activation of the multifidus muscle during rear impact may exacerbate capsular ligament strains (Mang et al., 2015; Siegmund, Blouin, et al., 2008), and evidence from animal models (Dong & Winkelstein, 2010; Lee et al., 2004; Lu et al., 2005), cadaveric studies (B. Deng et al., 2000; Ivancic et al., 2008; Pearson et al., 2004), and computational studies (Cronin, 2014; Fice et al., 2011; Stemper et al., 2005) suggests that increased capsular strain increases the risk of whiplash injury. Increased strain in the capsular ligament has been linked with non-neutral head-turned postures in cadaver models (Siegmund, Davis, et al., 2008), but our results suggest that MU activation in non-neutral postures may not be exacerbating this effect. In both left shoulder and passenger conditions we recorded decreased retraction (C0C1-T1 X displacement) and decreased head Y angular displacements compared to neutral posture tests. The decreased retraction and angular displacements were evidence of stiffening of both the rotational and linear sagittal plane head/neck response. Non-neutral axially rotated postures may bring into contact 81 adjacent surfaces of the zygapophyseal (or facet) joint and pre-strain the capsular ligament. During impact these structures potentially cause the observed increase in head-neck stiffness. In neutral posture rear-impact, reduced retraction is potentially associated with decreased capsular ligament strain (Siegmund, Myers, et al., 2001; Winkelstein et al., 2000). Although, non-neutral postures in whiplash type motions are in fact associated with increased capsular ligament strain and potentially increased risk of injury (Siegmund, Davis, et al., 2008). The difference potentially lies in pre-straining of the capsular ligament before impact. Beyond the sagittal plane in the left shoulder and passenger conditions, we saw increases in the lab reference head Y acceleration, X angular velocity, and X angular displacement. These motions are evidence of the coupled nature of cervical spine motions. Non-neutral postures will place the head centre of gravity out of the mid-sagittal plane and the inertial loads during rear impact will then cause further axial moments to be applied through the neck when the ligaments are already pre-strained. Due to the coupled nature of the motion of cervical spinal segments, the axial moment may lead to ipsilateral lateral bending moments of the segments (Bogduk & Mercer, 2000). This phenomenon may explain why we saw increased head Y acceleration and X rotations in our participants during non-neutral vs neutral posture rear impact. Computational human body models are a developing tool used by researchers to improve the safety of vehicles and improve understanding of the injury mechanisms in various crash modes. Muscle activation in these models has been shown to be important to capture the kinematics of humans during an impact (Brolin et al., 2005; Fice et al., 2011) and before an impact (Östh et al., 2015). Most recently, closed-loop PID feedback control has been used to model human responses in long duration braking events that precede some collisions (Meijer et al., 2013; Östh et al., 2015, 2012). Even these most advanced muscle activation controllers require tuning of feedback gains to match specific types of load cases. Muscle activation controllers also require validation data to ensure that muscle activation matches humans to ensure that the controller is not masking deficits in the mechanics of the underlying model. With the current data, researchers will be able to use the muscle activation as a function of time along with the kinematic corridors (Appendix Figure A.1, Figure A.2, and Figure A.3) to tune and/or validate their models of rear impact for the non-neutral postures presented. With the assumptions required to model the human body, the researchers could then increase impact severity to study injury mechanics of non-neutral postures more closely. Previous efforts to model occupants in non-neutral postures had to rely 82 on a muscle activation scheme from head forward rear impacts because of a lack of data (Shateri & Cronin, 2015). A limitation of this work is that only one automotive seat, one impact direction, and one impact severity were tested, and it is not certain that our results apply more generally. Further, the magnitude of the perturbation was limited to 0.78 m/s speed change and a peak acceleration of 2.1g, which was chosen to limit the risk of injury to our participants. It is not known how these results scale to more severe and injurious impact scenarios. Another limitation is the assumption of a rigid body between the torso accelerometer mount and the C7-T1 joint axis, which means that deformation of the torso between the sternum and C7-T1 will alter our torso kinematic measurements. Also, participants were aware that a perturbation would occur within a 5-10s window, which we did to ensure a consistent timing between conditions and because it was not feasible to have participants maintain the left shoulder check posture longer than this. Therefore, the results may not represent an unaware driver (Siegmund et al., 2003a). Finally, we do not have enough participants to separate our results for males and females, and females have been shown to have a higher risk of whiplash associated disorders (Carstensen et al., 2012; Jakobsson et al., 2004; Krafft et al., 2003). In summary, we described the head and torso kinematics and neck muscle responses of volunteers when they were exposed to a rear impact whiplash-like perturbation while their head was in a neutral posture and four non-neutral rotated-head postures drivers commonly adopt. When compared to a neutral head postures, we found that non-neutral postures increased the pre-impact activity of neck muscles and during impact they decreased the activity of the left multifidus for the left mirror check posture. Non-neutral postures led to widespread increases in accelerations and angular velocities of the head and angular velocities of the torso in directions outside the sagittal plane. These results provide knowledge of how different occupant postures before an impact can change their neuromuscular and kinematic responses to an impact. This work provides data that can inform injury prevention methods or devices and can be used by computational models to improve the modeling of drivers in non-neutral head postures before impact. 5.6 Data Availability Since data collected in this study will be useful to modelling researchers, we ensured that the we have permission from subjects and ethics approval to share the anonymized data collected in this study. The data will be available to download by contacting Dr. J.-S. Blouin (jsblouin@mail.ubc.ca). 83 Chapter 6. Neck muscle and head/neck kinematic responses while bracing against the steering wheel during frontal and rear impacts 6.1 Preamble A participant’s knowledge of an impending impact has been shown to influence their neck muscle responses and head/neck kinematics. Although, the effects of bracing oneself by pushing against the steering wheel on neck muscle responses and head/neck kinematics are less well understood. In this study, we measured the neck muscle responses and head kinematics of volunteers who pushed against the steering wheel before a frontal or rear-impact. In both chapters 5 and 6, the goal was to understand how occupant pre-impact factors (i.e. head posture and bracing) influenced their head and neck kinematic and neck muscle responses to whiplash-like perturbations. This chapter focuses on the effects bracing have on the kinematic and muscles responses. 6.2 Introduction Whiplash associated disorders (WAD) occur more commonly than any other motor vehicle injury (Quinlan et al., 2004; Styrke et al., 2012). Both frontal and rear impacts are known to cause WAD, accounting respectively for 23-38% and 38-52% of reported cases (Beattie & Lovell, 2010; Berglund et al., 2003; Galasko et al., 1993). In frontal impact, bracing against the steering wheel has been shown to significantly increase the risk of WAD (24 vs 38%; Jakobsson et al., 2004). With all collision directions combined, however, the proportion of WAD patients with neck or back pain symptoms were not associated with bracing before impact (Beattie & Lovell, 2010). Further, Beeman et al., (2011) showed that bracing prior to frontal impacts reduced displacements of the head and torso, suggesting lower cervical strains in braced compared to relaxed postures. These latter findings are potentially at odds with the increased injury risk from epidemiological evidence (Jakobsson et al., 2004). The role of impact direction is unknown in these discrepancies. Further, neck muscle activity is potentially linked to whiplash injury risk (Mang et al., 2012), and the role of neck muscle activity on the injury risk of braced occupants is unclear. The goals of this study were to re-examine braced volunteers in frontal impact and to extend these findings to rear impact and characterize neck muscle involvement. When volunteers maximally tensed their neck muscles before a rear, frontal, or side impact, head rotation occurred earlier and reached a lower peak value (Ejima et al., 2012, 2007; Ono et al., 1997). This observation has been supported by head-neck model simulations under similar circumstances in rear and frontal impacts (Dibb et al., 2013; Stemper et al., 2006), suggesting that pre-tensing neck muscles 84 before an impact may reduce injury risk. Occupants, however, may not fully tense their neck muscles before an impact, but rather brace themselves by pushing against the steering wheel and straightening their arms (Choi et al., 2005; Hault-Dubrulle et al., 2011). This type of bracing may involve neck muscle activity, due to the known link between neck muscle activity and isometric shoulder exertions (Karimi et al., 2016; Rahnama et al., 2015). In the present experiment, we exposed volunteers to frontal and rear whiplash-like perturbations while their upper extremities were either relaxed on the steering wheel or braced by pushing on the steering wheel. During these impacts, we measured the head and neck muscle responses in healthy participants. We expected participants would generate neck muscle activity due to bracing forces exerted on the steering wheel (Karimi et al., 2016; Rahnama et al., 2015). Thus, we hypothesized that pre-impact muscle activity would increase in the braced compared to relaxed trials. Further, it was expected that the increased pre-impact activity would lead to increased impact related neck muscle activity in braced trials compared to relaxed trials. Finally, we hypothesized that braced compared to relaxed trials would reduce displacements, as per Beeman et al. (2011), and increase acceleration and angular velocity of the torso and head. 6.3 Methods 6.3.1 Subjects Eleven subjects (Table 6.1) with no history of whiplash injury, neck/back pain, frequent/severe headaches, or neuromuscular injury participated in this study. Before the experiment, subjects were screened for neck pain using the Neck Disability Index. We found a median (range) score of 1 (0-3) indicating the absence of disability due to neck pain amongst the subjects (Vernon & Mior, 1991). Subjects provided written informed consent prior to participating in the study, which was approved by the UBC Clinical Research Ethics Board and conformed to the Declaration of Helsinki. 85 Table 6.1 - Anthropomorphic data of participants in this study. Volunteer Age (years) Height (cm) Weight (kg) Females 1 23 169 64 2 25 160 59 3 31 165 55 Mean (SD) 26.3 (4.2) 164.7 (4.5) 59.3 (4.5) Males 4 23 168 82 5 24 172 69 6 28 180 86 7 29 187 81 8 30 178 75 9 30 180 117 10 30 175 82 11 55 189 79 Mean (SD) 31.1 (10.0) 178.6 (7.1) 83.9 (14.3) 6.3.2 Instrumentation EMG activity was recorded with indwelling electrodes (0.05mm Stablohm 800A, California Fine Wire Company, Grover Beach, CA, USA) inserted unilaterally into the sternohyoid (STH), sternocleidomastoid (SCM), splenius capitis (SPL), semispinalis capitis (SSCAP), semispinalis capitis (SSCERV), C4 multifidus (MU), levator scapulae (LS), and trapezius (TRAP) under ultrasound guidance (MicroMaXX, Sonosite Inc., Bothell, WA; Siegmund et al., 2007). Bipolar EMG recordings were done with twisted electrode pairs that had a hooked tip with 2-4mm insulation removed. Electrodes were placed near the centre of each muscle’s horizontal cross section at approximately the C4-C5 level. In the SCM, the wire electrodes always remained superficial to the readily identifiable cleidomastoid subvolume (Kamibayashi & Richmond, 1998). EMG signals were amplified (×100), notch filtered at 60Hz, and bandpass filtered between 50-2000Hz using a Neurolog system (Digitimer, Welwyn Garden City, Hertfordshire, UK). In four subjects, the TRAP and LS were bandpass filtered between 10-2000Hz due to a failure in the 50Hz high pass filter. Head acceleration was measured with a nine-accelerometer array (8 Kistler 8302B20S1; 20 g, Amherst, NY; and 1 Silicon Design 2220–010; 10 g, Issaquah, WA) arranged in a 3-2-2-2 configuration (Padgaonkar et al., 1975) that measured both 3D angular and linear accelerations. These sensors were mounted to a headgear (Blouin, Siegmund, & Inglis, 2007) that was fitted tightly to the participant’s head. Torso kinematics were recorded using a triaxial accelerometer (Summit 34103A; 7.5g, Akron, OH) and a triaxial angular rate sensor (Y & Z axes DynaCube; 5730 deg/s, ATA Sensors, Albuquerque, NM; X-axis DTS ARS Pro-1500; 1500 deg/s, Seal Beach, CA, USA; Figure 6.1 for coordinate frame). These sensors were 86 mounted to a plate that was firmly affixed to the chest just below the sternal notch using adjusted shoulder straps and double-sided tape affixed to the skin. The relative motion overserved between the torso and head accelerometer mounts and the respective body parts they were mounted to was minimal during high speed video of a pilot subject. A motion capture system (Optotrak Certus, Northern Digital, Waterloo, ON, Canada) was used to measure head, torso, and sled displacements using four infrared light emitting diode (IRLED) markers on each of the head accelerometer array, torso chest plate, and car seat/sled platform. In addition, individual markers were placed on the right seat back (lateral side upper portion), shoulder (deltoid tuberosity of the humerus), elbow (lateral epicondyle), and hand (metacarpophalangeal joint of the middle finger). The location of the accelerometers and IRLED markers were digitized relative to anatomical landmarks so that the kinematics could be resolved to anatomically relevant locations (i.e. atlanto-occipital joint, head center of mass, seventh cervical and first thoracic vertebral joint). Sled acceleration was measured with a uniaxial accelerometer (Silicon Design 2220–100; 100 g). Accelerometer and angular rate signals were filtered with a 500Hz low pass filter. The force applied by the participant onto the steering wheel during bracing was measured with a 6 DoF load cell (Model 45E15, JR3 Incorporated, Woodland, CA) mounted between the steering wheel and its support structure, and these signals were filtered with a 1000Hz low pass filter. EMG, accelerometers, angular rate sensors, and force signals were recorded simultaneously at 4,000 Hz using a PXI DAQ system and custom LabVIEW code (EMG: PXI-4495 24-bit & other: PXI-6289 18-bit, National Instruments, Austin, TX). Motion capture data were acquired at 200 Hz, and each frame acquisition was triggered by the DAQ system to ensure synchronization. Figure 6.1 - Experimental setup with subject seated in a 2005 Volvo S40 drivers seat mounted to a feedback-controlled sled. The coordinate system is shown, and the Y axis comes out of the page. Green stars denote the locations used to calculate seatback angle. 87 6.3.3 Protocol After EMG wire insertion, seated participants performed maximum voluntary isometric contractions (MVIC) in ten directions: the six principal directions (flexion, extension, left/right lateral bending, and left/right axial rotation) and four directions comprising of equal combination of flexion/extension and lateral bending. Subjects sat with their torso constrained against a flat seat back and wore a tightly fitting helmet (The Classic, Pro-Tec, Santa Fe Springs, CA, USA) which was fixed to an inverted 6-axis force plate (OR6-7-1K-3985 AMTI Watertown, MA, USA) mounted above the subject. Visual feedback (Fice et al., 2014) and verbal encouragement were provided. These data were used to normalize the EMG recorded during perturbations. While subjects were seated in the car seat they performed an MVIC pushing in a forward direction against the steering wheel that was mounted to a 6 DoF load cell (Model 45E15, JR3 Incorporated, Woodland, CA). The seat back was mechanically locked in place to prevent rearward deflection. The target bracing load during braced perturbation trials was 60% of the resultant MVIC force. The 60% MVIC was chosen during pilot work as an upper limit that subjects could sustain for up to 15s before impact. Subjects were given two trials without perturbations to practise generating and maintaining 60% MVIC bracing load with the aid of verbal feedback on their current steering wheel loads from the experimenter. Participants were seated in a 2005 Volvo S40 driver’s seat that was mounted on a custom-built feedback-controlled sled powered by linear induction motors (Kollmorgen IC55-100A7, Kommack, NY; Siegmund et al., 2003a, 2003b; Figure 6.1). The steering wheel height and fore-aft position were adjusted for each participant to reflect how they would normally set their steering position when driving, while ensuring that their elbow was bent in a relaxed position. The feet were supported by footplates angled 55° from horizontal which were adjusted fore-aft to form a knee angle of 115°. A crash sound recorded from an actual vehicle-to-barrier crash (peak amplitude 109 dB, time-to peak 34 ms; Mang et al., 2012) accompanied the perturbation. The head restraint was removed, and no seat belts were used to simplify the boundary conditions for future modeling work. For all trials in both frontal and rear impacts, the sled perturbation was consistent with a speed change of 0.77 ± 0.03 m/s, peak acceleration of 19.9 ± 0.9 m/s2, and duration of 65.5 ± 1.2 ms. The onset of the acceleration matched that of a vehicle-to-vehicle crash producing a speed change of 8 km/h (Siegmund et al., 1997). In all trials, participants looked straight ahead before the perturbations and they were aware that the perturbation would occur within a 5-10s window. 88 In the experimental conditions, perturbations were performed with subjects maintaining one of two upper extremity postures: relaxed hands on wheel (relaxed) and braced with hands pushing against the wheel (braced). For the braced trials, the experimenter provided verbal instructions before the perturbation to help participants attain the target bracing load (60% MVIC). Habituation of muscle activity and kinematic responses to the sled perturbation was a potential confound in this experiment (Blouin et al., 2003; Siegmund et al., 2003b). To test the possibility of habituation, perturbations with the subjects relaxed hands in lap (control) were performed before, and after each experimental condition. The rear and frontal impacts were performed in separate blocks and the order of the blocks was randomly chosen. For each impact direction five perturbations were performed, with three control trials surrounding and separating the two randomly ordered experimental trials. The three exposures of the control trial (per impact direction) allowed the presence of habituation to be assessed and controlled for if present. Further, we reduced the potential for habituation by giving subjects at least five minutes between trials and not performing the same condition successively. 6.3.4 Data Analysis Steering wheel loads were filtered with a 30Hz zero-phase low-pass 8-pole Butterworth filter, which was chosen because the steering wheel loads were only used for setting the bracing level and those were isometric contractions. The loadcell offsets, which included the weight of the wheel & mountings, where removed by subtracting the average loadcell readings 4-5s before perturbation from all the hand on lap trials for each subject. Bracing MVIC was calculated as the maximum 100ms moving window mean resultant force during the pushing forward against the wheel MVIC trial. For the bracing trials, the percent of MVIC bracing was calculated as the mean resultant moment applied to the steering wheel during the 100ms before perturbation divided by the bracing MVIC. To remove sled electrical artifacts and movement artifacts, EMG data were bandpass filtered between 50-1000Hz with an 8-pole Butterworth filter and an IIR comb filter with 200Hz increments and a Q factor of 50, both zero-phase. Pre-impact EMG was calculated as the root-mean-square (RMS) of the EMG data in a 100ms window before the initiation of sled acceleration, which was normalized by the largest RMS (20ms moving window) calculated for each muscle across all the MVIC directions. Peak EMG was calculated as the maximum 20ms moving windows again normalized by the MVIC RMS EMG. EMG onset was calculated using an approximate generalized log-likelihood ratio approach (Staude, 2001; Staude & Wolf, 1999). These onset times were visually confirmed and adjusted for 7% of the trials. Onsets were omitted in trials where the normalized peak EMG was less than 0.05 above preimpact levels, which 89 occurred in 4% of the trials. Due to movement artifacts that saturated the pre-amplifier, we omitted the EMG data from the following number of subjects (trials) for each muscle 1 (2) SCM, 2 (3) SPL, 2 (3) SSCAP, 1 (2) SSCERV, 3 (3) MU, 2 (2) LS, and 3 (9) TRAP. Head and torso kinematics were presented in a lab fixed coordinate system with the Z-axis aligned with gravity (positive values downwards), the X-axis aligned with the sled axis (positive values forward), and the Y-axis with positive values to the right. Gravity was removed from head and torso accelerometers using the Optotrak data to calculate the sensors orientation as a function of time (Blouin, Siegmund, & Inglis, 2007). The bias of the accelerometers and angular rate sensors was then removed using the mean of each signal over the 100ms preceding impact. Rigid body transformations were used to resolve the head accelerations to the head centre of mass (CG) (Blouin, Siegmund, & Inglis, 2007), which was estimated to lie in the midsagittal plane, rostral to the interaural axis by 17% of the distance between the interaural axis and the vertex (NASA, 1978). Head angular acceleration was computed from the 9-accelerometer array (Padgaokar, 1975) and then integrated to calculate angular velocity. Assuming the torso deformations are minimal, we again used rigid body transformation to resolve torso accelerations to the seventh cervical and first thoracic vertebral joint (C7-T1), which was assumed to be midway between the sternal notch and C7 spinous process (Queisser et al., 1994). For this transformation, angular velocities were differentiated to calculate angular accelerations, and both were low-pass filtered at 30Hz with a zero-phase 8-pole Butterworth filter. Due to a suspected faulty ground connection that caused a floating DC offset on several measurement channels, the kinematic data from one subject were omitted. Peak values and their timing were calculated for the head and torso acceleration and angular velocity (see Figure 6.2 for the peaks used). The onsets of these variables were calculated using a finite difference approach (Siegmund, 2001), before visual confirmation which resulted in adjustment in 5% of cases. Peak head and torso angular position and their timings were calculated from the Optotrak data and were calculated relative to the initial position, defined as the average position 100ms before impact. The peak head retraction was defined as the maximum X-axis relative distance between the C7-T1 joint and the atlanto-occipital joint (AOJ), which was defined as 24mm posterior and 37mm inferior to the head CG (Siegmund et al., 2007). Initial head posture was quantified as the average X distance between head CG and the seat hinge from 100ms before impact and the average Y angle of the Frankfurt plane (plane passing through external auditory meatus and inferior orbitals) over the same time. Initial upper arm angle was calculated as the 90 average angle between Z-axis and the vector connecting the shoulder and elbow IRLED markers during the 100ms before perturbation. Similarly, initial elbow angle was defined as the average during the 100ms before impact of the angle between the vectors made up of elbow and shoulder markers and the elbow and hand markers. Preimpact seatback angle was calculated as the average angle 100ms before impact of a line connecting the digitized seat hinge location and seatback IRLED (Figure 6.1 for reference locations) from the Z-axis in the sagittal plane. 6.3.5 Statistics Habituation is a potential confound in this study (Blouin et al., 2003; Siegmund et al., 2003b). We first tested for its presence in the following variables: peak EMG of the eight muscles and peak head and torso acceleration, angular rate, angular displacement, and retraction in both rear and frontal impact. To test for habituation, we compared the three control trials for each variable with a separate analysis for rear and frontal impact using a linear mixed model with control trial as a repeated measure fixed effect and an unstructured covariance matrix. This model was chosen to avoid losing statistical power due to our missing data points, and the covariance matrix utilized improved model fit (assessed with Akaike's Information Criterion). The residuals from this model were found to deviate from a normal distribution using the Kolmogorov-Smirnov test in 15 of the 30 variables. We then performed a rank transform to all the data, repeated the analysis, and found that the residuals were normally distributed in all cases. We found evidence of habituation (i.e. main effect P0.05) in 4 of the 30 variables: peak SPL activity (F2,10=7.31; P=0.011), torso acceleration (F2,9=18.78; P=0.001) and head angular displacement (F2,9=5.58; P=0.027) in rear impact; and head angular displacement (F2,9=6.23; P=0.020) in frontal impact. When comparing the largest and smallest median responses in these variables, the SPL activity in rear impact decreased by 8.2% MVIC (%difference: 40.3%) from the second to final trial; torso X acceleration in rear impact increased by 0.30g (%difference: 14.3%) from the first to last trial; head angular displacement in rear impact increased by 0.63° (%difference: 3.3%) from the first to second trial; head angular displacement in frontal impact decreased by 3.47° (%difference: 24.6%) from the first to last trial. Compared to the data reported in Blouin et al. (2003) and Siegmund et al. (2003b) the 4/30 variables showing habituation was considered minimal evidence of habituation and thus we continued the analysis assuming no habituation, i.e. the control response was calculated as the average of the three control trials. We hypothesized that the pre-impact and peak muscle activity in our eight muscles would increase in the braced vs. relaxed trials. Further, we hypothesized that the peak head/neck displacements would 91 decrease while accelerations and angular velocities would increase in the braced vs. relaxed trials. The variables for these hypotheses were pre-impact and peak EMG activity, and peak head & torso X acceleration, Y angular velocity, Y angular displacement, and X retraction (Figure 6.1 for coordinate system). In addition, we were also interested in the onset of EMG activity, the timing of these kinematics variables, and the onset of head and torso acceleration and angular rate. In the variables of interest, we tested the effect of bracing (i.e. relaxed vs. braced trials) using a repeated measures linear mixed model with an unstructured covariance matrix, and we performed separate analyses for each variable and impact direction. The residuals from this model were found to deviate from a normal distribution using the Kolmogorov-Smirnov test in 42 of 84 variables, so the tests were rerun with all the data rank transformed and all residuals were not different from normal. This latter analysis is presented in the rest of this manuscript. To test if the initial head posture varied between trials, we performed separate linear mixed models for head X position and Y angle with initial arm position as one repeated measures factor (five levels: 3 controls, relaxed, braced) and impact direction (rear/frontal) as another. To test if the arm postures of the braced and relaxed trials were different, we performed separate linear mixed model analyses for upper arm and elbow angle, with relaxed/braced as one repeated measures factor, and impact direction as the other. Finally, the seatback angle before impact was tested for a difference between the braced and relaxed trials using a linear mixed model with relaxed/braced as one repeated measures factor, and impact direction as the other. These analyses used a compound symmetry covariance matrix to improve model fit as per Akaike's Information Criterion, and the residuals were normally distributed as per the Kolmogorov-Smirnov test. Data analyses were performed with MATLAB 2017a (Mathworks, Natwick, MA, USA) and statistical analyses were performed with SPSS v25 (IBM, Armonk, NY, USA). Significance was set at P<0.05. Data will be presented as either mean ± standard deviation when a normal distribution was assumed or median (1st, 3rd quartile) when it was not. 6.4 Results Subjects adopted similar head postures before control trials and relaxed trials, but not before the braced trials (head position: F4,94=44.98, P<0.0001; head angle: F4,94=17.49, P<0.0001). For the braced trials, the average initial head COG position (relative to the seat hinge) and Frankfort plane angles were -107.8 ± 35.7mm and 7.9 ± 6.1° compared to -61.5 ± 26.0mm and 13.5 ± 4.5° for the other trials (multiple 92 P<0.0004). No significant differences in head postures were observed between control and relaxed trials (multiple P0.05). Impact direction did not influence initial head position (F1,94=0.022, P=0.883) or angle (F1,94=0.806, P=0.371). During braced trials, subjects had more shoulder flexion and less elbow flexion when compared to relaxed trials. The angle between vertical and the upper arm was on average 53.3 ± 6.6° before relaxed trials but was 76.0 ± 7.4° before braced trials (F1,31=154.19, P<0.0001). Similarly, the mean elbow angle was more flexed (112.8 ± 11.6°) before relaxed trials than before braced trials (135.6 ± 18.6°; F1,31=49.04, P<0.0001). Upper arm (F1,31=0.050, P=0.825) and elbow (F1,31=0.023, P=0.880) angle did not change with impact direction. The bracing MVIC resultant force while pushing on the wheel was 1130 ± 292 N. While bracing, subjects exerted an average of 57.5 ± 6.5% of their bracing MVIC on the steering wheel before impact, which corresponded to an average resultant force of 651 ± 192 N. Bracing caused seat back rotation, with an average pre-impact seatback angle of 28.9 ± 0.6° for braced compared to 27.5 ± 0.4° for relaxed trials (F1,31=259.82, P<0.0001). Impact direction did not significantly alter the pre-impact seatback angle (F1,31=2.46, P=0.127). Pre-impact muscle activity was generally low (all muscles, 75th percentile <4% MVIC) for relaxed trials, and bracing caused widespread increases in both frontal and rear impacts (Table 6.2; Figure 6.2; Figure 6.3). The median (1st, 3rd Quartile) of the subject-wise increase for braced compared to relaxed trials for muscles that reached significance was 1.1 (0.3, 2.3)% MVIC for rear and 1.5 ( 0.7, 4.1)% MVIC for frontal impacts, while the STH & TRAP for rear impact and SSCERV, MU, & TRAP for frontal impact were not significantly different (Table 6.2; Figure 6.4). During impact, all the neck muscles exhibited robust activity in response to the perturbations, with the smallest median peak activity of 12 (7, 39)% MVIC from the SSCAP in the control condition in rear impact, and the largest of 114 (42, 198)% MVIC from the SCM in the control condition during frontal impact (Table 6.3; Figure 6.4). In rear impact, bracing did not lead to any significant changes in peak activity for any muscles. In frontal impact, the median TRAP and MU muscle activity in the braced trials dropped by 51 and 21% MVIC compared to the relaxed trials (Table 6.3; Figure 6.4). The impact related onset of muscle activity was generally consistent in braced compared to relaxed trials, and no statistically significant differences were found for any muscle or impact direction (Table 6.4). Braced posture compared to relaxed led to widespread changes in the kinematics of volunteers in both frontal and rear impacts (Figure 6.2; Figure 6.3). One of those changes was earlier onset of the head (Δ-93 33ms in frontal) and torso acceleration (Δ-11 & -17ms in rear & frontal) in braced compared to relaxed trials (Table 6.5). Peak kinematic measures were affected by bracing when compared to relaxed in all cases except head and torso angular velocity, head angle, and retraction in rear impact, and torso acceleration in frontal impact (Table 6.5; Figure 6.5). When comparing median values for braced to relaxed trials in rear impact, head acceleration decreased (Δ-0.4g), torso acceleration increased (Δ1.1g), and the angular displacement of the torso decreased (Δ-8.2°). For the same comparison in frontal impact, peak head acceleration increased (Δ0.4g), peak head (Δ124.1°/s) and torso (Δ100.5°/s) angular velocity increased, peak head (Δ1.5°) and torso (Δ10.1°) angular displacement increased, and retraction decreased (Δ10.1mm). The timing of these kinematic peaks occurred generally sooner for braced vs relaxed trials (Table 6.6; Figure 6.5). In both rear and frontal impact, braced compared to relaxed peak kinematics occurred earlier for the head (Δ34ms rear & Δ66ms frontal) and torso (Δ17ms rear & Δ58ms frontal) acceleration, head angular velocity (Δ38ms rear & Δ48ms frontal), head angle (Δ46ms rear & Δ108ms frontal), and in frontal impact peak retraction occurred earlier (Δ26ms). 94 Table 6.2 - The median (1st, 3rd Quartile) pre-impact normalized RMS EMG activity for the muscles tested in both rear and frontal impact. STH, sternohyoid; SCM, sternocleidomastoid; SPL, splenius capitis; SSCAP, semispinalis capitis; SSCERV, semispinalis cervicis; MU, multifidus; LS, levator scapulae; TRAP, Trapezius. *Denotes a significant (P<0.05) difference between the relaxed and braced trials in the repeated measures linear mixed model analysis. Rear Impacts Frontal Impacts Muscle Control Relaxed Braced Relaxed vs. Braced Statistics Control Relaxed Braced Relaxed vs. Braced Statistics STH 0.015 (0.012, 0.018) 0.017 (0.010, 0.021) 0.035 (0.015, 0.077) F1,10=5.20; P=0.046* 0.014 (0.007, 0.017) 0.012 (0.007, 0.016) 0.033 (0.019, 0.116) F1,10=12.91; P=0.005* SCM 0.020 (0.011, 0.027) 0.018 (0.011, 0.031) 0.026 (0.022, 0.058) F1,9.5=15.37; P=0.003* 0.015 (0.009, 0.024) 0.014 (0.007, 0.022) 0.027 (0.021, 0.067) F1,10=22.38; P=0.001* SPL 0.008 (0.005, 0.016) 0.008 (0.004, 0.012) 0.034 (0.015, 0.045) F1,10=15.33; P=0.003* 0.009 (0.007, 0.016) 0.009 (0.006, 0.010) 0.021 (0.013, 0.043) F1,9.8=23.82; P=0.001* SSCAP 0.007 (0.002, 0.010) 0.005 (0.003, 0.010) 0.009 (0.005, 0.020) F1,10=44.79; P=0.000* 0.007 (0.004, 0.012) 0.007 (0.003, 0.008) 0.012 (0.004, 0.029) F1,8.2=27.67; P=0.001* SSCERV 0.006 (0.003, 0.009) 0.006 (0.003, 0.008) 0.012 (0.008, 0.019) F1,10=23.29; P=0.001* 0.006 (0.005, 0.013) 0.009 (0.006, 0.011) 0.013 (0.008, 0.027) F1,10=5.28; P=0.044* MU 0.004 (0.004, 0.010) 0.004 (0.003, 0.012) 0.015 (0.011, 0.023) F1,9=12.16; P=0.007* 0.009 (0.006, 0.028) 0.009 (0.006, 0.024) 0.013 (0.006, 0.030) F1,9.2=0.25; P=0.629 LS 0.009 (0.005, 0.016) 0.006 (0.005, 0.018) 0.023 (0.021, 0.115) F1,10=9.72; P=0.011* 0.011 (0.006, 0.014) 0.011 (0.005, 0.020) 0.059 (0.022, 0.085) F1,10=11.29; P=0.007* TRAP 0.014 (0.013, 0.016) 0.013 (0.010, 0.020) 0.015 (0.012, 0.020) F1,9.1=1.90; P=0.201 0.018 (0.013, 0.031) 0.016 (0.010, 0.030) 0.015 (0.010, 0.021) F1,9.5=0.26; P=0.624 95 Table 6.3 - The median (1st, 3rd Quartile) peak normalized RMS EMG activity for the muscles tested in both rear and frontal impact. STH, sternohyoid; SCM, sternocleidomastoid; SPL, splenius capitis; SSCAP, semispinalis capitis; SSCERV, semispinalis cervicis; MU, multifidus; LS, levator scapulae; TRAP, Trapezius. *Denotes a significant (P<0.05) difference between the relaxed and braced trials in the repeated measures linear mixed model analysis. Rear Impacts Frontal Impact Muscle Control Relaxed Braced Relaxed vs. Braced Statistics Control Relaxed Braced Relaxed vs. Braced Statistics STH 0.63 (0.50, 0.92) 0.59 (0.51, 0.75) 0.67 (0.43, 1.07) F1,10=0.23; P=0.639 0.53 (0.43, 0.83) 0.58 (0.26, 0.88) 0.54 (0.37, 0.81) F1,10=0.23; P=0.638 SCM 0.89 (0.66, 1.04) 0.91 (0.71, 1.10) 0.72 (0.49, 0.93) F1,9.4=2.51; P=0.146 1.14 (0.42, 1.98) 0.84 (0.46, 1.42) 0.98 (0.51, 1.31) F1,10=0.02; P=0.887 SPL 0.21 (0.16, 0.26) 0.21 (0.15, 0.46) 0.26 (0.17, 0.39) F1,10=1.36; P=0.271 1.08 (0.30, 1.22) 0.71 (0.23, 1.64) 0.40 (0.31, 0.69) F1,9.3=1.64; P=0.232 SSCAP 0.12 (0.07, 0.39) 0.20 (0.05, 0.50) 0.27 (0.06, 0.37) F1,10=0.05; P=0.835 0.37 (0.24, 0.56) 0.19 (0.09, 0.58) 0.27 (0.08, 0.38) F1,8=4.67; P=0.063 SSCERV 0.29 (0.10, 0.46) 0.31 (0.10, 0.46) 0.20 (0.15, 0.58) F1,10=1.78; P=0.212 0.36 (0.29, 0.63) 0.52 (0.29, 0.68) 0.41 (0.15, 0.76) F1,9.4=1.29; P=0.284 MU 0.21 (0.17, 0.71) 0.24 (0.12, 0.59) 0.29 (0.16, 0.41) F1,9.7=0.46; P=0.515 0.45 (0.37, 1.56) 0.49 (0.42, 2.27) 0.28 (0.25, 0.68) F1,9.1=13.10; P=0.005* LS 0.24 (0.14, 0.35) 0.28 (0.11, 0.40) 0.31 (0.16, 0.45) F1,10=0.55; P=0.477 0.62 (0.25, 0.90) 0.44 (0.24, 0.60) 0.29 (0.21, 0.46) F1,10=0.83; P=0.385 TRAP 0.45 (0.23, 0.87) 0.61 (0.21, 0.92) 0.41 (0.19, 0.52) F1,9.3=2.82; P=0.126 0.80 (0.48, 2.20) 0.96 (0.34, 1.42) 0.45 (0.22, 0.57) F1,9.4=11.88; P=0.007* 96 Table 6.4 - The median (1st, 3rd quartile) onset of RMS EMG activity and select kinematic measures in both rear and frontal impact. STH, sternohyoid; SCM, sternocleidomastoid; SPL, splenius capitis; SSCAP, semispinalis capitis; SSCERV, semispinalis cervicis; MU, multifidus; LS, levator scapulae; TRAP, trapezius; CG, centre of gravity; T1, first thoracic vertebrae; a, linear acceleration; ω, angular velocity. *Denotes a significant (P<0.05) difference between the relaxed and braced trials in the repeated measures linear mixed model analysis. Rear Impacts Frontal Impacts Control Relaxed Braced Relaxed vs. Braced Statistics Control Relaxed Braced Relaxed vs. Braced Statistics EMG Onset (ms) STH 57 (55, 69) 59 (58, 65) 56 (51, 67) F1,10=0.89; P=0.368 61 (54, 73) 59 (54, 94) 60 (55, 75) F1,10=0.03; P=0.874 SCM 67 (65, 70) 66 (65, 69) 61 (52, 68) F1,8.3=2.80; P=0.132 70 (67, 76) 74 (60, 119) 64 (60, 73) F1,10=0.65; P=0.440 SPL 60 (55, 79) 64 (59, 79) 60 (57, 73) F1,9.3=0.05; P=0.836 64 (57, 67) 64 (52, 68) 55 (42, 61) F1,9.3=3.55; P=0.091 SSCAP 60 (58, 90) 66 (58, 105) 68 (54, 81) F1,7=0.66; P=0.445 63 (54, 68) 67 (55, 74) 56 (48, 69) F1,7.7=0.97; P=0.354 SSCERV 92 (69, 150) 74 (57, 140) 62 (54, 85) F1,9=1.30; P=0.283 69 (63, 78) 65 (58, 71) 61 (55, 75) F1,8.9=0.09; P=0.766 MU 92 (74, 106) 68 (60, 151) 52 (40, 66) F1,8.7=4.82; P=0.057 65 (57, 73) 59 (57, 68) 59 (53, 67) F1,10=0.00; P=0.961 LS 68 (57, 73) 62 (55, 94) 54 (34, 72) F1,10=1.40; P=0.265 63 (58, 77) 64 (57, 70) 58 (53, 66) F1,10=1.13; P=0.312 TRAP 63 (53, 91) 62 (53, 71) 58 (31, 60) F1,9.5=1.55; P=0.243 63 (56, 71) 68 (54, 76) 74 (50, 78) F1,8.7=0.00; P=0.978 Kinematic Onset (ms) Head CG ax 47 (37, 59) 40 (34, 55) 36 (32, 38) F1,9=1.93; P=0.198 67 (60, 75) 77 (51, 101) 34 (32, 35) F1,9=9.09; P=0.015* Torso T1 ax 30 (27, 34) 31 (28, 32) 20 (16, 24) F1,9=29.29; P=0.000* 35 (26, 39) 35 (29, 41) 18 (14, 22) F1,9=45.61; P=0.000* Head ωy 51 (43, 60) 45 (34, 48) 46 (43, 49) F1,9=1.15; P=0.311 56 (48, 64) 52 (30, 55) 43 (37, 51) F1,9=0.70; P=0.426 Torso ωy 31 (26, 37) 32 (20, 35) 21 (14, 30) F1,9=4.49; P=0.063 30 (28, 34) 28 (24, 35) 25 (19, 35) F1,9=0.46; P=0.516 97 Table 6.5 - The median (1st, 3rd quartile) peak kinematic variables in both rear and frontal impact. a, linear acceleration; ω, angular velocity; θ, angular position; x, forward direction; y, rightward direction; CG, center of gravity; T1, first thoracic vertebrae. *Denotes a significant (P<0.05) difference between the relaxed and braced trials in the repeated measures linear mixed model analysis. Kinematic Variables Rear Impacts Frontal Impacts Control Relaxed Braced Relaxed vs. Braced Statistics Control Relaxed Braced Relaxed vs. Braced Statistics Head CG ax (g) 2.5 (2.4, 2.9) 2.7 (2.6, 2.9) 2.3 (2.2, 2.5) F1,9=15.77; P=0.003* -1.0 (-1.1, -1.0) -1.5 (-1.8, -1.2) -1.9 (-2.3, -1.8) F1,9=8.03; P=0.020* Torso T1 ax (g) 2.2 (1.7, 2.4) 2.3 (2.0, 2.9) 3.4 (3.2, 3.8) F1,9=11.92; P=0.007* -1.3 (-1.6, -1.1) -2.8 (-3.9, -1.9) -2.9 (-3.2, -2.8) F1,9=0.02; P=0.901 Head ωy (°/s) 332.3 (308.2, 374.7) 360.8 (332.4, 370.1) 385.5 (337.9, 409.2) F1,9=0.79; P=0.398 -139.2 (-167.3, -86.4) -143.1 (-162.6, -78.6) -267.2 (-298.7, -228.8) F1,9=14.62; P=0.004* Torso ωy (°/s) 115.1 (99.9, 167.9) 133.1 (111.3, 268.0) 139.5 (108.9, 154.2) F1,9=0.17; P=0.693 -81.1 (-96.8, -64.9) -81.5 (-128.7, -52.9) -182.0 (-199.6, -168.0) F1,9=14.91; P=0.004* Head θy (°) 18.9 (18.3, 21.9) 20.1 (19.7, 20.9) 21.0 (18.4, 22.6) F1,9=0.07; P=0.793 -13.5 (-14.7, -12.3) -12.5 (-15.3, -11.1) -17.9 (-18.9, -16.3) F1,9=10.04; P=0.011* Torso θy (°) 12.9 (9.9, 17.1) 15.4 (12.2, 28.4) 7.2 (2.4, 8.7) F1,9=23.61; P=0.001* -9.0 (-10.7, -8.4) -5.8 (-7.3, -4.0) -7.3 (-8.8, -6.5) F1,9=8.18; P=0.019* Retraction (mm) -26.0 (-27.3, -23.2) -26.2 (-26.9, -23.0) -20.1 (-32.8, -14.5) F1,9=2.12; P=0.179 41.2 (32.1, 55.2) 43.5 (32.5, 58.4) 33.4 (30.7, 45.6) F1,9=6.54; P=0.031* 98 Table 6.6 - The median (1st, 3rd quartile) timing of peak kinematic variables in both rear and frontal impact. Data presented in ms units. a, linear acceleration; ω, angular velocity; θ, angular position; x, forward direction; y, rightward direction; CG, center of gravity; T1, first thoracic vertebrae. *Denotes a significant (P<0.05) difference between the relaxed and braced trials in the repeated measures linear mixed model analysis. Kinematic Variables Rear Impacts Frontal Impacts Control Relaxed Braced Relaxed vs. Braced Statistics Control Relaxed Braced Relaxed vs. Braced Statistics Head CG ax 123 (121, 127) 123 (119, 131) 89 (82, 104) F1,9=71.47; P=0.000* 185 (163, 192) 160 (145, 179) 94 (89, 96) F1,9=34.42; P=0.000* Torso T1 ax 85 (84, 88) 84 (81, 85) 67 (66, 74) F1,9=314.18; P=0.000* 143 (125, 147) 131 (127, 146) 73 (70, 84) F1,9=26.63; P=0.001* Head ωy 143 (142, 146) 144 (140, 145) 106 (103, 118) F1,9=93.83; P=0.000* 128 (120, 140) 155 (114, 178) 107 (101, 115) F1,9=24.47; P=0.001* Torso ωy 106 (98, 149) 156 (84, 164) 147 (73, 168) F1,9=0.19; P=0.671 122 (104, 164) 149 (83, 226) 188 (167, 196) F1,9=0.01; P=0.909 Head θy 175 (172, 182) 177 (167, 182) 131 (127, 148) F1,9=72.50; P=0.000* 268 (244, 306) 239 (217, 251) 131 (125, 150) F1,9=29.58; P=0.000* Torso θy 179 (157, 194) 194 (176, 207) 130 (99, 197) F1,9=5.03; P=0.052 216 (189, 231) 143 (112, 174) 160 (95, 231) F1,9=0.04; P=0.845 Retraction 128 (124, 184) 127 (124, 128) 137 (98, 182) F1,9=0.46; P=0.516 244 (234, 251) 215 (204, 228) 189 (179, 202) F1,9=11.77; P=0.008* 99 Figure 6.2 - Exemplar traces for EMG and kinematic data from a single subject in rear impact for hands in lap, relaxed and braced experimental conditions. Vertical scale bars are aligned with the onset of perturbation and represent 2 g, 30 mm, 400 °/s2, and 25°. STH, sternohyoid; SCM, sternocleidomastoid; SPL, splenius capitis; SSCAP, semispinalis capitis; SSCERV, semispinalis cervicis; MU, multifidus; LS, levator scapulae; TRAP, Trapezius; a, linear acceleration; ω, angular velocity; θ, angular position; x, forward direction; y, rightward direction; CG, center of gravity; T1, first thoracic vertebrae. 100 Figure 6.3 - Exemplar traces for EMG and kinematic data from a single subject in frontal impact for hands in lap, relaxed and braced experimental conditions. Vertical scale bars are aligned with the onset of perturbation and represent 2 g, 30 mm, 400 °/s2, and 25°. STH, sternohyoid; SCM, sternocleidomastoid; SPL, splenius capitis; SSCAP, semispinalis capitis; SSCERV, semispinalis cervicis; MU, multifidus; LS, levator scapulae; TRAP, Trapezius; a, linear acceleration; ω, angular velocity; θ, angular position; x, forward direction; y, rightward direction; CG, center of gravity; T1, first thoracic vertebrae. 101 Figure 6.4 - Normalized RMS EMG pre-impact and peak during impact for both rear and frontal impacts of the eight muscles studied. STH, sternohyoid; SCM, sternocleidomastoid; SPL, splenius capitis; SSCAP, semispinalis capitis; SSCERV, semispinalis cervicis; MU, multifidus; LS, levator scapulae; TRAP, Trapezius. *Denotes a significant (P<0.05) difference between the relaxed and braced trials in the repeated measures linear mixed model analysis. 102 Figure 6.5 - Peak and the timing of peaks for kinematic measures in both rear and frontal impact. a, linear acceleration; α, angular acceleration; ω, angular velocity; θ, angular position; x, forward direction; y, rightward direction; CG, center of gravity; T1, first thoracic vertebrae. *Denotes a significant (P<0.05) difference between the relaxed and braced trials in the repeated measures linear mixed model analysis.103 6.5 Discussion The goal of this study was to examine how a vehicle occupant placing their hands on the steering wheel while relaxed compared to braced affected their head and torso kinematic and neck muscle response to rear and frontal impact. We found widespread increases in the muscle activity before impact in the braced compared to relaxed trials for both rear and frontal impacts, which confirms our first hypothesis (Table 6.7). In rear impact, the peak muscle activity was not different for the braced compared to relaxed case, and peak activity decreased in the MU and TRAP in frontal impact, so we reject our second hypothesis (Table 6.7). Finally, we hypothesized that when comparing braced to relaxed trials head and torso peak acceleration and angular velocity would increase while angular displacements and retraction would decrease in both rear and frontal impacts. For torso kinematics, we partially accept this hypothesis in rear impact and reject it in frontal impact (Table 6.7). For head kinematics, we reject the final hypothesis in rear impact and partially accept it in frontal impact (Table 6.7). In automobile collisions, increased displacements of joints may relate to increased tissue strains, while increased accelerations are related to increased forces. In frontal impact, our results show bracing increased acceleration of the head, which may lead to higher forces on the neck, and increased angular displacement, which may lead to higher strains in the neck. The combination of potentially higher forces and strains could explain the increased WAD injury risk for braced drivers (Jakobsson et al., 2004). In rear impact, bracing may be protective against whiplash injury as it led to decreased head acceleration, but epidemiologic evidence is lacking to support this. Further, pre-impact muscle activity may be contributing to the altered kinematics during braced compared to relaxed impact. Finally, the data can be used to validate or improve computer simulations of humans in these impact scenarios (See Appendix Figure A.4 and Figure A.5 for kinematic response corridors). 104 Table 6.7 - Summary of hypothesis testing for braced versus relaxed trials in both rear and frontal impact. a, linear acceleration; ω, angular velocity; θ, angular position; x, forward direction; y, rightward direction; CG, center of gravity; T1, first thoracic vertebrae. Metric Hypothesized Rear Frontal Hypothesis Accept/Reject Pre-impact muscle activity Increase Increased in 7 of 8 muscles Increased in 6 of 8 muscles Hypothesis 1: Accepted Peak muscle activity Increase No sig. diff. Decreased in MU & TRAP Hypothesis 2: Rejected Torso Kinematics T1 ax Increase Increased No sig. diff. Hypothesis 3: For torso, partially accepted in rear impact and rejected in frontal impact ωy Increase No sig. diff. Increased θy Decrease Decreased Increased Head Kinematics CG ax Increase Decreased Increased Hypothesis 3: For head, rejected in rear impact and partially accepted in frontal impact ωy Increase No sig. diff. Increased θy Decrease No sig. diff. Increased Retraction Decrease No sig. diff. Decreased Bracing against the steering wheel compared to relaxed hands on the steering wheel led to increased pre-impact neck muscle activity in all muscles except STH & TRAP before rear impact, and SSCERV, MU, & TRAP before frontal impact. These increases in median pre-impact activity were limited to a range of 0.4% MVIC (SSCAP) to 2.6% MVIC (SPL) in rear impact, and 0.6% MVIC (SSCAP) to 4.8% MVIC (LS) in frontal impact. It is not clear if these levels of muscle activation alone were sufficient to stiffen joints and alter kinematics, but it is possible that the pre-impact activity altered joint angles and the load path through the spine, contributing to the altered kinematics. The increase in pre-impact muscle activity in braced vs. relaxed trials was generally larger in frontal impact, and it did not always occur in the same muscle for each impact direction. Given that subjects knew the impact direction before perturbation, it is possible that participants utilized a direction-specific preparatory response. The potential preparatory response was further explored using a post-hoc analysis consisting of separate paired t-tests on the rank transformed pre-impact muscle activity for each muscle comparing the data before the braced frontal and rear impacts. No significant differences were found (t10=-1.097:1.218, P=0.251:0.952), which does not provide any evidence for a direction-specific preparatory response. Separate from neck muscle activation, the lack of changes found in the arm, elbow, seat back, and Frankfort plane angles before frontal and rear impacts does not support directionally specific preparation. The MU muscle is potentially important in the etiology of whiplash injury, because of its direct insertion on the capsular ligament (J. S. Anderson et al., 2005; Winkelstein et al., 2001), which have been linked to the source of whiplash injury in 54-60% of cases (Barnsley et al., 1995; Lord et al., 1996). It has been proposed that activation of the multifidus muscle during rear impact may exacerbate capsular ligament strains (Mang et al., 2015; Siegmund, Blouin, et al., 2008), and evidence from animal models (Dong & 105 Winkelstein, 2010; Lee et al., 2004; Lu et al., 2005), cadaver models (B. Deng et al., 2000; Ivancic et al., 2008; Pearson et al., 2004), and computational studies (Cronin, 2014; Fice et al., 2011; Stemper et al., 2005) suggests that increased capsular strain increases the risk of whiplash injury. In our study, we found a small (1.0% MVIC) increases in median MU activity pre-impact in braced vs. relaxed trials for rear impact, and no difference for frontal impact. It isn’t known if this activity level could affect capsular strain in a meaningful way, but we also found that peak activity decreased in braced vs relaxed trials for frontal impact (-20.8% MVIC). Taken together, our results suggest that bracing before impact could provide a protective measure against muscle-induced capsular strain in frontal impact, but the effect in rear impact is unclear. The potentially protective nature of bracing on the capsular ligament is at odds with the increase in WAD symptoms reported for braced frontal impact (Jackobsson et al., 2004), so it is possible that a different injury mechanism is responsible. There was one subject that exhibited an increase of 5.6% MVIC in peak MU activity during the frontal braced trial, and in rear impact 7 of 10 subjects exhibited increased (mean (SD) of 11.2 (15.2) %MVIC) peak MU activity, which opens the possibility that the mean trend might not always explain the injury of an individual. Further, we found that the average peak MU response for the three females tested was 33.9 and 167.1 % MVIC higher than the males in the relaxed condition for rear and frontal impact respectively and 19.7 and 92.4% higher in the braced condition again for rear and frontal impact. We don’t have enough female subjects to statistically test these differences, but our results suggest a possible link to the increase in WAD risk for female occupants (Carstensen et al., 2012; Jakobsson et al., 2004; Krafft et al., 2003). In the frontal impact when comparing braced vs. relaxed trials, we found increased peak head and torso angular rate, earlier onset of head and torso acceleration, and earlier timing of peak head and torso acceleration, head angular velocity, and head angle. In rear impact when comparing braced vs. relaxed trials, peak torso acceleration increased, torso angular displacement decreased, and we found earlier peaks of head and torso acceleration, head angular velocity, and head angle. These kinematic changes in the braced compared to relaxed trials are generally consistent with a stiffer torso and neck, or tighter mechanical coupling between the subject and sled, for both the frontal and rear impacts. With a stiffer response, we expect perturbation energy to be absorbed over a shorter period of time which results in smaller displacements, higher peak acceleration/velocity, and earlier peak kinematic measures. Exceptions for the expected response of a stiffer system were the decrease in head acceleration during rear impact, and the increase in head and torso angular displacement in frontal impact. The increase in head angle during braced compared to relaxed trials may be related to the increase in extension angle of the head before perturbation in braced trials (Δ5.6°). If the pre-impact neck muscle activity led to 106 extension of the head, then there would be a larger range of motion available for flexion during frontal impact, and less range of motion during rear impact. Previous studies in frontal impact, also showed deviations from a stiffer response with head and shoulder excursions reduced, but no differences found in head or torso accelerations and reaction forces for volunteers braced against a steering wheel compared to relaxed (Beeman et al., 2016, 2011; Kemper et al., 2014). Computational human body models are a developing tool used by automotive companies to improve the safety of vehicles. Muscle activation in these models has been shown to be important to capture the kinematics of humans during an impact (Brolin et al., 2005; Fice et al., 2011) and before impact (Östh et al., 2015). Most recently, closed-loop PID feedback control has been used to control muscle activity in human body models during long duration braking events that lead to a collision (Östh et al., 2015, 2012). Even these most advanced muscle activation controllers require tuning of feedback gains to match specific types of load cases. Muscle activation controllers also require validation data to ensure that muscle activation matches humans to ensure that the controller is not masking deficits in the mechanics of the underlying model. With access to these data, researchers will be able to use the muscle activation as a function of time along with the kinematic corridors (Appendix Figure A.4 and Figure A.5) to tune and/or validate their models in rear or frontal impact for braced or relaxed drivers and relaxed passengers. The researchers could then increase the impact severity to study injury mechanics of braced occupants more closely and inform injury prevention approaches. A limitation of this work is that only one automotive seat, one impact axis, and one impact severity were tested, and it is not certain that our results apply more generally. Further, the perturbation was limited to 0.77 m/s speed change and a peak acceleration of 2.0g, which was chosen to limit the risk of injury to our subjects. It is not known how these results would scale to more severe and injurious impact scenarios. Another limitation is the assumption of a rigid body between the torso accelerometer mount and the C7-T1 joint axis, which means that deformation of the torso between the sternum and C7-T1 will alter our torso kinematic measurements. Also, subjects were aware that a perturbation would occur within a 5-10s window, which we did to ensure a consistent timing between trial types and it would have been fatiguing to have subjects maintain a 60% MVIC force on the steering wheel for longer than this during braced trials. Awareness is not a limitation for the braced condition, as these trials represent an aware driver, but the kinematics and muscle responses of the control and relaxed may not represent an unprepared passenger or driver as desired (Siegmund et al., 2003b). Finally, we do not have enough 107 subjects to separate our results for males and females, and females have been shown to have a higher risk of whiplash injury (Carstensen et al., 2012; Jakobsson et al., 2004; Krafft et al., 2003). In summary, we described the head and torso kinematics and neck muscle response of volunteers when they put their hands on the steering wheel while relaxed and braced by pushing against the wheel in both rear and frontal impact. When compared to hands relaxed on the wheel, we found that bracing increases the pre-perturbation activity of most muscles for both frontal and rear impacts and reduces the peak neck muscle activity for some muscles during frontal impacts. Bracing led to increased angular velocity and displacements of the head and torso, increased head acceleration, and decreased retraction in frontal impact. In rear impact, bracing reduced head acceleration, increased torso acceleration and decreased torso angle. These results provide knowledge of how occupant behaviour before an impact can change their response to an impact. Further, this work provides data that can be used to inform injury prevention approaches or improve computational modeling of drivers in both braced and relaxed states before impact. 6.6 Data Availability As we know the data collected in this study would be useful to modelling researchers we ensured that the we have ethics approval and consent from our subjects to share the anonymized data. The data will be available to download by contacting Dr. J.-S. Blouin (jsblouin@mail.ubc.ca). 108 Chapter 7. General Discussion and Conclusions The overall goal of this dissertation was to understand how occupant behaviours before a collision, including non-neutral head postures and bracing against the steering wheel, influenced neck muscle activity and head/neck kinematics during impact. An extension of this goal was to provide data and knowledge that can be incorporated into computational models of the head and neck to prevent or mitigate neck injuries in automobile collisions. To accomplish this, we conducted a series of five experiments. In the first (Chapter 2), we discovered how neck maximum voluntary isometric contractions scale in directions beyond the principal axes. This is important to reproduce in computational models before seeking to simulate non-principal-axis loadings such as non-neutral postures. In the next experiment (Chapter 3), we showed that the biomechanical line of action of a neck muscle is not an accurate predictor of how the body uses that muscle, which will guide the development of neck muscle controllers in computational models. In Chapter 4, we quantified the non-neutral head postures drivers adopt under naturalistic conditions, providing experimentally derived boundary conditions for experimental and computational research. Next, these data were used in Chapter 5, where we explored the role of neck muscle activation for drivers subjected to rear impact while in non-neutral postures. In the last experiment (Chapter 6), we explored the role of neck muscle activation for drivers and passengers when bracing before rear or frontal impacts. The motivation for the experiment in Chapter 2 was that validation of strength in computational models of the head-neck models is limited to principal axis directions. In this experiment, we found that off-axis strength when normalized by principal axis strength was generally equal to unity, except for combinations of ipsilateral axial rotation and lateral bending where the normalized strength exceeded one. The pattern of off-axis strength relative to principal axis strength was consistent amongst subjects, allowing computational head and neck models to be validated in many directions using only principal axis strength data. Off-axis neck strength is important for omni-directional impacts and occupants in non-neutral postures. Ensuring the biomechanics (i.e. strength) in multiple directions in a model are representative of humans will give confidence in the basics of the muscle model before trying to apply complex muscle activation schemes or muscle controllers to model dynamic scenarios including pre or post crash occupant responses. This step-wise approach to validation of models is important to build confidence and prevent the application of muscle controllers to mask deficiencies in the base muscle or structural components. Therefore, for model validation it is important to use the data from Chapter 2 first before proceeding to using perturbation data from Chapter 5 or 6 which looked at braced and non-109 neutral postures during low speed impact. Further, this experiment was also motivated to provide experimental researchers a way to scale contractions in non-principal axis direction as a function of MVIC. This non-principal axis scaling of MVIC was important for the experimental design of Chapter 3. The motivation behind Chapter 3 was to provide fundamental knowledge of the relationship between the neural control and biomechanics of neck muscles to shape future efforts in modelling human motor control. Currently, muscle control in human body models have advanced to closed-loop PID control based on either muscle length or joint angles (Meijer et al., 2013; Östh et al., 2015), but it is known that simple feedback control would be unstable in humans due to neural delays and noise (Franklin & Wolpert, 2011). As modelling of muscle control advances towards a more physiologically-based solution, certain simplifying assumption(s) may be tempting. One such assumption is that the line of action of a muscle will predict how the motor controller would utilize that muscle. In Chapter 3, we found this assumption to be invalid because the preferred activation directions, a measure of the motor control, were 23, 39, & 21° different from the electrically-stimulated directions which were a measure of muscle line of action in the sternocleidomastoid, splenius capitis, and semispinalis capitis respectively. Further, we observed less variability in the preferred activation direction within subjects when the neck contraction moment increased. Taken together, we showed that the human neck motor controller does not optimize individual muscle biomechanics but, as activation increases, biomechanical constraints in part dictate the activation of neck muscles. In Chapter 2, a base level of validation was presented that should be applied to neck models in omni directions, and the experiment in Chapter 3 provided fundamental knowledge to aid in the design of a muscle controller, the rest of the experiments in this dissertation were focussed on providing knowledge of the role of neck muscles and model validation data in injury relevant crash scenarios. The first such scenario is when occupants do not have their head in a neutral posture, i.e. looking straight ahead before impact. When occupants were in non-neutral head postures before an impact, their injury risk increased to 36% from 23% for neutral posture (Jakobsson et al., 2008). Before we investigated neck muscle and kinematic responses of occupants in non-neutral postures, we first needed to quantify the head postures of actual drivers, which was the objective of Chapter 5. We found that while driving a vehicle on public roads drivers (left hand drive) had peak yaw angles of the head relative to the vehicle of -81.5° for left shoulder check, -34.3° for left mirror check, 16.2° for rear-view mirror check, 42.1° for right mirror check, 58.2° when looking at their passenger, and 84.3° for right shoulder check. The head postures quantified for common driving tasks were then used as boundary conditions for the 110 experiment in Chapter 5 and can be used likewise for other modelling or experimental research. We also found that drivers spent a larger proportion of time in non-neutral postures when the vehicle was stopped (17.5%) compared to moving (8.2%) and they moved their head further from neutral when the car was stationary compared to moving. These findings provide a possible explanation for why drivers are more likely to be stationary compared to moving at the time they were injured from a rear impact (Deans et al., 1987; Gibson et al., 2000; Norris & Watt, 1983; Ryan et al., 1994; Sturzenegger et al., 1995). It is known that whiplash injuries occur more frequently when occupants have their head turned at the time of impact, but most volunteer impact studies have studied volunteers in neutral postures. The goal of the experiment in Chapter 5 was to understand the response of neck muscles and head/neck kinematics when volunteers maintained one of five postures: neutral, left shoulder check, left mirror check, rear-view mirror check, and look at passenger, before a perturbation that mimicked a low speed rear impact. The head angles used to represent these common driving tasks were derived from the results in Chapter 4 of this dissertation. When compared to a neutral head posture, we found that several non-neutral postures increased the pre-impact activity of neck muscles and the left mirror check posture reduced multifidus activity. Non-neutral postures led to widespread increases in accelerations and angular velocities of the head and angular velocities of the torso in directions outside the sagittal plane. Taken together, these results suggest that non-neutral postures may have changed the spinal load paths and the resulting kinematics during impact, potentially through facet joint orientations and pre-strain of tissues, and these effects may be more important for injury risk than changes in peak muscle activity. Possibly, increased initial muscle activity has contributed to the altered load paths in the spine. In addition to the understanding of the role neck muscle activation plays in non-neutral head posture impact, the data presented in this work are valuable validation data for models of this type of loading. As with Chapter 5, the motivation for the last experiment was again that people are not always relaxed and looking straight ahead before an automotive rear impact. A common way for occupants to react to an impending collision is to brace themselves by pushing against the steering wheel and straightening their arms (Choi et al., 2005; Hault-Dubrulle et al., 2011). Previous work has shown that awareness to an impact (Siegmund et al., 2003a) and co-contracting your neck muscles before impact (Ejima et al., 2012, 2008; Ono et al., 1997) alter head/neck kinematics, but the effect of bracing against the steering wheel on neck muscle activation was unknown. In Chapter 6, we investigated the changes of neck muscle 111 activity and head/neck kinematics when subjects maintained their hands on the steering while either relaxed or braced against the steering wheel before low-speed frontal and rear impact perturbations. We found that when compared to the relaxed condition, bracing increases the pre-perturbation activity of most muscles for both frontal and rear impacts and reduced the peak neck muscle activity in trapezius and multifidus during frontal impacts. Bracing led to widespread increased accelerations and angular rates, and peaks that occurred earlier in the head and torso in frontal impact, and to a lesser extent in rear impact. These results show that bracing leads kinematic changes expected from a stiffer neck, but the effects were less pronounced in rear impact. In frontal impact, the stiffening of the neck runs contrary to the reduced peak muscle activity in two muscles expected to resist neck flexion, which could be due to pre-impact muscle activity that may have caused intervertebral extension, which is partially supported by an increase in head extension before braced trials, reducing the gap in adjacent facet surfaces stiffening the neck in protraction. As with the results in Chapter 5, these results can be used to inform the design of injury preventions methods and for validation data for models in this type of loading. The results presented in this dissertation improve our understanding of neck muscle biomechanics with how strength scales in non-principal-axis directions, fundamental aspects of the relationship between neck muscle biomechanics and neural control, and how neck muscles respond to impact when the head is turned, or the occupant braces themselves against the steering wheel. For modelling research, this body of work can be used to validate omni-directional biomechanics, inform muscle control design, and validate models exposed to impacts for occupants in non-neutral head postures or braced postures. 7.1 Implications for modelling research Part of the motivation for this dissertation was to provide data that could be used to improve computational models of the human body. With this goal in mind, the subject data from Chapter 2 is available in the Appendix of this dissertation and the subject data from Chapter 5 and 6 have been made available to download by contacting J.-S. Blouin (jsblouin@mail.ubc.ca). The data made available will include, for each subject and trial, the processed EMG time histories of each muscle and the processed kinematic time histories of the variables presented. The sex and anthropometry of the subject represented in each data trace will be provided. An important aspect of validating human body models is step-wise validation at increasing scales of complexity. For example with a head and neck model, one can begin with tissue level validation 112 (Mattucci et al., 2013), followed by functional cervical spinal segments in quasi-static and dynamic loadings (Barker et al., 2017; Panzer & Cronin, 2009), full cervical spine in quasi-static loading (Panzer et al., 2011), full neck maximum isometric contractions (de Bruijn et al., 2016; Mortensen et al., 2018; Vasavada et al., 1998), and finally simulating vehicle collisions (Fice & Cronin, 2012; Panzer et al., 2011). This step-wise approach helps to provide confidence that errors at one level of complexity are not masking underlying deficiencies. For example, head/neck kinematics in a collision simulation could be matched to volunteers using non-physiologic muscle forces to mask an inappropriately stiff cervical spine, and the step-wise validation approach would limit the possibility of this kind of compounding error. At the step of validating the maximum isometric contraction of the full neck model, currently available data limit validation to principle axis directions (Cagnie et al., 2007; Jordan et al., 1999; Mayoux-Benhamou et al., 1989; Portero et al., 2001; Queisser et al., 1994; Vasavada et al., 2001). The data in Chapter 2 (Fice et al., 2014) allow validation of computational neck models in several non-principle axis directions. This type of validation will provide confidence in the underlying omni-directional biomechanics (i.e. strength) of a model before advancing to simulations of omni-directional collisions or non-neutral postures that result in non-principal-axis head motions. When simulating omni-directional collisions, it is important that the muscle activity patterns for computational human neck models are based on volunteer experimental data. It may be tempting to assume that human neck muscles will be recruited maximally when loads on or displacements of the head align with their biomechanical line of action, but in Chapter 3 it was shown that this assumption is invalid. More specifically, in Chapter 3 it was shown that the biomechanics, associated with isolated activation of the muscle, cannot be used to predict when or how the human neural controller will utilize that muscle for a given task or situation. For computational models, it is possible to use an optimization approach with a cost function such as minimize energy expenditure to establish omni-directional muscle activation schemes that utilize the biomechanics of the neck to resist the applied loads/accelerations. Most optimization approaches will under estimate the co-contraction of antagonist muscles used to stiffen the neck, due to the nature of the energy use this would entail. Although in humans co-contraction may be important for joint stability and it has been postulated as a method of human motor control, known as impedance control (Hogan, 1984). In the shoulder for example, a computation model with an EMG driven muscle activation scheme compared to an optimized scheme resulted in up to 45% improvement in the predicted joint reaction forces of patients with instrumented hemi-arthroplasty (Nikooyan et al., 2012). The difference in the two muscle activation schemes was shown to result from more co-contraction in the EMG driven muscle activity. In the neck, co-contraction may act to stiffen the 113 spinal column because without neuromuscular support the cervical spine is unstable and prone to buckling (Crisco et al., 1992; Panjabi et al., 1998). Therefore, using data from volunteer experiments will allow the prediction of accurate omni-directional muscle activation patterns in the neck for computational models, which may provide more accurate predictions of the underlying stress and strain on tissues. Such data for omni-directional seated perturbation exist (Ólafsdóttir et al., 2015, 2018), but these data do not provide information about muscle activation during head axial rotation and the data in Chapter 3 can augment the data from Ólafsdóttir et al. (2015, 2018) for a subset of neck muscles. Muscle control of head axial rotation may be important for maneuvers or collisions that result in yaw motion of the vehicle, such as lane changes or spins. To improve the usability of the whiplash-perturbation studies in this dissertation (Chapter 5 and 6) by computational modellers, certain factors were considered. Firstly, muscle activity was normalized by the MVIC for each muscle. In this format, the data are comparable to the Hill muscle model activation state which scales peak muscle force by a factor of zero to one, where one represents maximal excitation (Equation 1.1; Figure 1.4). In presenting muscle activity data this way, the muscle activity data collected in Chapters 5 and 6 can be used in the activation state of a human body model to partially drive its kinematic response, or they can be directly compared to the activation state output of a model’s muscle controller. A potential limitation of this approach is that the MVIC measured may not correspond with a volunteer’s true maximum voluntary effort, and the true maximum voluntary effort may not correspond with the peak capabilities of the muscle. In fact it is common to measure neck muscle activity during sled perturbation that exceeds MVIC (Table 5.4; Table 6.3; Ólafsdóttir et al., 2015; Siegmund, Blouin, et al., 2008). Another consideration to ensure usability of the data for human body modellers was the measurement of the sled acceleration as a function of time, which allows accurate reproduction of the loading conditions used in Chapters 5 and 6. Finally, so that the boundary conditions of the sled perturbations could be reproduced, we took measurements and reported the position of the head, arms, legs, and seatback. These measurements allow for accurate positioning of human body models to match the volunteers and reduce the potential sources of error between the model and the volunteer data. Further, no head restraints or seatbelts were utilized because these components generate complex boundary conditions for the model, and their omission again reduces potential sources of error in the model implementation. A potential limitation when modeling the boundary conditions of the sled experiments in Chapter 5 and 6 is that we do not have a computational model of the automotive seat available to share, but the driver’s seat from a 2005 Volvo S40 was used here, and modeling researchers could create the appropriate seat model from an exemplar seat. 114 Developments in computational human body models have led researchers away from using feed-forward predefined muscle activation states in their muscle models and towards muscle controllers that modify the activation state in a manner loosely based on human motor control (Iwamoto & Nakahira, 2015; Meijer et al., 2013; Östh et al., 2015, 2012). One approach is to use PID controllers to activate muscles based on joint angle or muscle length relative to a desired set-point, usually the neutral or initial posture before impact (Meijer et al., 2013; Östh et al., 2015, 2012). The researchers of this PID approach have recognized that the gains of their controller cannot be constant values for different loading situations, which is because the human body flexibly alters reflex gains contextually. For example, the cervicocollic reflex (CCR) is suppressed when a volunteer’s body was rotated under them with sinusoidal motions while their head was free (Keshner et al., 1999; Keshner & Peterson, 1995; Peng et al., 1996). The suppression of CCR is likely because the volunteer’s objective was to keep their head fixed in space, and the CCR would have worked against this objective. Another example of changing feedback gains is the suppression of neck afferent feedback when the frequency of torso perturbation was sufficiently high to exceed the natural frequency of the head-neck system (Forbes et al., 2013). This suppression potentially occurred because, at such frequencies, delay in the afferent feedback may led to oscillatory behaviour. Further, it has been shown that the nervous system only reacts to task relevant perturbations, which requires flexible afferent feedback gains (Franklin & Wolpert, 2008; Todorov & Jordan, 2002; Valero-Cuevas et al., 2009). In fact, the ability to flexibly control reflex gains is foundational to the optimal feedback control theory for human motor control, which has been able to demonstrate many of the features of human motor control, including only responding to task-relevant perturbations without explicitly coding for these features (review: S. H. Scott, 2012; Todorov & Jordan, 2002). The data presented in Chapters 5 and 6 will allow the gains of muscle controllers in human body models to be tuned to situations wherein occupants are in non-neutral positions or braced against the steering wheel before impact. As discussed further below, these situations are important to model due the increased injury risk for non-neutral (Jakobsson et al., 2008) and braced postures (Jakobsson et al., 2004). Modelling drivers in non-neutral and braced postures may generate insight into the stresses and strains on tissues in the neck and help identify the sources of injury, but note that physiological muscle activity and forces are required to accurately predict these stresses and strains on neck tissues. 7.2 Aetiology of whiplash In Chapter 4, we showed that drivers adopted axially rotated head postures further from neutral when their vehicle was stopped compared to when it was moving. This finding potentially explains why 115 symptomatic patients were significantly more likely than asymptomatic patients to be in a stationary vehicle with their head rotated at the time of impact (Sturzenegger et al., 1995). We also found that drivers spend proportionally more time in non-neutral postures while their vehicle was stopped compared to moving. This finding potentially explains why a large proportion (54-90%) of patients with neck pain after a rear impact report that their vehicle was stopped at the time of impact (Deans et al., 1987; Gibson et al., 2000; Norris & Watt, 1983; Ryan et al., 1994; Sturzenegger et al., 1995), but it is potentially inconsistent with studies that did not find a statistical link between injury and stationary vehicles (Norris & Watt,1983; Sturzenegger et al.,1995; Radanov et al.,1995). During automobile collisions, increased joint displacements may relate to increased tissue strains, while increased accelerations are related to increased forces. Our findings of braced compared to relaxed occupants in frontal collisions showed that both linear acceleration and angular rotation increased with bracing, which support epidemiologic evidence that braced occupants experience higher injury rates in frontal collisions (Jakobsson et al., 2004). In rear impact, braced subjects had decreased head acceleration when compared to relaxed subjects. This change in kinematics could be protective, but there is not yet epidemiologic evidence to suggest that bracing in rear impact reduces injury risk. When occupants were performing a left shoulder check or looking at their passenger, we found that sagittal plane retraction and head angular displacement were reduced, which taken together in isolation could suggest that these postures may be protective (Kumar et al., 2005). This perspective, however, ignores the fact these displacements are acting on joints with connective tissue that is already pre-strained by the rotated head posture (Siegmund, Davis, et al., 2008). It also ignores the large increases in head rotation and accelerations in directions outside the sagittal plane. Our findings of increased acceleration and rotations beyond the sagittal plane may increase forces and tissue strains in the cervical spine and therefore these findings support epidemiologic evidence that head turned postures increase injury risk and the persistence of symptoms during rear impact (Jakobsson et al., 2008; Radanov et al., 1995). In both braced perturbations and non-neutral head postures, pre-impact activity generally increased for most neck muscles. This increased activity may alter the load paths in the neck, and thereby contribute to the altered kinematics and changes in injury risks compared to relaxed neutral postures. Pre-impact activity may change the load paths in the neck through pre-stress on ligaments, intervertebral discs, and muscle tissue in addition to altering the relative alignment of adjacent facet surfaces. We did not find widespread changes to the peak muscle activity during impact for braced or non-neutral head postures when compared to relaxed neutral postures. Therefore, it is unclear what role, if any, changes in evoked peak muscle activity have on the injury risk during these collision scenarios. 116 7.3 General limitations The number of muscles measured in the experiments in this dissertation is a potential limitation because the results can only inform modelers and others as to the responses of specific muscles, requiring additional assumptions or simplifications to apply the current results to the whole neck. When using surface electrodes, there are only a limited number of superficial muscles that can be measured before the muscle signals become difficult to interpret because of cross-talk and geometric displacement. For deeper muscles, indwelling electrodes are required and were used in this dissertation. The experiment in Chapter 4 measuring from three unilateral muscles with two insertions per muscle, the experiment in Chapter 5 measuring from four bilateral muscle pairs, and the experiment in Chapter 6 measuring from eight unilateral muscles. The experiments in Chapters 2 and 4 did not record muscle activity. The number of muscles recorded with indwelling electrodes can be limited due to the size of the neck and the proximity of wires to one another as the number of electrodes increases, because wires hitting each other outside of the neck during dynamic movements can lead to movement artifacts which can saturate amplifiers. In the future, it may be possible to alleviate some of this concern if small pre-amplifiers were taped directly to the subject. Although, from experience in our group, it has been found that the discomfort and time requirements for subjects grows increasingly as the number of indwelling electrodes increases (Forbes et al., 2018). In Chapter 2 and 3, a potential limitation was that the results were limited to a single seated posture with the subject looking straight ahead and in isometric conditions. It is not known how the results would apply to dynamic events or isometric moments in other postures. A potential limitation of Chapter 4, 5, and 6 is that each study was collected in one automotive seat. In Chapter four, the head postures were collected while driving a 2010 Subaru Impreza and in Chapters 5 and 6, a 2005 Volvo S40 driver’s seat was mounted to the perturbation sled. The geometry of the vehicle’s mirrors and windows may have an influence on head angles measured in Chapter 4, and the type of seat could influence loads applied to the torso and potentially change the head and neck responses in Chapters 5 and 6. While the experiments were conducted with relevant automotive boundary conditions, they may not generalize to all vehicles on the road. Limitations specific to Chapters 5 and 6 were the perturbation characteristics. We limited the collision pulses to a 0.78 m/s speed change and a peak acceleration of 2.1g, to minimize the risk of injury to our subjects. It is not known how these results would scale to more severe and potentially injurious impact 117 scenarios. Also, subjects were aware that a perturbation would occur within a 5-10s window, which we did to ensure a consistent timing between conditions. Therefore, the results may not represent a fully unaware driver (Siegmund et al., 2003). Finally, the experiments in Chapters 2 and 3 were conducted with male subjects and we do not have enough subjects to compare our results between males and females in Chapters 5 and 6. Females have been shown to have a higher risk of whiplash injury (Carstensen et al., 2012; Jakobsson et al., 2004; Krafft et al., 2003). It is important for future work continuing that presented in this dissertation to focus on female volunteers. 7.4 Future Research 7.4.1 Future volunteer experimental work Autonomous vehicles are likely to increase their presence on public roads, but it remains unclear how long it will take for widespread adoption. People have been found to be less tolerant of mistakes made by artificial intelligence when compared to humans (Prahl & Van Swol, 2017), and this may translate to people being less tolerant of injuries resulting from autonomous vehicles when compared to injuries caused by human driver error. Even without mistakes by autonomous vehicles, there will likely still be situations where a collision is unavoidable due to the physical limitations of the vehicle and the operating environment. Because autonomous vehicles are unlikely to avoid injuries completely and the public will be potentially less tolerant to these injuries, passive safety will remain an important aspect of autonomous vehicles. A complication of autonomous vehicle collisions is that the state (i.e. position, posture and muscle contraction levels) of the occupants involved in autonomous crashes will likely be further from the sitting straight up, looking straight ahead, and relaxed state explored in most previous volunteer sled impact tests. It will be beneficial for future work on human volunteer responses to whiplash-like perturbations to focus on postures and situations relevant to occupants of autonomous vehicles. Many of these situations will be almost immediately relevant to cars on the road today as semi-autonomous features become common. For human volunteer perturbation experiments, the first suggestion for a load case that is relevant to semi-autonomous or fully autonomous vehicles is performing braking and/or steering maneuvers on unaware occupants before an impact. Researchers have quantified how autonomous steering and braking influence the movements of volunteers which result in non-neutral postures before a potential collision, which may increase the chance of injury (Ghaffari et al., 2018; Ólafsdóttir et al., 2013; Östh et 118 al., 2012; van Rooij et al., 2013). Although, none of these studies have followed up the pre-impact maneuvers with an impact-like perturbation to assess the potential change in injury risk. Volunteer testing is relevant for this type of collision scenario because muscle activity is necessary to represent the position of occupants after relatively long-duration steering or braking maneuvers prior to impact. Further non-neutral postures that could be more common for semi-autonomous or autonomous vehicles and should be investigated in human volunteer perturbation experiments include looking downwards at an electronic device, and reclined postures. 7.4.2 Future computational work Computational human body models are a valuable tool that is currently being utilized by automotive manufacturers to develop improved passive safety features in their vehicles. A potential criticism of these models is that they can only be used to reproduce physical tests that have already been conducted due to the required validation, limiting their applicability to understanding novel situations. The muscle activity response of computational human body models, even when using advanced feedback control methods (Meijer et al., 2013; Östh et al., 2015), needs to be tuned to a specific load case as discussed in the ‘Implications for modellers’ section of this chapter. Another criticism of human body models is that by trying to represent an average occupant, they are not modeling the potentially most vulnerable occupants. Extensive work is being conducted to morph human body models into a variety of body sizes and proportions (Hu, 2018; Hwang et al., 2016; John et al., 2019; Zhang et al., 2017). Although beyond geometric variability, humans also display marked variability in their neuromuscular responses as can be seen in the EMG measurements in Chapters 3, 5, and 6, and this variability is not currently being modelled. Chapter 3 in particular showed that variability in subjects’ neural control was greater than the variability in their biomechanical line of action. The limitations of using human body models in novel situations and the lack of representative muscle response variability could be potentially addressed with a modelled muscle controller that more closely mimics human motor control. Advances in this area have been made where physiological constants were used to mimic the frequency response of human VCR and CCR in a multibody model of the head-neck (Happee et al., 2017). A potential solution to further the physiological basis of human body model muscle controllers is an implementation of optimal feedback control (review: S. H. Scott, 2012; Todorov & Jordan, 2002). This control paradigm has been shown to match many features of human motor control, while only programming behavioural level task objectives. The details of optimal feedback control theory applied 119 to human motor control are beyond the scope of this thesis. Briefly, however, the noisy and delayed sensory feedback from the body is combined with a copy of the output motor commands in a Kalman filter to establish the optimal state estimation, which is then combined with task objectives in the optimal feedback control law to produce motor commands. In such a model, stochastic noise could be added to the measurement of joint angles or muscle length, and the task objectives could be manipulated to generate a distribution of muscle activation responses for the same load inputs, which could potentially provide insight into the response of a wider range of occupants. Further, because feedback gains are decided by the controller, tuning of the controller to each load case may not be necessary, and thus this muscle controller could potentially be used in novel situations. More research is required to understand how the task objectives should be defined for vehicle occupants, and how such a model of human motor control accounts for the startle response, which has been shown to be part of exposure to whiplash-like perturbations (Blouin et al., 2006a, 2006b). Also, the optimal state estimator requires a mathematically invertible human body model to estimate sensory consequences of motor outputs, and it is not clear how this could be accomplished with today’s complex finite element or multibody human models used in vehicle safety research. 7.5 Conclusions The experiments presented here improve our understanding of neck muscle biomechanics and neuromuscular responses in a context applied to frontal and rear end automotive collisions. In this dissertation I have presented how neck strength scales in non-principal-axis directions, fundamental aspects of the relationship between neck muscle biomechanics and neural control, and how neck muscles and head/neck kinematics respond to impacts when the head is turned or the occupant braces themselves by pushing against the steering wheel. The results of this dissertation show that occupant pre-impact behaviours are important to predict their impact response and potential for whiplash injury. Further, pre-impact behaviours may become more diverse with increasing levels of autonomy in passenger vehicles and need to be taken into account to improve vehicle safety. For modelling research, this body of work can be used to validate omni-directional biomechanics, inform muscle controller design, and validate models exposed to impacts for occupants in non-neutral head postures or braced postures. These experiments will help inform injury prevention methods, improve computational human body models, and ultimately lead to safer automobiles. 120 Bibliography Abbott, R., Pedler, A., Sterling, M., Hides, J., Murphey, T., Hoggarth, M., & Elliott, J. M. (2015). The geography of fatty infiltrates within the cervical multifidus and semispinalis cervicis in individuals with chronic whiplash-associated disorders. The Journal of Orthopaedic and Sports Physical Therapy, 45(4), 281–288. Abrahams, V. C. (1977). The physiology of neck muscles; their role in head movement and maintenance of posture. Canadian Journal of Physiology and Pharmacology, 55(3), 332–338. Ackland, D. C., Merritt, J. S., & Pandy, M. G. (2011). Moment arms of the human neck muscles in flexion, bending and rotation. Journal of Biomechanics, 44(3), 475–486. Alexander, C. M., & Harrison, P. J. (2002). The bilateral reflex control of the trapezius muscle in humans. Experimental Brain Research, 142(3), 418–424. Anderson, J. S., Hsu, A. W., & Vasavada, A. N. (2005). Morphology, architecture, and biomechanics of human cervical multifidus. Spine, 30(4), E86-91. Anderson, M. E. (1977). Segmental reflex inputs to motoneurons innervating dorsal neck musculature in the cat. Experimental Brain Research, 28(1–2), 175–187. Anderst, W. J., Donaldson, W. F., Lee, J. Y., & Kang, J. D. (2015). Cervical motion segment contributions to head motion during flexion\extension, lateral bending, and axial rotation. The Spine Journal, 15(12), 2538–2543. Baldwin, G., Mahony, R., & Trumpf, J. (2007). Complementary filter design on the Special Euclidean group SE(3). In European Control Conference (ECC) (pp. 3763–3770). Kos, Greece. Barker, J. B., Cronin, D. S., & Nightingale, R. W. (2017). Lower cervical spine motion segment computational model validation: Kinematic and kinetic response for quasi-static and dynamic loading. Journal of Biomechanical Engineering, 139(6), 61009. Barnsley, L., Lord, S. M., Wallis, B. J., & Bogduk, N. (1995). The prevalence of chronic cervical zygapophysial joint pain after whiplash. Spine, 20(1), 20–26. 121 Bartz, P. (2016). Building an AHRS using the SparkFun “9DOF Razor IMU” or “9DOF Sensor Stick.” Retrieved April 1, 2018, from https://github.com/Razor-AHRS/razor-9dof-ahrs/wiki/tutorial Batschelet, E. (1981). Circular Statistics in Biology. Academic Press. Beattie, N., & Lovell, M. E. (2010). Can patients with low energy whiplash associated disorder develop low back pain? Injury, 41(2), 144–146. Beeman, S. M., Kemper, A. R., & Duma, S. M. (2016). Neck forces and moments of human volunteers and post mortem human surrogates in low-speed frontal sled tests. Traffic Injury Prevention, 17(sup1), 141–149. Beeman, S. M., Kemper, A. R., Madigan, M. L., & Duma, S. M. (2011). Effects of bracing on human kinematics in low-speed frontal sled tests. Annals of Biomedical Engineering, 39(12), 2998. Berglund, A., Alfredsson, L., Jensen, I., Bodin, L., & Nygren, A. (2003). Occupant- and crash-related factors associated with the risk of whiplash injury. Annals of Epidemiology, 13(1), 66–72. Bernstein, N. A. (1967). The Co-Ordination and Regulation of Movements. Oxford, UK: Pergamon Press. Binder-Macleod, S. A., Halden, E. E., & Jungles, K. A. (1995). Effects of stimulation intensity on the physiological responses of human motor units. Medicine and Science in Sports and Exercise, 27(4), 556–565. Binder-Macleod, S. A., Lee, S. C., Fritz, A. D., & Kucharski, L. J. (1998). New look at force-frequency relationship of human skeletal muscle: effects of fatigue. Journal of Neurophysiology, 79(4), 1858–1868. Blouin, J.-S., Descarreaux, M., Bélanger-Gravel, A., Simoneau, M., & Teasdale, N. (2003). Attenuation of human neck muscle activity following repeated imposed trunk-forward linear acceleration. Experimental Brain Research, 150(4), 458–464. Blouin, J.-S., Inglis, J. T., & Siegmund, G. P. (2006a). Auditory startle alters the response of human subjects exposed to a single whiplash-like perturbation. Spine, 31(2), 146–154. Blouin, J.-S., Inglis, J. T., & Siegmund, G. P. (2006b). Startle responses elicited by whiplash perturbations. 122 The Journal of Physiology, 573(Pt 3), 857–867. Blouin, J.-S., Siegmund, G. P., Carpenter, M. G., & Inglis, J. T. (2007). Neural control of superficial and deep neck muscles in humans. Journal of Neurophysiology, 98(2), 920–928. Blouin, J.-S., Siegmund, G. P., & Inglis, J. T. (2007). Interaction between acoustic startle and habituated neck postural responses in seated subjects. Journal of Applied Physiology, 102(4), 1574–1586. Bogduk, N. (2002). Innervation and Pain Patterns of the Cervical Spine. In R. Grant (Ed.), Physical Therapy of the Cervical and Thoracic Spine (3rd ed., pp. 1–12). Churchill Livingstone. Bogduk, N., & Mercer, S. (2000). Biomechanics of the cervical spine. I: Normal kinematics. Clinical Biomechanics, 15(9), 633–648. Bose, D., Crandall, J. R., Untaroiu, C. D., & Maslen, E. H. (2010). Influence of pre-collision occupant parameters on injury outcome in a frontal collision. Accident Analysis & Prevention, 42(4), 1398–1407. Boyd-Clark, L. C., Briggs, C. A., & Galea, M. P. (2002). Muscle spindle distribution, morphology, and density in longus colli and multifidus muscles of the cervical spine. Spine, 27(7), 694–701. Brault, J. R., Siegmund, G. P., & Wheeler, J. B. (2000). Cervical muscle response during whiplash: evidence of a lengthening muscle contraction. Clinical Biomechanics, 15(6), 426–435. Brolin, K., Halldin, P., & Leijonhufvud, I. (2005). The effect of muscle activation on neck response. Traffic Injury Prevention, 6(1), 67–76. Automotive Technology & Engineering; Biomechanics; Road Transport Industries; Transportation Engineering. Brown, P., Rothwell, J. C., Thompson, P. D., Britton, T. C., Day, B. L., & Marsden, C. D. (1991). New observations on the normal auditory startle reflex in man. Brain : A Journal of Neurology, 114 ( Pt 4, 1891–1902. Buchanan, T. S., Rovai, G. P., & Rymer, W. Z. (1989). Strategies for muscle activation during isometric torque generation at the human elbow. Journal of Neurophysiology, 62(6), 1201–1212. https://doi.org/10.1152/jn.1989.62.6.1201 123 Cagnie, B., Cools, A., De Loose, V., Cambier, D., & Danneels, L. (2007). Differences in isometric neck muscle strength between healthy controls and women with chronic neck pain: the use of a reliable measurement. Archives of Physical Medicine and Rehabilitation, 88(11), 1441–1445. Carlson, E. J., Tominaga, Y., Ivancic, P. C., & Panjabi, M. M. (2007). Dynamic vertebral artery elongation during frontal and side impacts. The Spine Journal, 7(2), 222–228. Carriot, J., Jamali, M., Cullen, K. E., & Chacron, M. J. (2017). Envelope statistics of self-motion signals experienced by human subjects during everyday activities: Implications for vestibular processing. PloS One, 12(6), e0178664. Carstensen, T. B. W., Frostholm, L., Oernboel, E., Kongsted, A., Kasch, H., Jensen, T. S., & Fink, P. (2012). Are there gender differences in coping with neck pain following acute whiplash trauma? A 12-month follow-up study. European Journal of Pain, 16(1), 49–60. Cavanaugh, J. M. (2000). Neurophysiology and Neuroanatomy of Neck Pain. In N. Yoganandan & F. A. Pintar (Eds.), Frontiers in Whiplash Trauma: Clinical and Biomechanical (pp. 79–96). Amsterdam: IOS Press. Chan, Y. S., Kasper, J., & Wilson, V. J. (1987). Dynamics and directional sensitivity of neck muscle spindle responses to head rotation. Journal of Neurophysiology, 57(6), 1716–1729. Chancey, V. C., Nightingale, R. W., Van Ee, C. A., Knaub, K. E., & Myers, B. S. (2003). Improved estimation of human neck tensile tolerance: reducing the range of reported tolerance using anthropometrically correct muscles and optimized physiologic initial conditions. Stapp Car Crash Journal, 47, 135–153. Choi, H. Y., Sah, S. J., Lee, B., Cho, H. S., Kang, S. J., Mun, M. S., … Lee, J. (2005). Experimental and numerical studies of muscular activations of bracing occupant. In 19th ESV Conference, Paper. Cox, C. A., Dibb, A. T., Cutcliffe, H. C., Nightingale, R. W., Myers, B. S., Vasavada, A. N., … Bass, C. R. (2014). The influence of muscle modeling methods and paths on head and neck response. In International Center for Numerical Methods in Engineering. Crago, P. E., Peckham, P. H., & Thrope, G. B. (1980). Modulation of muscle force by recruitment during intramuscular stimulation. IEEE Transactions on Bio-Medical Engineering, 27(12), 679–684. 124 Crisco, J. J., Panjabi, M. M., Yamamoto, I., & Oxland, T. R. (1992). Euler stability of the human ligamentous lumbar spine. Part II: Experiment. Clinical Biomechanics (Bristol, Avon), 7(1), 27–32. Cronin, D. S. (2014). Finite element modeling of potential cervical spine pain sources in neutral position low speed rear impact. Journal of the Mechanical Behavior of Biomedical Materials, 33, 55–66. Curatolo, M., Bogduk, N., Ivancic, P. C., McLean, S. A., Siegmund, G. P., & Winkelstein, B. A. (2011). The role of tissue damage in whiplash-associated disorders: discussion paper 1. Spine, 36(25 Suppl), S309-15. Daenen, L., Nijs, J., Raadsen, B., Roussel, N., Cras, P., & Dankaerts, W. (2013). Cervical motor dysfunction and its predictive value for long-term recovery in patients with acute whiplash-associated disorders: A systematic review. Journal of Rehabilitation Medicine, 45(2), 113–122. Davidsson, J., Deutscher, C., Hell, W., Lövsund, P., & Svensson, M. (2001). Human volunteer kinematics in rear-end sled collisions. Traffic Injury Prevention, 2(4), 319–333. Automotive Technology & Engineering; Biomechanics; Road Transport Industries; Transportation Engineering. de Bruijn, E., der Helm, F. C. T., & Happee, R. (2016). Analysis of isometric cervical strength with a nonlinear musculoskeletal model with 48 degrees of freedom. Multibody System Dynamics, 36(4), 339–362. De Jager, M., Sauren, A., Thunnissen, J., & Wismans, J. (1994). A three-dimensional head-neck model: validation for frontal and lateral impacts. SAE Transactions, 1660–1676. Deans, G. T., Magalliard, J. N., Kerr, M., & Rutherford, W. H. (1987). Neck sprain - a major cause of disability following car accidents. Injury, 18(1), 10–12. Demandt, B. (2018). Global car sales analysis 2017. Retrieved April 2, 2018, from http://carsalesbase.com/global-car-sales-2017/ Deng, B., Begeman, P. C., Yang, K. H., Tashman, S., & King, A. I. (2000). Kinematics of human cadaver cervical spine during low speed rear-end impacts. Stapp Car Crash Journal, 44, 171–188. Deng, Y.-C., & Goldsmith, W. (1987). Response of a human head/neck/upper-torso replica to dynamic loading—II. Analytical/numerical model. Journal of Biomechanics, 20(5), 487–497. 125 Dibb, A. T., Cox, C. A., Nightingale, R. W., Luck, J. F., Cutcliffe, H. C., Myers, B. S., … Bass, C. R. (2013). Importance of muscle activations for biofidelic pediatric neck response in computational models. Traffic Injury Prevention, 14 Suppl, S116-27. Dong, L., & Winkelstein, B. A. (2010). Simulated whiplash modulates expression of the glutamatergic system in the spinal cord suggesting spinal plasticity is associated with painful dynamic cervical facet loading. Journal of Neurotrauma, 27(1), 163–174. Eichberger, A., Darok, M., Steffan, H., Leinzinger, P. E., Boström, O., & Svensson, M. Y. (2000). Pressure measurements in the spinal canal of post-mortem human subjects during rear-end impact and correlation of results to the neck injury criterion. Accident; Analysis and Prevention, 32(2), 251–260. Ejima, S., Ito, D., Satou, F., Mikami, K., Ono, K., Kaneoka, K., & Shiina, I. (2012). Effects of Pre‐impact Swerving/Steering on Physical Motion of the Volunteer in the Low‐Speed Side‐impact Sled Test. In Ircobi (pp. 352–366). Ejima, S., Ono, K., Holcombe, S., Kaneoka, K., & Fukushima, M. (2007). A study on occupant kinematics behaviour and muscle activities during pre-impact braking based on volunteer tests. In Proceedings of ircobi (international research council on the biomechanics of injury) conference 2007, held maastricht, the netherlands, september 2007. Ejima, S., Zama, Y., Satou, F., Holcombe, S., Ono, K., Kaneoka, K., & Shiina, I. (2008). Prediction of the physical motion of the human body based on muscle activity during pre-impact braking. In Proceedings of the IRCOBI Conference (pp. 163–175). Elliott, J. M. (2011). Are there implications for morphological changes in neck muscles after whiplash injury? Spine, 36(25 Suppl), S205-10. Elliott, J. M., Jull, G., Noteboom, J. T., Darnell, R., Galloway, G., & Gibbon, W. W. (2006). Fatty Infiltration in the Cervical Extensor Muscles in Persistent Whiplash-Associated Disorders. Spine, 31(22), E847–E855. Euston, M., Coote, P., Mahony, R., Jonghyuk Kim, & Hamel, T. (2008). A complementary filter for attitude estimation of a fixed-wing UAV. In 2008 IEEE/RSJ International Conference on Intelligent 126 Robots and Systems (pp. 340–345). IEEE. Falla, D., Bilenkij, G., & Jull, G. (2004). Patients with chronic neck pain demonstrate altered patterns of muscle activation during performance of a functional upper limb task. Spine, 29(13), 1436–1440. Farshadmanesh, F., Byrne, P., Keith, G. P., Wang, H., Corneil, B. D., & Crawford, J. D. (2012). Cross-validated models of the relationships between neck muscle electromyography and three-dimensional head kinematics during gaze behavior. Journal of Neurophysiology, 107(2), 573–590. Farshadmanesh, F., Byrne, P., Wang, H., Corneil, B. D., & Crawford, J. D. (2012). Relationships between neck muscle electromyography and three-dimensional head kinematics during centrally induced torsional head perturbations. Journal of Neurophysiology, 108(11), 2867–2883. Fice, J. B., Blouin, J. S., & Siegmund, G. P. (2018). Head postures during naturalistic driving. Traffic Injury Prevention, 19(6), 637–643. https://doi.org/10.1080/15389588.2018.1493582 Fice, J. B., & Cronin, D. S. (2012). Investigation of whiplash injuries in the upper cervical spine using a detailed neck model. Journal of Biomechanics, 45(6), 1098–1102. Fice, J. B., Cronin, D. S., & Panzer, M. B. (2011). Cervical spine model to predict capsular ligament response in rear impact. Annals of Biomedical Engineering, 39(8), 2152–2162. https://doi.org/10.1007/s10439-011-0315-4 Fice, J. B., Siegmund, G. P., & Blouin, J.-S. (2014). Prediction of three dimensional maximum isometric neck strength. Annals of Biomedical Engineering, 42(9), 1846–1852. https://doi.org/10.1007/s10439-014-1046-0 Fice, J. B., Siegmund, G. P., & Blouin, J.-S. (2018). Neck muscle biomechanics and neural control. Journal of Neurophysiology, 120(1), 361–371. Fisher, N. I., Lewis, T., & Embleton, B. J. J. (1987). Statistical Analysis of Spherical Data. Cambridge: Cambridge University Press. Fitzpatrick, R. C., & Day, B. L. (2004). Probing the human vestibular system with galvanic stimulation. Journal of Applied Physiology (Bethesda, Md. : 1985), 96(6), 2301–2316. 127 Forbes, P. A., Dakin, C. J., Vardy, A. N., Happee, R., Siegmund, G. P., Schouten, A. C., & Blouin, J.-S. (2013). Frequency response of vestibular reflexes in neck, back, and lower limb muscles. Journal of Neurophysiology, 110(8), 1869–1881. Forbes, P. A., Fice, J. B., Siegmund, G. P., & Blouin, J. S. (2018). Electrical vestibular stimuli evoke robust muscle activity in deep and superficial neck muscles in humans. Frontiers in Neurology, 9(JUL). https://doi.org/10.3389/fneur.2018.00535 Forssberg, H., & Hirschfeld, H. (1994). Postural adjustments in sitting humans following external perturbations: muscle activity and kinematics. Experimental Brain Research, 97(3), 515–527. Franklin, D. W., & Wolpert, D. M. (2008). Specificity of reflex adaptation for task-relevant variability. Journal of Neuroscience, 28(52), 14165–14175. Franklin, D. W., & Wolpert, D. M. (2011). Computational mechanisms of sensorimotor control. Neuron, 72(3), 425–442. https://doi.org/10.1016/j.neuron.2011.10.006 Fung, Y. C. (1993). Biomechanics: mechanical properties of living tissues (2md ed.). New York: Springer. Gabriel, D. A., Matsumoto, J. Y., Davis, D. H., Currier, B. L., & An, K.-N. (2004). Multidirectional neck strength and electromyographic activity for normal controls. Clinical Biomechanics (Bristol, Avon), 19(7), 653–658. Galasko, C. S., Murray, P. M., Pitcher, M., Chambers, H., Mansfield, S., Madden, M., … Hodson, M. (1993). Neck sprains after road traffic accidents: a modern epidemic. Injury, 24(3), 155–157. Garrett, W. E. (1996). Muscle strain injuries. The American Journal of Sports Medicine, 24(6 Suppl), S2-8. Ghaffari, G., Brolin, K., Bråse, D., Pipkorn, B., Svanberg, B., Jakobsson, L., & Davidsson, J. (2018). Passenger kinematics in Lane change and Lane change with Braking Manoeuvres using two belt configurations: standard and reversible pre‐pretensioner. IRCOBI Conference 2018, 493–511. Gibson, T., Bogduk, N., Macpherson, J., & McIntosh, A. (2000). Crash Characteristics of Whiplash Associated Chronic Neck Pain. Journal of Musculoskeletal Pain, 8(1–2), 87–95. Goldberg, J. M., & Cullen, K. E. (2011). Vestibular control of the head: possible functions of the 128 vestibulocollic reflex. Experimental Brain Research, 210(3–4), 331–345. Grauer, J. N., Panjabi, M. M., Cholewicki, J., Nibu, K., & Dvorak, J. (1997). Whiplash produces an S-shaped curvature of the neck with hyperextension at lower levels. Spine, 22(21), 2489–2494. Gresty, M. A. (1974). Coordination of head and eye movements to fixate continuous and intermittent targets. Vision Research, 14(6), 395–403. Grosse, P., & Brown, P. (2003). Acoustic startle evokes bilaterally synchronous oscillatory EMG activity in the healthy human. Journal of Neurophysiology, 90(3), 1654–1661. Happee, R. (1994). Inverse dynamic optimization including muscular dynamics, a new simulation method applied to goal directed movements. Journal of Biomechanics, 27(7), 953–960. Happee, R., de Bruijn, E., Forbes, P. A., & van der Helm, F. C. T. (2017). Dynamic head-neck stabilization and modulation with perturbation bandwidth investigated using a multisegment neuromuscular model. Journal of Biomechanics, 58, 203–211. Harris, C. M., & Wolpert, D. M. (1998). Signal-dependent noise determines motor planning. Nature, 394(6695), 780–784. https://doi.org/10.1038/29528 Hault-Dubrulle, A., Robache, F., Pacaux, M.-P., & Morvan, H. (2011). Determination of pre-impact occupant postures and analysis of consequences on injury outcome. Part I: A driving simulator study. Accident Analysis & Prevention, 43(1), 66–74. Herzog, W. (2013). Mechanisms of enhanced force production in lengthening (eccentric) muscle contractions. Journal of Applied Physiology, 116(11), 1407–1417. Hill, A. V. (1938). The heat of shortening and the dynamic constants of muscle. Proc. R. Soc. Lond. B, 126(843), 136–195. Hoffman, D. S., & Strick, P. L. (1999). Step-tracking movements of the wrist. IV. Muscle activity associated with movements in different directions. Journal of Neurophysiology, 81(1), 319–333. Hogan, N. (1984). Adaptive control of mechanical impedance by coactivation of antagonist muscles. IEEE Transactions on Automatic Control, 29(8), 681–690. 129 Hu, J. (2018). Parametric human modeling. In Basic finite element method as applied to injury biomechanics (pp. 417–445). Elsevier. Hunter, B. V, Thelen, D. G., & Dhaher, Y. Y. (2009). A three-dimensional biomechanical evaluation of quadriceps and hamstrings function using electrical stimulation. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 17(2), 167–175. Hwang, E., Hu, J., Chen, C., Klein, K. F., Miller, C. S., Reed, M. P., … Hallman, J. J. (2016). Development, evaluation, and sensitivity analysis of parametric finite element whole-body human models in side impacts. Inami, S., Shiga, T., Tsujino, A., Yabuki, T., Okado, N., & Ochiai, N. (2001). Immunohistochemical demonstration of nerve fibers in the synovial fold of the human cervical facel joint. Journal of Orthopaedic Research, 19(4), 593–596. Ito, S., Ivancic, P. C., Pearson, A. M., Tominaga, Y., Gimenez, S. E., Rubin, W., & Panjabi, M. M. (2005). Cervical intervertebral disc injury during simulated frontal impact. European Spine Journal, 14(4), 356–365. Ivancic, P. C. (2012). Cervical neural space narrowing during simulated rear crashes with anti-whiplash systems. European Spine Journal, 21(5), 879–886. Ivancic, P. C., Ito, S., Tominaga, Y., Carlson, E. J., Rubin, W., & Panjabi, M. M. (2006). Effect of rotated head posture on dynamic vertebral artery elongation during simulated rear impact. Clinical Biomechanics, 21(3), 213–220. Ivancic, P. C., Ito, S., Tominaga, Y., Rubin, W., Coe, M. P., Ndu, A. B., … Panjabi, M. M. (2008). Whiplash causes increased laxity of cervical capsular ligament. Clinical Biomechanics, 23(2), 159–165. Iwamoto, M., & Nakahira, Y. (2015). Development and validation of the Total HUman Model for Safety (THUMS) version 5 containing multiple 1d muscles for estimating occupant motions with muscle activation during side impacts. Iwamoto, M., Nakahira, Y., Kimpara, H., Sugiyama, T., & Min, K. (2012). Development of a human body finite element model with multiple muscles and their controller for estimating occupant motions and impact responses in frontal crash situations. Stapp Car Crash Journal, 56(October), 231–268. 130 Jakobsson, L., Isaksson-Hellman, I., & Lindman, M. (2008). WHIPS (Volvo cars’ Whiplash Protection System)-the development and real-world performance. Traffic Injury Prevention, 9(6), 600–605. Jakobsson, L., Norin, H., & Svensson, M. Y. (2004). Parameters influencing AIS 1 neck injury outcome in frontal impacts. Traffic Injury Prevention, 5(2), 156–163. Jang, J.-H., & Han, C.-S. (1997). The sensitivity analysis of yaw rate for a front wheel steering vehicle: In the frequency domain. KSME International Journal, 11(1), 56–66. John, J. D., Kumar, G. S., & Yoganandan, N. (2019). Cervical spine morphology and ligament property variations: A finite element study of their influence on sagittal bending characteristics. Journal of Biomechanics. Jones, K. E., Hamilton, A. F., & Wolpert, D. M. (2002). Sources of signal-dependent noise during isometric force production. Journal of Neurophysiology, 88(3), 1533–1544. https://doi.org/10.1152/jn.2002.88.3.1533 Jordan, A., Mehlsen, J., Bülow, P. M., Ostergaard, K., & Danneskiold-Samsøe, B. (1999). Maximal isometric strength of the cervical musculature in 100 healthy volunteers. Spine, 24(13), 1343–1348. Juul-Kristensen, B., Clausen, B., Ris, I., Jensen, R. V., Steffensen, R. F., Chreiteh, S. S., … Søgaard, K. (2013). Increased neck muscle activity and impaired balance among females with whiplash-related chronic neck pain: a cross-sectional study. Journal of Rehabilitation Medicine, 45(4), 376–384. Kaale, B. R., Krakenes, J., Albrektsen, G., & Wester, K. (2005a). Head position and impact direction in whiplash injuries: associations with MRI-verified lesions of ligaments and membranes in the upper cervical spine. Journal of Neurotrauma, 22(11), 1294–1302. Kaale, B. R., Krakenes, J., Albrektsen, G., & Wester, K. (2005b). Whiplash-associated disorders impairment rating: neck disability index score according to severity of MRI findings of ligaments and membranes in the upper cervical spine. Journal of Neurotrauma, 22(4), 466–475. Kamibayashi, L. K., & Richmond, F. J. (1998). Morphometry of human neck muscles. Spine, 23(12), 1314–1323. Kaneoka, K., Ono, K., Inami, S., & Hayashi, K. (1999). Motion analysis of cervical vertebrae during 131 whiplash loading. Spine, 24(8), 763–770. Karimi, N., Rezasoltani, A., Rahnama, L., Noori-Kochi, F., & Jaberzadeh, S. (2016). Ultrasonographic analysis of dorsal neck muscles thickness changes induced by isometric contraction of shoulder muscles: a comparison between patients with chronic neck pain and healthy controls. Manual Therapy, 22, 174–178. Kemper, A. R., Beeman, S. M., Madigan, M. L., & Duma, S. M. (2014). Human occupants in low-speed frontal sled tests: effects of pre-impact bracing on chest compression, reaction forces, and subject acceleration. Traffic Injury Prevention, 15 Suppl 1, S141-50. Kent, J. T. (1982). The Fisher-Bingham distribution on the sphere. Journal of the Royal Statistical Society. Series B (Methodological) Journal of the Royal Statistical Society. Series B J. R. Statist. Soc. B, 44(1), 71–80. Keshner, E. A., Campbell, D., Katz, R. T., & Peterson, B. W. (1989). Neck muscle activation patterns in humans during isometric head stabilization. Experimental Brain Research, 75(2), 335–344. Keshner, E. A., Hain, T. C., & Chen, K. J. (1999). Predicting control mechanisms for human head stabilization by altering the passive mechanics. Journal of Vestibular Research : Equilibrium & Orientation, 9(6), 423–434. Keshner, E. A., & Peterson, B. W. (1995). Mechanisms controlling human head stabilization. I. Head-neck dynamics during random rotations in the horizontal plane. Journal of Neurophysiology, 73(6), 2293–2301. Knaub, K. E., & Myers, B. S. (1998). Cervical Spine Muscle. Krafft, M., Kullgren, A., Lie, A., & Tingvall, C. (2003). The risk of whiplash injury in the rear seat compared to the front seat in rear impacts. Traffic Injury Prevention, 4(2), 136–140. Krakenes, J., & Kaale, B. R. (2006). Magnetic resonance imaging assessment of craniovertebral ligaments and membranes after whiplash trauma. Spine, 31(24), 2820–2826. Kumar, S., Ferrari, R., & Narayan, Y. (2005). Effect of head rotation in whiplash-type rear impacts. The Spine Journal : Official Journal of the North American Spine Society, 5(2), 130–139. 132 Kumar, S., Narayan, Y., & Amell, T. (2001). Cervical strength of young adults in sagittal, coronal, and intermediate planes. Clinical Biomechanics (Bristol, Avon), 16(5), 380–388. Kurtzer, I., Pruszynski, J. A., Herter, T. M., & Scott, S. H. (2006). Primate upper limb muscles exhibit activity patterns that differ from their anatomical action during a postural task. Journal of Neurophysiology, 95(1), 493–504. Lee, K. E., Franklin, A. N., Davis, M. B., & Winkelstein, B. A. (2006). Tensile cervical facet capsule ligament mechanics: failure and subfailure responses in the rat. Journal of Biomechanics, 39(7), 1256–1264. Lee, K. E., Thinnes, J. H., Gokhin, D. S., & Winkelstein, B. A. (2004). A novel rodent neck pain model of facet-mediated behavioral hypersensitivity: implications for persistent pain and whiplash injury. Journal of Neuroscience Methods, 137(2), 151–159. Leong, P., & Carlile, S. (1998). Methods for spherical data analysis and visualization. Journal of Neuroscience Methods, 80(2), 191–200. Li, Q., Shen, H., & Li, M. (2013). Magnetic resonance imaging signal changes of alar and transverse ligaments not correlated with whiplash-associated disorders: a meta-analysis of case-control studies. European Spine Journal, 22(1), 14–20. Lord, S. M., Barnsley, L., Wallis, B. J., & Bogduk, N. (1996). Chronic cervical zygapophysial joint pain after whiplash. A placebo-controlled prevalence study. Spine, 21(15), 1737–1745. Lu, Y., Chen, C., Kallakuri, S., Patwardhan, A., & Cavanaugh, J. M. (2005). Neurophysiological and biomechanical characterization of goat cervical facet joint capsules. Journal of Orthopaedic Research, 23(4), 779–787. Maak, T. G., Tominaga, Y., Panjabi, M. M., & Ivancic, P. C. (2006). Alar, transverse, and apical ligament strain due to head-turned rear impact. Spine, 31(6), 632–638. Mahony, R., Cha, S.-H., & Hamel, T. (2006). A coupled estimation and control analysis for attitude stabilisation of mini aerial vehicles. In Proceedings of the Australasian Conference on Robotics & Automation. Auckland, New Zealand. Mahony, R., Hamel, T., & Pflimlin, J.-M. (2008). Nonlinear Complementary Filters on the Special 133 Orthogonal Group. IEEE Transactions on Automatic Control, 53(5), 1203–1218. Manchikanti, L., Singh, V., Rivera, J., & Pampati, V. (2002). Prevalence of cervical facet joint pain in chronic neck pain. Pain Physician, 5(3), 243–249. Mang, D. W. H., Siegmund, G. P., & Blouin, J.-S. (2014). Whiplash evokes descending muscle recruitment and sympathetic responses characteristic of startle. The Journal of the Canadian Chiropractic Association, 58(2), 109–118. Mang, D. W. H., Siegmund, G. P., Brown, H. J., Goonetilleke, S. C., & Blouin, J.-S. (2015). Loud preimpact tones reduce the cervical multifidus muscle response during rear-end collisions: a potential method for reducing whiplash injuries. The Spine Journal : Official Journal of the North American Spine Society, 15(1), 153–161. Mang, D. W. H., Siegmund, G. P., Inglis, J. T., & Blouin, J.-S. (2012). The startle response during whiplash: a protective or harmful response? Journal of Applied Physiology, 113(4), 532–540. Markley, F. L., Cheng, Y., Crassidis, J. L., & Oshman, Y. (2007). Averaging Quaternions. Journal of Guidance, Control, and Dynamics, 30(4), 1193–1197. Matsumoto, M., Okada, E., Ichihara, D., Chiba, K., Toyama, Y., Fujiwara, H., … Takahata, T. (2010). Prospective ten-year follow-up study comparing patients with whiplash-associated disorders and asymptomatic subjects using magnetic resonance imaging. Spine, 35(18), 1684–1690. Mattucci, S. F. E., Moulton, J. A., Chandrashekar, N., & Cronin, D. S. (2013). Strain rate dependent properties of human craniovertebral ligaments. Journal of the Mechanical Behavior of Biomedical Materials, 23, 71–79. Maurice, M., Gioanni, H., & Abourachid, A. (2006). Influence of the behavioural context on the optocollic reflex (OCR) in pigeons (Columba livia). The Journal of Experimental Biology, 209(Pt 2), 292–301. Mayoux-Benhamou, M. A., Wybier, M., & Revel, M. (1989). Strength and cross-sectional area of the dorsal neck muscles. Ergonomics, 32(5), 513–518. Meijer, M., Broos, J., Elrofai, H., de Bruijn, E., Forbes, P., & Happee, R. (2013). Modelling of bracing in a multi-body active human model. 134 Mortensen, J. D., Vasavada, A. N., & Merryweather, A. S. (2018). The inclusion of hyoid muscles improve moment generating capacity and dynamic simulations in musculoskeletal models of the head and neck. PloS One, 13(6), e0199912. Nikooyan, A. A., Veeger, H. E. J., Westerhoff, P., Bolsterlee, B., Graichen, F., Bergmann, G., & der Helm, F. C. T. (2012). An EMG-driven musculoskeletal model of the shoulder. Human Movement Science, 31(2), 429–447. Norris, S. H., & Watt, I. (1983). The prognosis of neck injuries resulting from rear-end vehicle collisions. The Journal of Bone and Joint Surgery. British Volume, 65(5), 608–611. Nozaki, D., Nakazawa, K., & Akai, M. (2005). Muscle activity determined by cosine tuning with a nontrivial preferred direction during isometric force exertion by lower limb. Journal of Neurophysiology, 93(5), 2614–2624. Ólafsdóttir, J. M., Brolin, K., Blouin, J.-S., & Siegmund, G. P. (2015). Dynamic spatial tuning of cervical muscle reflexes to multidirectional seated perturbations. Spine, 40(4), E211-9. Ólafsdóttir, J. M., Fice, J. B., Mang, D. W. H., Brolin, K., Davidsson, J., Blouin, J. S., & Siegmund, G. P. (2018). Trunk muscle recruitment patterns in simulated precrash events. Traffic Injury Prevention, 19(sup1), S186–S188. https://doi.org/10.1080/15389588.2018.1426917 Ólafsdóttir, J. M., Östh, J. K. H., Davidsson, J., & Brolin, K. B. (2013). Passenger Kinematics and Muscle Responses in Autonomous Braking Events with Standard and Reversible Pre‐tensioned Restraints, 46(0), 602–617. Ono, K., Kaneoka, K., Wittek, A., & Kajzer, J. (1997). Cervical injury mechanism based on the analysis of human cervical vertebral motion and head-neck-torso kinematics during low-speed rear impacts. Stapp Car Crash Journal, 41, 339–356. Östh, J., Brolin, K., & Bråse, D. (2015). A human body model with active muscles for simulation of pretensioned restraints in autonomous braking interventions. Traffic Injury Prevention, 16(3), 304–313. https://doi.org/10.1080/15389588.2014.931949 Östh, J., Brolin, K., Carlsson, S., Wismans, J., & Davidsson, J. (2012). The occupant response to autonomous braking: a modeling approach that accounts for active musculature. Traffic Injury 135 Prevention, 13(3), 265–277. Oude Nijhuis, L. B., Janssen, L., Bloem, B. R., van Dijk, J. G., Gielen, S. C., Borm, G. F., & Overeem, S. (2007). Choice reaction times for human head rotations are shortened by startling acoustic stimuli, irrespective of stimulus direction. The Journal of Physiology, 584(Pt 1), 97–109. Padgaokar, A. J. (1975). Measurement of angular acceleration of a rigid body using linear accelerometers. Transactions of the American Society of Mechanical Engineers, 75, 522–526. Padgaonkar, A. J., Krieger, K. W., & King, A. I. (1975). Measurement of Angular Acceleration of a Rigid Body Using Linear Accelerometers. Journal of Applied Mechanics, 42(3), 552. Panjabi, M. M., Cholewicki, J., Nibu, K., Grauer, J., Babat, L. B., & Dvorak, J. (1998). Critical load of the human cervical spine: an in vitro experimental study. Clinical Biomechanics (Bristol, Avon), 13(1), 11–17. Panjabi, M. M., Ito, S., Pearson, A. M., & Ivancic, P. C. (2004). Injury mechanisms of the cervical intervertebral disc during simulated whiplash. Spine, 29(11), 1217–1225. Panjabi, M. M., Ivancic, P. C., Maak, T. G., Tominaga, Y., & Rubin, W. (2006). Multiplanar cervical spine injury due to head-turned rear impact. Spine, 31(4), 420–429. Panjabi, M. M., Maak, T. G., Ivancic, P. C., & Ito, S. (2006). Dynamic intervertebral foramen narrowing during simulated rear impact. Spine, 31(5), 128–134. Panjabi, M. M., Pearson, A. M., Ito, S., Ivancic, P. C., Gimenez, S. E., & Tominaga, Y. (2004). Cervical spine ligament injury during simulated frontal impact. Spine, 29(21), 2395–2403. Panzer, M. B., & Cronin, D. S. (2009). C4-C5 segment finite element model development, validation, and load-sharing investigation. Journal of Biomechanics, 42(4), 480–490. Panzer, M. B., Fice, J. B., & Cronin, D. S. (2011). Cervical spine response in frontal crash. Medical Engineering and Physics, 33(9), 1147–1159. https://doi.org/10.1016/j.medengphy.2011.05.004 Pearson, A. M., Ivancic, P. C., Ito, S., & Panjabi, M. M. (2004). Facet joint kinematics and injury mechanisms during simulated whiplash. Spine, 29(4), 390–397. 136 Pearson, A. M., Panjabi, M. M., Ivancic, P. C., Ito, S., Cunningham, B. W., Rubin, W., & Gimenez, S. E. (2005). Frontal impact causes ligamentous cervical spine injury. Spine, 30(16), 1852–1858. Peng, G. C., Hain, T. C., & Peterson, B. W. (1996). A dynamical model for reflex activated head movements in the horizontal plane. Biological Cybernetics, 75(4), 309–319. Peterson, B. W., Choi, H., Hain, T., Keshner, E. A., & Peng, G. C. (2001). Dynamic and kinematic strategies for head movement control. Annals of the New York Academy of Sciences, 942, 381–393. Peterson, B. W., Goldberg, J., Bilotto, G., & Fuller, J. H. (1985). Cervicocollic reflex: its dynamic properties and interaction with vestibular reflexes. Journal of Neurophysiology, 54(1), 90–109. Pettersson, K., Hildingsson, C., Toolanen, G., Fagerlund, M., & Björnebrink, J. (1997). Disc pathology after whiplash injury. A prospective magnetic resonance imaging and clinical investigation. Spine, 22(3), 283–288. Popovic, D., Gordon, T., Rafuse, V. F., & Prochazka, A. (1991). Properties of implanted electrodes for functional electrical stimulation. Annals of Biomedical Engineering, 19(3), 303–316. Portero, P., Bigard, A. X., Gamet, D., Flageat, J. R., & Guézennec, C. Y. (2001). Effects of resistance training in humans on neck muscle performance, and electromyogram power spectrum changes. European Journal of Applied Physiology, 84(6), 540–546. Prahl, A., & Van Swol, L. (2017). Understanding algorithm aversion: When is advice from automation discounted? Journal of Forecasting, 36(6), 691–702. Premerlani, W., & Bizard, P. (2009). Direction Cosine Matrix IMU: Theory. Retrieved from http://owenson.me/build-your-own-quadcopter-autopilot/DCMDraft2.pdf Proske, U., & Allen, T. J. (2005). Damage to skeletal muscle from eccentric exercise. Exercise and Sport Sciences Reviews, 33(2), 98–104. Queisser, F., Blüthner, R., & Seidel, H. (1994). Control of positioning the cervical spine and its application to measuring extensor strength. Clinical Biomechanics, 9(3), 157–161. Quinlan, K. P., Annest, J. L., Myers, B., Ryan, G., & Hill, H. (2004). Neck strains and sprains among motor 137 vehicle occupants-United States, 2000. Accident Analysis and Prevention, 36(1), 21–27. Quinn, K. P., & Winkelstein, B. A. (2007). Cervical facet capsular ligament yield defines the threshold for injury and persistent joint-mediated neck pain. Journal of Biomechanics, 40(10), 2299–2306. Radanov, B. P., Sturengger, M., & Di Stefano, G. (1995). Long-Term Outcome after Whiplash Injury: A 2-Year Follow-Up Considering Features of Injury Mechanism and Somatic, Radiologic, and Psychosocial Findings. Medicine, 74(5), 281–297. Rahnama, L., Rezasoltani, A., Zavieh, M. K., NooriKochi, F., & Baghban, A. A. (2015). Differences in cervical multifidus muscle thickness during isometric contraction of shoulder muscles: a comparison between patients with chronic neck pain and healthy controls. Journal of Manipulative and Physiological Therapeutics, 38(3), 210–217. Rapoport, S. (1979). Reflex connexions of motoneurones of muscles involved in head movement in the cat. The Journal of Physiology, 289, 311–327. Reddy, M., Reddy, B., Schöggl, A., Saringer, W., & Matula, C. (2002). The complexity of trauma to the cranio-cervical junction: correlation of clinical presentation with Doppler flow velocities in the V3-segment of the vertebral arteries. Acta Neurochirurgica, 144(6), 575–580. Richmond, F. J., & Abrahams, V. C. (1975). Morphology and distribution of muscle spindles in dorsal muscles of the cat neck. Journal of Neurophysiology, 38(6), 1322–1339. Richmond, F. J., MacGillis, D. R., & Scott, D. A. (1985). Muscle-fiber compartmentalization in cat splenius muscles. Journal of Neurophysiology, 53(4), 868–885. Riener, R., Quintern, J., & Schmidt, G. (1996). Biomechanical model of the human knee evaluated by neuromuscular stimulation. Journal of Biomechanics, 29(9), 1157–1167. Roucoux, A., & Crommelinck, M. (1988). Control of head movement during visual orientation. Control of Head Movement. Oxford University Press, New York, 208–223. Ryan, G. A., Taylor, G. W., Moore, V. M., & Dolinis, J. (1994). Neck strain in car occupants: injury status after 6 months and crash-related factors. Injury, 25(8), 533–537. 138 Schap, J. M., Koya, B., & Gayzik, F. S. (2019). Objective Evaluation of Whole Body Kinematics in a Simulated, Restrained Frontal Impact. Annals of Biomedical Engineering, 47(2), 512–523. Schmidt, R. A., Zelaznik, H., Hawkins, B., Frank, J. S., & Quinn, J. T. (1979). Motor-output variability: A theory for the accuracy of rapid motor acts. Psychological Review, 86(5), 415–451. https://doi.org/10.1037/0033-295X.86.5.415 Scott, S. H. (2012). The computational and neural basis of voluntary motor control and planning. Trends in Cognitive Sciences, 16(11), 541–549. Scott, S., & Sanderson, P. L. (2002). Whiplash: a biochemical study of muscle injury. European Spine Journal, 11(4), 389–392. Šerić, V., Blažić-Čop, N., & Demarin, V. (2000). Haemodynamic changes in patients with whiplash injury measured by transcranial Doppler sonography (TCD). Collegium Antropologicum, 24(1), 197–204. Shateri, H., & Cronin, D. S. (2015). Out-of-Position Rear Impact Tissue-Level Investigation Using Detailed Finite Element Neck Model. Traffic Injury Prevention, 16(7), 698–708. Shemmell, J. (2015). Interactions between stretch and startle reflexes produce task-appropriate rapid postural reactions. Frontiers in Integrative Neuroscience, 9, 2. Shugg, J. A. J., Jackson, C. D., & Dickey, J. P. (2011). Cervical spine rotation and range of motion: pilot measurements during driving. Traffic Injury Prevention, 12(1), 82–87. Siegmund, G. P. (2001). The reflex response of human neck muscles to whiplash-like perturbations. University of British Columbia. Siegmund, G. P., Blouin, J.-S., Brault, J. R., Hedenstierna, S., & Inglis, J. T. (2007). Electromyography of superficial and deep neck muscles during isometric, voluntary, and reflex contractions. Journal of Biomechanical Engineering, 129(1), 66–77. Siegmund, G. P., Blouin, J.-S., Carpenter, M. G., Brault, J. R., & Inglis, J. T. (2008). Are cervical multifidus muscles active during whiplash and startle? An initial experimental study. BMC Musculoskeletal Disorders, 9, 80. 139 Siegmund, G. P., Davis, M. B., Quinn, K. P., Hines, E., Myers, B. S., Ejima, S., … Winkelstein, B. A. (2008). Head-turned postures increase the risk of cervical facet capsule injury during whiplash. Spine, 33(15), 1643–1649. Siegmund, G. P., Inglis, J. T., & Sanderson, D. J. (2001). Startle response of human neck muscles sculpted by readiness to perform ballistic head movements. The Journal of Physiology, 535(Pt 1), 289–300. Siegmund, G. P., King, D. J., Lawrence, J. M., Wheeler, J. B., Brault, J. R., & Smith, T. A. (1997). Head/Neck Kinematic Response of Human Subjects in Low-Speed Rear-End Collisions. Siegmund, G. P., Myers, B. S., Davis, M. B., Bohnet, H. F., & Winkelstein, B. A. (2001). Mechanical evidence of cervical facet capsule injury during whiplash: a cadaveric study using combined shear, compression, and extension loading. Spine, 26(19), 2095–2101. Siegmund, G. P., Sanderson, D. J., Myers, B. S., & Inglis, J. T. (2003a). Awareness affects the response of human subjects exposed to a single whiplash-like perturbation. Spine, 28(7), 671–679. Siegmund, G. P., Sanderson, D. J., Myers, B. S., & Inglis, J. T. (2003b). Rapid neck muscle adaptation alters the head kinematics of aware and unaware subjects undergoing multiple whiplash-like perturbations. Journal of Biomechanics, 36(4), 473–482. Siegmund, G. P., Vasavada, A. N., & Blouin, J.-S. (2011). The human splenius capitis muscle is primarily a head-neck rotator. In Annual meeting of the American Society of Biomechanics. Long Beach, CA, USA. Siegmund, G. P., Winkelstein, B. A., Ivancic, P. C., Svensson, M. Y., & Vasavada, A. N. (2009). The anatomy and biomechanics of acute and chronic whiplash injury. Traffic Injury Prevention, 10(2), 101–112. Staude, G. H. (2001). Precise onset detection of human motor responses using a whitening filter and the log-likelihood-ratio test. IEEE Transactions on Biomedical Engineering, 48(11), 1292–1305. Staude, G. H., & Wolf, W. (1999). Objective motor response onset detection in surface myoelectric signals. Medical Engineering & Physics, 21(6–7), 449–467. Stemper, B. D., Yoganandan, N., Cusick, J. F., & Pintar, F. A. (2006). Stabilizing Effect of Precontracted 140 Neck Musculature in Whiplash. Spine, 31(20), E733–E738. Stemper, B. D., Yoganandan, N., & Pintar, F. A. (2005). Effects of abnormal posture on capsular ligament elongations in a computational model subjected to whiplash loading. Journal of Biomechanics, 38(6), 1313–1323. Sturzenegger, M., Radanov, B. P., & Di Stefano, G. (1995). The effect of accident mechanisms and initial findings on the long-term course of whiplash injury. Journal of Neurology, 242(7), 443–449. Styrke, J., Stålnacke, B.-M., Bylund, P.-O., Sojka, P., & Björnstig, U. (2012). A 10-year incidence of acute whiplash injuries after road traffic crashes in a defined population in northern Sweden. PM & R : The Journal of Injury, Function, and Rehabilitation, 4(10), 739–747. Suderman, B. L., Krishnamoorthy, B., & Vasavada, A. N. (2012). Neck muscle paths and moment arms are significantly affected by wrapping surface parameters. Computer Methods in Biomechanics and Biomedical Engineering, 15(7), 735–744. https://doi.org/10.1080/10255842.2011.558085 Svensson, M. Y., Boström, O., Davidsson, J., Hansson, H. A., Håland, Y., Lövsund, P., … Säljö, A. (2000). Neck injuries in car collisions--a review covering a possible injury mechanism and the development of a new rear-impact dummy. Accident; Analysis and Prevention, 32(2), 167–175. Szabo, T. J., & Welcher, J. B. (1996). Subject kinematics and electromyographic activity during low speed rear impacts. Stapp Car Crash Journal, 40, 295–315. Todorov, E., & Jordan, M. I. (2002). Optimal feedback control as a theory of motor coordination. Nature Neuroscience, 5(11), 1226–1235. Tominaga, Y., Maak, T. G., Ivancic, P. C., Panjabi, M. M., & Cunningham, B. W. (2006). Head-turned rear impact causing dynamic cervical intervertebral foramen narrowing: implications for ganglion and nerve root injury. Journal of Neurosurgery. Spine, 4(5), 380–387. Tominaga, Y., Ndu, A. B., Coe, M. P., Valenson, A. J., Ivancic, P. C., Ito, S., … Panjabi, M. M. (2006). Neck ligament strength is decreased following whiplash trauma. BMC Musculoskeletal Disorders, 7, 103. Ulbrich, E. J., Añon, J., Hodler, J., Zimmermann, H., Sturzenegger, M., Anderson, S. E., & Boesch, C. (2014). Does normalized signal intensity of cervical discs on T2 weighted MRI images change in 141 whiplash patients? Injury, 45(4), 784–791. Valero-Cuevas, F. J. (2000). Predictive modulation of muscle coordination pattern magnitude scales fingertip force magnitude over the voluntary range. Journal of Neurophysiology, 83(3), 1469–1479. https://doi.org/10.1152/jn.2000.83.3.1469 Valero-Cuevas, F. J., Venkadesan, M., & Todorov, E. (2009). Structured variability of muscle activations supports the minimal intervention principle of motor control. Journal of Neurophysiology, 102(1), 59–68. Valls-Solé, J., Kumru, H., & Kofler, M. (2008). Interaction between startle and voluntary reactions in humans. Experimental Brain Research, 187(4), 497–507. https://doi.org/10.1007/s00221-008-1402-0 van den Kroonenberg, A., Philippens, M., Cappon, H., Wismans, J., Hell, W., & Langwieder, K. (1998). Human Head-Neck Response During Low-Speed Rear End Impacts. https://doi.org/10.4271/983158 van Rooij, L., Elrofai, H., Philippens, M. M. G. M., & Daanen, H. A. M. (2013). Volunteer kinematics and reaction in lateral emergency maneuver tests. Stapp Car Crash Journal, 57(November), 313–342. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/24435737 van Zuylen, E. J., Gielen, C. C., & Denier van der Gon, J. J. (1988). Coordination and inhomogeneous activation of human arm muscles during isometric torques. Journal of Neurophysiology, 60(5), 1523–1548. Vasavada, A. N., Brault, J. R., & Siegmund, G. P. (2007). Musculotendon and fascicle strains in anterior and posterior neck muscles during whiplash injury. Spine, 32(7), 756–765. Vasavada, A. N., Li, S., & Delp, S. L. (1998). Influence of muscle morphometry and moment arms on the moment-generating capacity of human neck muscles. Spine, 23(4), 412–422. Vasavada, A. N., Li, S., & Delp, S. L. (2001). Three-dimensional isometric strength of neck muscles in humans. Spine, 26(17), 1904–1909. Vasavada, A. N., Peterson, B. W., & Delp, S. L. (2002). Three-dimensional spatial tuning of neck muscle activation in humans. Experimental Brain Research, 147(4), 437–448. 142 https://doi.org/10.1007/s00221-002-1275-6 Vernon, H., & Mior, S. (1991). The Neck Disability Index: a study of reliability and validity. Journal of Manipulative and Physiological Therapeutics, 14(7), 409–415. Wilson, V. J., Precht, W., & Dieringer, N. (1983). Responses of different compartments of cat’s splenius muscle to optokinetic stimulation. Experimental Brain Research, 50(1), 153–156. Winkelstein, B. A., McLendon, R. E., Barbir, A., & Myers, B. S. (2001). An anatomical investigation of the human cervical facet capsule, quantifying muscle insertion area. Journal of Anatomy, 198(Pt 4), 455–461. Winkelstein, B. A., Nightingale, R. W., Richardson, W. J., & Myers, B. S. (2000). The cervical facet capsule and its role in whiplash injury: a biomechanical investigation. Spine, 25(10), 1238–1246. Winters, J. M. (1995). How detailed should muscle models be to understand multi-joint movement coordination? Human Movement Science, 14(4–5), 401–442. Winters, J. M., & Stark, L. (1988). Estimated mechanical properties of synergistic muscles involved in movements of a variety of human joints. Journal of Biomechanics, 21(12), 1027–1041. Winters, J. M., & Woo, S. L. Y. (1990). Multiple Muscle Systems: Biomechanics and Movement Organization. New York: Springer-Verlag, New York. Yao, H.-D., Svensson, M. Y., & Nilsson, H. (2018). Deformation of dorsal root ganglion due to pressure transients of venous blood and cerebrospinal fluid in the cervical vertebral canal. Journal of Biomechanics, 76, 16–26. Yeomans, J. S., & Frankland, P. W. (1995). The acoustic startle reflex: neurons and connections. Brain Research. Brain Research Reviews, 21(3), 301–314. Ylinen, J., Salo, P., Nykänen, M., Kautiainen, H., & Häkkinen, A. (2004). Decreased isometric neck strength in women with chronic neck pain and the repeatability of neck strength measurements. Archives of Physical Medicine and Rehabilitation, 85(8), 1303–1308. Yoganandan, N., Pintar, F. A., & Cusick, J. F. (2002). Biomechanical analyses of whiplash injuries using an 143 experimental model. Accident; Analysis and Prevention, 34(5), 663–671. Zhang, K., Cao, L., Fanta, A., Reed, M. P., Neal, M., Wang, J.-T., … Hu, J. (2017). An automated method to morph finite element whole-body human models with a wide range of stature and body shape for both men and women. Journal of Biomechanics, 60, 253–260. 144 Appendix Table A.1 - Chapter 2: Magnitude of the resultant moment produced by each subject for all of the directions tested. R. = right, L. = left, Flex. = flexion, Lat. Bend. = lateral bending, Ext. = extension. Data given in Nm. Axial Moment Direction Horizontal Plane Direction Subject 1 Subject 2 Subject 3 Subject 4 Subject 5 Subject 6 Subject 7 Subject 8 Subject 9 No Axial Flexion 28.3 24.9 37.9 21.8 32.7 40.3 27.6 32.1 25.4 Flex. + R. Lat. Bend. 32.6 23.2 33.9 23.6 26.5 41.6 27.0 28.8 31.5 R. Lat. Bend. 35.1 20.7 35.7 19.8 25.7 47.2 30.2 34.5 42.5 Ext. + R. Lat. Bend 38.3 21.9 46.2 38.1 28.0 48.7 29.8 40.3 55.2 Extension 42.3 35.8 45.7 45.8 48.6 67.1 45.4 57.7 66.0 R. Axial Flexion 22.8 21.1 25.9 15.0 18.3 32.7 23.2 29.0 20.1 Flex. + R. Lat. Bend. 33.2 25.5 25.6 18.7 19.7 40.8 30.2 34.4 30.8 R. Lat. Bend. 45.8 30.8 29.2 28.5 28.7 54.2 25.9 28.5 48.5 Ext. + R. Lat. Bend 43.8 31.4 41.7 34.8 31.5 59.7 33.2 39.7 53.9 Extension 35.1 23.5 39.4 25.3 17.7 51.3 27.6 44.4 41.7 L. Axial Flexion 23.9 19.6 28.1 17.7 23.8 31.9 19.1 29.7 21.6 Flex. + R. Lat. Bend. 24.8 20.5 31.3 17.8 15.4 32.7 23.1 31.4 31.5 R. Lat. Bend. 24.0 18.0 36.2 14.9 19.2 31.2 18.8 30.3 35.1 Ext. + R. Lat. Bend 25.6 16.7 34.0 20.2 18.1 35.5 23.4 39.5 37.9 Extension 35.2 24.4 27.7 20.3 21.5 47.5 31.9 43.8 46.5 R. Axial None 17.2 8.1 10.9 7.9 6.9 19.4 11.3 14.8 17.9 L. Axial None 15.9 8.0 10.6 8.3 7.0 18.9 12.9 13.9 18.9 145 Table A.2 - Chapter 2: Normalized 3D moments produced by each subject for all of the directions tested. R. = right, L. = left, Flex. = flexion, Lat. Bend. = lateral bending, Ext. = extension. Axial Moment Direction Horizontal Plane Direction Subject 1 Subject 2 Subject 3 Subject 4 Subject 5 Subject 6 Subject 7 Subject 8 Subject 9 No Axial Flexion 1.00 1.01 1.01 1.02 1.04 1.00 1.00 1.00 1.00 Flex. + R. Lat. Bend. 1.05 1.09 0.93 1.15 0.96 0.95 0.96 0.88 0.97 R. Lat. Bend. 1.01 1.03 1.01 1.00 1.01 1.01 1.01 1.00 1.01 Ext. + R. Lat. Bend 1.03 0.92 1.16 1.37 0.82 0.91 0.79 0.97 1.06 Extension 1.00 1.01 1.00 1.01 1.01 1.00 1.00 1.01 1.00 R. Axial Flexion 0.86 1.08 0.93 0.79 1.02 0.90 1.01 1.04 0.80 Flex. + R. Lat. Bend. 1.16 1.48 0.93 1.05 1.06 1.08 1.30 1.26 1.02 R. Lat. Bend. 1.53 2.03 1.17 1.69 1.93 1.39 1.13 1.00 1.36 Ext. + R. Lat. Bend 1.36 1.85 1.49 1.81 1.74 1.40 1.26 1.17 1.32 Extension 0.98 1.13 1.34 0.99 0.76 1.03 0.84 1.12 0.86 L. Axial Flexion 0.90 1.03 1.06 0.95 1.32 0.93 0.82 1.04 0.88 Flex. + R. Lat. Bend. 0.86 1.11 1.17 0.99 0.87 0.85 0.94 1.09 1.06 R. Lat. Bend. 0.76 1.01 1.37 0.83 1.05 0.76 0.72 1.01 0.94 Ext. + R. Lat. Bend 0.77 0.74 1.23 0.95 0.94 0.79 0.77 1.16 0.92 Extension 0.99 1.05 0.91 0.74 0.97 0.95 0.98 1.15 0.96 R. Axial None 1.04 0.96 1.03 0.95 0.90 1.02 0.94 1.03 0.98 L. Axial None 0.97 1.08 0.99 1.08 1.13 0.99 1.09 0.98 1.04 146 Figure A.1 - Chapter 5 X-axis mean kinematic subject data shown with ± standard deviation response corridors for neutral, left shoulder check, left mirror check, rear-view mirror check, and looking at passenger (right side) experimental conditions. a, linear acceleration; Δd, change in distance; α, angular acceleration; ω, angular velocity; Δθ, change in angular position relative to initial position; CG, center of gravity; C0C1, atlanto-occipital joint; T1, first thoracic vertebrae. 147 Figure A.2 - Chapter 5 Y-axis mean kinematic subject data shown with ± standard deviation response corridors for neutral, left shoulder check, left mirror check, rear-view mirror check, and looking at passenger (right side) experimental conditions. a, linear acceleration; Δd, change in distance; α, angular acceleration; ω, angular velocity; Δθ, change in angular position relative to initial position; CG, center of gravity; C0C1, atlanto-occipital joint; T1, first thoracic vertebrae. 148 Figure A.3 - Chapter 5 Z-axis mean kinematic subject data shown with ± standard deviation response corridors for neutral, left shoulder check, left mirror check, rear-view mirror check, and looking at passenger (right side) experimental conditions. a, linear acceleration; Δd, change in distance; α, angular acceleration; ω, angular velocity; Δθ, change in angular position relative to initial position; CG, center of gravity; C0C1, atlanto-occipital joint; T1, first thoracic vertebrae. 149 Figure A.4 - Chapter 6 rear impact kinematic subject data shown with response corridors for control, relaxed and braced experimental conditions. a, linear acceleration; ω, angular velocity; θ, angular position; x, forward direction; y, rightward direction; z, downward direction; CG, center of gravity; T1, first thoracic vertebrae. 150 Figure A.5 - Chapter 6 frontal impact kinematic subject data shown with response corridors for hands in lap, relaxed and braced experimental conditions. a, linear acceleration; ω, angular velocity; θ, angular position; x, forward direction; y, rightward direction; z, downward direction; CG, center of gravity; T1, first thoracic vertebrae.