Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Regional activation within vastus medialis and lateralis in clinical and experimental pain Gallina, Alessio 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2018_may_gallina_alessio.pdf [ 4.38MB ]
Metadata
JSON: 24-1.0364687.json
JSON-LD: 24-1.0364687-ld.json
RDF/XML (Pretty): 24-1.0364687-rdf.xml
RDF/JSON: 24-1.0364687-rdf.json
Turtle: 24-1.0364687-turtle.txt
N-Triples: 24-1.0364687-rdf-ntriples.txt
Original Record: 24-1.0364687-source.json
Full Text
24-1.0364687-fulltext.txt
Citation
24-1.0364687.ris

Full Text

REGIONAL ACTIVATION WITHIN VASTUS MEDIALIS AND LATERALIS IN CLINICAL AND EXPERIMENTAL PAIN by  Alessio Gallina  B.Sc.P.T., Universitá degli Studi di Torino, 2009 M.Sc., Universitá degli Studi di Pisa, 2012  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Rehabilitation Sciences)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  March 2018  © Alessio Gallina, 2018 ii  Abstract The possibility to preferentially activate the distal region of the vastus medialis (VM) has been debated for a long time in physiotherapy for its potential relevance for patellofemoral pain (PFP). However, there is little experimental evidence that quadriceps muscle activation can be modulated regionally, either in voluntary tasks or in the presence of pain. This thesis examined whether the human neuromuscular system can preferentially control regions within the quadriceps muscle, and if pain results in regional modulation of quadriceps muscle activation. Methods: Regional activation within the VM (Chapters 2 to 5) and vastus lateralis (VL; Chapters 4 and 5) was investigated using high-density surface electromyography. To investigate whether the human spinal cord has the neuromuscular circuitry to selectively recruit motor units located in different muscle regions, regional stretch reflexes were elicited by applying mechanical stimuli along the VM insertion on the patella (Chapter 2). The possibility to regionally activate the VM in dynamic voluntary contractions was studied (Chapter 3). Principal component analysis was used to identify differences in regional activation patterns between 36 females with PFP and 20 age- and sex-matched controls (Chapter 4). Finally, the effect of location-specific acute nociceptive input on muscle activation within VM and VL was studied (Chapter 5). Results: Regionalized stretch reflexes provided evidence that the human spinal cord has the neuromuscular circuitry to selectively modulate the activation of regions within the VM. Redistribution between VM and VL and within the VM was observed in controls when comparing concentric and eccentric contractions. In the same task, less redistribution was observed in females with PFP, especially those with higher maximal knee extension strength. iii  Region-specific adaptations of VM activation were observed in response to localized nociceptive input. Discussion: This dissertation provides evidence for regional activation between quadriceps heads and within the VM in reflex and voluntary contractions. Regional modulation of quadriceps muscle activation was observed in PFP and with experimental pain. This thesis highlights the role of regional activation within the quadriceps muscle in health and pain, and provides support for current theories on how the human nervous system adapts to pain. iv  Lay Summary Pain changes the way we move by altering how muscles work. It is currently unknown whether pain around the knee changes the activation of the quadriceps (the muscle in front of the thigh) as a whole, or if regions within this muscle adapt to pain differently. This dissertation demonstrates that regions within the human quadriceps muscle can be activated preferentially, both in voluntary and reflex contractions. Females with pain around the kneecap showed less regional activation of quadriceps muscle than females without pain. When pain was experimentally induced in different regions around the knee by injecting a mixture of water and salt, the changes in regional activation of the quadriceps muscle depended on the pain location. This dissertation furthers our understanding on how the human body controls muscles in the presence of pain. v  Preface The work in this dissertation was conceived, conducted, and written by Alessio Gallina. Research described in this dissertation was approved by the University of British Columbia's (UBC) Clinical Research Ethics Board: H14-02035, H14-01088 and H15-01263; and by the University of Queensland Human Research Ethics committee: 2004000654. Chapters 1 and 6 were written by Alessio Gallina. Drs. Jayne Garland, Michael Hunt, James Wakeling and Paul Hodges assisted in editing these chapters.  Chapter 2 is based on work conducted by Alessio Gallina, and Drs. Jayne Garland, Jean-Sebastien Blouin, and Tanya Ivanova. Alessio Gallina was responsible for the study design, data collection, analyses and interpretation, and writing and revising the manuscript. Drs. Garland, Bloiun, and Ivanova assisted in designing the study, data collection, analysis, interpretation and editing the manuscript. A version of Chapter 2 has been published: Gallina A, Blouin JS, Ivanova TI, Garland JS. Regionalization of the stretch reflex in the human vastus medialis. J Physiol, 2017; 595: 4991–5001. Chapter 3 is based on work conducted by Alessio Gallina, and Drs. Jayne Garland, and Tanya Ivanova. Alessio Gallina was responsible for the study design, data collection, analyses and interpretation, and writing and revising the manuscript. Drs. Garland and Ivanova assisted in designing the study, data collection, analysis, interpretation and editing the manuscript. A version of Chapter 3 has been published: Gallina A, Ivanova TD, Garland SJ. Regional activation within the vastus medialis in stimulated and voluntary contractions. J Appl Physiol 2016; 121, 466-77. Chapter 4 is based on work conducted by Alessio Gallina, and Drs. Jayne Garland, Michael Hunt, James Wakeling and Paul Hodges. Alessio Gallina was responsible for the study vi  design, data collection, analyses and interpretation, and writing and revising the manuscript. Drs. Garland, Hunt, Wakeling, and Hodges assisted in designing the study, data collection, analysis, interpretation and editing the manuscript. Chapter 5 is based on work conducted by Alessio Gallina, and Drs. Jayne Garland and Paul Hodges. Alessio Gallina was responsible for the study design, data collection, analyses and interpretation, and writing and revising the manuscript. Drs. Garland, Hodges and Tucker assisted in designing the study, data collection, analysis, interpretation and editing the manuscript. Drs. Salomoni and Hall helped with data collection. A version of Chapter 5 has been submitted: Gallina A, Salomoni SE, Hall L, Tucker K, Garland SJ, Hodges PW. Location specific responses to nociceptive input support the purposeful nature of pain adaptation. vii  Table of Contents Abstract .......................................................................................................................................... ii Lay Summary ............................................................................................................................... iv Preface .............................................................................................................................................v Table of Contents ........................................................................................................................ vii List of Figures ............................................................................................................................. xiii List of Abbreviations ...................................................................................................................xv Acknowledgements .................................................................................................................... xvi Dedication .................................................................................................................................. xvii Chapter 1: Introduction ................................................................................................................1 1.1 General aim ..................................................................................................................... 1 1.2 Pain adaptation theory..................................................................................................... 1 1.3 Quadriceps muscle and patellofemoral joint structure and function .............................. 3 1.4 Motor unit anatomy and control ..................................................................................... 5 1.5 Spinal stretch reflex and neuromuscular control ............................................................ 7 1.6 Pain and neuromuscular control ...................................................................................... 8 1.7 Patellofemoral pain ....................................................................................................... 10 1.7.1 Prevalence ................................................................................................................. 10 1.7.2 Clinical management ................................................................................................ 10 1.7.3 Etiopathogenesis of patellofemoral pain ................................................................... 11 1.7.4 Neuromuscular dysfunction associated with patellofemoral pain ............................ 12 1.8 Methodological approach.............................................................................................. 14 1.8.1 Electromyography ..................................................................................................... 14 viii  1.8.2 Surface electromyography ........................................................................................ 15 1.8.3 High-density surface electromyography ................................................................... 16 1.8.4 Factorization algorithms and high-density surface electromyography ..................... 18 1.9 Thesis outline ................................................................................................................ 19 1.10 Objectives and hypotheses ............................................................................................ 20 Chapter 2: Regionalization of the stretch reflex in the human vastus medialis.....................22 2.1 Introduction ................................................................................................................... 22 2.2 Methods......................................................................................................................... 24 2.2.1 Participants ................................................................................................................ 24 2.2.2 High-density surface electromyography ................................................................... 24 2.2.3 Intramuscular recordings .......................................................................................... 26 2.2.4 Protocol ..................................................................................................................... 27 2.2.5 Data analysis ............................................................................................................. 28 2.2.6 Statistical analysis ..................................................................................................... 36 2.3 Results ........................................................................................................................... 37 2.4 Discussion ..................................................................................................................... 41 Chapter 3: Regional activation within the vastus medialis in stimulated and voluntary contractions ..................................................................................................................................46 3.1 Introduction ................................................................................................................... 46 3.2 Methods......................................................................................................................... 48 3.2.1 Participants ................................................................................................................ 48 3.2.2 Experimental setup.................................................................................................... 48 3.2.3 Anatomical reference ................................................................................................ 50 ix  3.2.4 Intramuscular stimulation protocol ........................................................................... 52 3.2.5 Dynamic contractions protocol ................................................................................. 53 3.2.6 Data analysis ............................................................................................................. 53 3.2.7 Statistical analysis ..................................................................................................... 57 3.3 Results ........................................................................................................................... 58 3.3.1 Participants ................................................................................................................ 58 3.3.2 Intramuscular stimulation protocol ........................................................................... 58 3.3.3 Changes in M-wave amplitude with distance ........................................................... 59 3.3.4 Intramuscular stimulation ......................................................................................... 60 3.3.5 Dynamic contractions ............................................................................................... 62 3.3.6 Isometric contractions ............................................................................................... 65 3.4 Discussion ..................................................................................................................... 65 Chapter 4: Regional activation within vastus medialis and lateralis in a dynamic task is altered in patellofemoral pain .....................................................................................................71 4.1 Introduction ................................................................................................................... 71 4.2 Methods......................................................................................................................... 73 4.2.1 Participants ................................................................................................................ 73 4.2.2 Clinical tests .............................................................................................................. 74 4.2.3 Protocol ..................................................................................................................... 74 4.2.4 Data collection .......................................................................................................... 74 4.2.5 Data analysis ............................................................................................................. 76 4.2.6 Statistical analysis ..................................................................................................... 80 4.3 Results ........................................................................................................................... 81 x  4.4 Discussion ..................................................................................................................... 88 Chapter 5: Location specific responses to nociceptive input support the purposeful nature of pain adaptation ........................................................................................................................92 5.1 Introduction ................................................................................................................... 92 5.2 Methods......................................................................................................................... 94 5.2.1 Participants ................................................................................................................ 94 5.2.2 Experimental protocol ............................................................................................... 94 5.2.3 Electromyography and force recordings ................................................................... 95 5.2.4 Experimental knee pain............................................................................................. 97 5.2.5 Voluntary contractions .............................................................................................. 98 5.2.6 Data analysis ............................................................................................................. 99 5.2.7 Statistical analysis ................................................................................................... 100 5.3 Results ......................................................................................................................... 101 5.4 Discussion ................................................................................................................... 107 Chapter 6: General discussion ..................................................................................................113 6.1 Summary ..................................................................................................................... 113 6.2 Methodological advancements.................................................................................... 113 6.2.1 Combining intramuscular stimulation and high-density surface electromyography to infer muscle architecture ..................................................................................................... 113 6.2.2 Influence of crosstalk .............................................................................................. 114 6.2.3 Interpretation of surface electromyographic amplitude in dynamic contractions .. 115 6.2.4 Stretch reflexes elicited by mechanical taps applied to the muscle ........................ 115 6.2.5 Twitching and cramping in response to hypertonic saline solution injection ......... 116 xi  6.3 Anatomical advancements .......................................................................................... 117 6.3.1 Localized muscle fibre distribution in vastus medialis motor units ....................... 117 6.3.2 Localized spinal circuitry in vastus medialis .......................................................... 118 6.4 Neurophysiological advancements ............................................................................. 120 6.4.1 Relevance of localized stretch reflexes ................................................................... 120 6.4.2 Neuromuscular adaptation to acute pain is location-specific ................................. 121 6.4.3 Vasti neuromuscular activation in females with and without patellofemoral pain . 123 6.5 Clinical advancements ................................................................................................ 124 6.5.1 Exercise prescription to target regions within the vasti .......................................... 124 6.5.2 Link between pain location and presentations of musculoskeletal disorders ......... 125 6.5.3 Association between knee extension strength and neuromuscular dysfunction ..... 125 6.6 Limitations .................................................................................................................. 126 6.7 Implications and future directions .............................................................................. 128 6.7.1 Representativeness of surface electromyographic recordings ................................ 128 6.7.2 Regionalization of reflexes ..................................................................................... 129 6.7.3 Implications of location-specific adaptation to pain ............................................... 129 6.7.4 Unbalanced vasti activation in patellofemoral pain: clinical implications ............. 130 6.8 Conclusions ................................................................................................................. 131 References ...................................................................................................................................132 xii  List of Tables Table 2.1: Intramuscular EMG activation in response to mechanical taps................................... 40 Table 4.1: Anthropometric and clinical measures, differences between groups. ......................... 82 Table 4.2: Association between neuromuscular activation indicators and clinical measures.. .... 88  xiii  List of Figures Figure 1.1: Pain Adaptation Model ................................................................................................. 2 Figure 2.1: Experimental setup ..................................................................................................... 25 Figure 2.2: Identification of a surface EMG response .................................................................. 29 Figure 2.3: Example of responses to mechanical taps for one participants .................................. 31 Figure 2.4: Responses to individual mechanical taps ................................................................... 33 Figure 2.5: Identification of size and location of EMG responses across the grid ....................... 34 Figure 2.6: Muscle twitches in response to mechanical taps ........................................................ 37 Figure 2.7: Effect of tap location on the localization of the response .......................................... 39 Figure 3.1: Experimental setup ..................................................................................................... 51 Figure 3.2: Example of M-waves elicited in a representative participant .................................... 54 Figure 3.3: Example of action potentials along columns and rows of the grid ............................ 55 Figure 3.4: EMG amplitude distributions at different knee angles in a participant ...................... 60 Figure 3.5: Shift of EMG amplitude distribution along columns and rows of the grid ................ 61 Figure 3.6: EMG amplitude distributions in the dynamic contraction ......................................... 63 Figure 3.7: Peak EMG amplitude distribution along the columns of the electrode grid .............. 64 Figure 4.1: Experimental setup ..................................................................................................... 75 Figure 4.2: Example of PCA analysis of HDsEMG signals for a control participant .................. 78 Figure 4.3: Example of PCA analysis of HDsEMG signals for a participant with PFP ............... 79 Figure 4.4: Minimum number of PCs that explains at least 90% of the variance ........................ 83 Figure 4.5: Comparison of temporal coefficients of PC1 and PC2 .............................................. 86 Figure 4.6: Scatter plot of KES and VM/VL redistribution index ................................................ 87 Figure 5.1: Experimental setup ..................................................................................................... 95 xiv  Figure 5.2: Position of the electrode grids and pain areas for the four locations ......................... 96 Figure 5.3: Experimental protocol ................................................................................................ 98 Figure 5.4: Involuntary muscle activation recorded during pain at rest ..................................... 102 Figure 5.5: Effect of experimental pain on knee extension force direction. ............................... 103 Figure 5.6: Effect of experimental pain on EMG amplitude ...................................................... 105 Figure 5.7: Regional activation in response to experimental pain .............................................. 106 Figure 5.8: Changes in surface EMG amplitude distribution ..................................................... 107  xv  List of Abbreviations ARV, Average rectified value BAS, Baseline BMI, Body mass index CNS, Central nervous system CoV, Coefficient of variation EMG, Electromyographic FP, Infrapatellar fat pad HDsEMG, High-density surface electromyography KES, Knee extension strength MVC, maximal voluntary contraction MVEA, Maximal voluntary electrical activation nKES, Normalized knee extension strength NCS, Normalized change scores PCA, Principal component analysis PCn, Principal component [number] PF, Patellofemoral PFP, Patellofemoral pain RF, Rectus femoris VL, Vastus lateralis VM, Vastus medialis VMD, Vastus medialis, distal VMP, Vastus medialis, proximal xvi  Acknowledgements I will start by saying that I am extremely grateful to my supervisor, Dr. Jayne Garland, for guiding me throughout my PhD. Her advice was essential to make the best of this experience, substantially contributing to my growth as a researcher and as a person. My gratitude also extends to my supervisory committee, Drs. Paul Hodges, Michael Hunt and James Wakeling. Paul, your work inspired me since I started doing research. I have deeply enjoyed our discussions on pain adaptation and neuromuscular control; our in-person meetings have been an incredible drive for my research. Michael and James, thank you for your mentorship in research areas I was not familiar with, expanding my interests and making me a more complete researcher. I wish to acknowledge the Vanier Canada Graduate Scholarship program; receiving this award meant a lot to me. Besides the needed financial support, knowing that reviewers from government funding agencies believed in me as a researcher and in my work was a big motivational push during my PhD. I also wish to thank important collaborators. I wish to thank Drs. Tanya Ivanova, Sue Peters and Kim Miller for their support during my PhD; I feel like we have done much learning together, and my work benefited a lot from this collaboration. My sincerest gratitude also to Drs. Jean-Sébastien Blouin and Lara Boyd for their availability to help me in my research projects; working with them was a truly enriching experience. My PhD experience would not have been the same without my fellow graduate students, postdocs and staff. I feel blessed to be able to call many of them friends today. Thank you all for the time together, in the lab and outside, and the constructive discussions about science - and life in general! Special thanks are owed to my family. To my wife Alice, who supported my career choices by embracing a life together overseas. To my parents Marco and Luciana, and to my sister Arianna, who always believed in me, and greatly contributed to make me who I am today. And to all my Italian friends, who continue to greatly influence my life even from a distance. My PhD would not have been possible without all these people. xvii  Dedication To Alice, my life and adventure partner.  1  Chapter 1: Introduction 1.1 General aim The objective of this thesis was to determine whether pain alters the activation of regions within the quadriceps muscle. Due to their unique architectural features and putative relevance in pathologies such as patellofemoral pain (PFP), vastus medialis (VM) and vastus lateralis (VL) were used as a model to investigate how regions within a muscle adapt to experimental and clinical pain. 1.2 Pain adaptation theory Pain is a common symptom of pathologies treated by physiotherapists and other medical specialists. Sensorimotor adaptations to pain have been studied for years, and different theories have been proposed to explain how neuromuscular activation changes in the presence of pain. One attempt to describe the adaptation to pain was the “vicious cycle theory” (Travell et al., 1942; Roland, 1986), which predicted that the agonist muscle would be tonically active in response to muscle pain, leading to ischemia and further pain. Later, a different theory predicted opposite changes, with inhibition of the agonist for painful movements and facilitation of the antagonist muscle (Lund et al., 1991). The most recent theory (Hodges & Tucker, 2011; Hodges & Smeets, 2014) conceived a purposeful adaptation of the motor system to pain with the final goal of protecting the irritated tissues and avoid further injury (Figure 1.1). This is achieved by modifying motor output and mechanical behaviour (e.g.: by redistributing the load within a joint) through changes in neuromuscular activation both between- and within-muscle, coordinated at different levels of the nervous system. While these changes are beneficial in the short term, they may sustain pain and cause further injury in the long term. Overall, this theory offers a direct link between the treatment of specific neuromuscular dysfunction and clinical outcomes of 2  rehabilitative interventions. Of relevance, physiotherapy has the means to intervene at different stages of this cycle (Hodges, 2011). This theoretical model may be useful in clinical practice to set treatment goals relevant for each stage of the adaptation, such as: 1) obtain early pain reduction to reduce impact on motor control and learning; 2) identify which techniques can be effective to influence motor adaptation at different levels of the nervous system; 3) identify and train redistribution of muscle activation and of load within the joints; 4) enhance the components of the adaptation that are helpful, while reducing unnecessary components that may result in long term consequences.    Figure 1.1: Theoretical model (Hodges, 2011) of how the human body adapts to pain, and how rehabilitation can influence the different stages of the adaptation. (© 2011 Elsevier, by permission).  3  1.3 Quadriceps muscle and patellofemoral joint structure and function According to the current pain adaptation theory, redistribution of load within a joint in response to pain may be achieved by altering the relative activation between muscles and between regions of a muscle. Due to its potential for force production along different directions and its common involvement in musculoskeletal disorders, the human quadriceps muscle offers a model to study muscle adaptation to clinical and experimental knee pain.  The quadriceps muscle comprises four heads. The VM, VL and vastus intermedius originate from the proximal region of the femur and insert on the patella through the common quadricipital tendon, although the VM also has a direct insertion on the medial side of the patella (Holt et al., 2008). The rectus femoris (RF) is the only head of the quadriceps muscle that crosses the hip joint, originating from the anterior inferior iliac spine and inserting on the quadricipital tendon. The main function of the quadriceps muscle as a whole is knee extension, and the RF also creates a hip flexion vector. However, due to specific anatomical characteristics, each vastus creates force vectors in different directions around the patellofemoral (PF) joint. For instance, VM and VL are synergists for knee extension, but their fibre orientation and insertions on the patellar tendon adds off-axis vectors (Wilson & Sheehan, 2010), meaning that regions within VM and VL may produce forces different from pure knee extension.  Movements of the PF joint are described as translations and rotations in three directions (antero-posterior, proximal-distal, medio-lateral); some of these motions, especially mediolateral translation and rotations, influence PF contact mechanics (i.e., where and how much of the patella is in contact with the femur). Different contributions of VM and VL result in different patellar movement (Lin et al., 2004), distribution of pressure within the PF joint (Elias et al., 4  2009; Wünschel et al., 2011; Lorenz et al., 2012) and tibiofemoral joint kinematics (Wünschel et al., 2011).  The vasti are complex muscles, with differences in fibre orientation across their volume (Weinstabl et al., 1989; Peeler et al., 2005; Blazevich et al., 2006; also reviewed in: Becker et al., 2009, 2010; Smith et al., 2009b) and flat, large insertions (Weinstabl et al., 1989; Peeler et al., 2005). Indeed, several studies demonstrated differences in muscle thickness (Blazevich et al., 2006; Gallina et al., 2018a), fibre pennation (angle on a plane perpendicular to the skin; Blazevich et al., 2006; Gallina et al., 2018a) and fibre orientation (angle on a plane parallel to the skin; Peeler et al., 2005; Gallina et al., 2018a) when proximal and distal VM were compared. Local variations in muscle architecture were proven to result in specific force vectors and lines of action for proximal and distal regions within VM (Lin et al., 2004; Wilson & Sheehan, 2010; Gallina & Vieira, 2015) and VL (Wilson & Sheehan, 2010). This leads to different functional roles of the two muscle regions: the proximal, less oblique fibres substantially contribute to pull the patella proximally (with knee extension), the distal fibres have a predominant action on mediolateral patellar kinematics (Lin et al., 2004; Lefebvre et al., 2007). Blocking the activation of the distal region of the VM with anaesthetics, for instance, resulted in increased lateral patellar shift, lateral tibial shift and rotation during dynamic knee extensions (Sheehan et al., 2012).  The different force vectors created by individual quadriceps muscle heads may be used to redistribute load within the PF joint in response to pain. To be able to take full advantage of such heterogeneity for movement production, the central nervous system (CNS) should be able to coordinate not only individual muscles, but also specific regions within those muscles. 5  1.4 Motor unit anatomy and control The smallest unit that can be controlled by the human CNS is the motor unit, which consists of a motoneurone and all the muscle fibres it innervates. Motor units can be considered the actuators of the CNS, as they provide the conduit to convert neural drive from the CNS into forces and movement (Heckman & Enoka, 2012). Increased muscle force production as a result of increased neural drive is achieved using two mechanisms: 1) recruitment of additional motor units; 2) increase in discharge rate of active motor units. The main determinant of motor unit recruitment order is the size of its soma, with smaller motoneurones being activated before larger ones (Henneman, 1957). This principle, commonly referred to as Henneman’s size principle, has proven to be valid in most of the conditions tested, with few exceptions (reviewed in Heckman & Enoka, 2012). In addition, motor unit discharge rate tends to vary in concert both within a muscle and between synergist muscles (De Luca & Erim, 2002; Laine et al., 2015), as active motor units share a common drive from the CNS (De Luca & Erim, 1994). Taken together, these two principles seem to suggest that there may be little possibility to preferentially drive motor units based on their mechanical action. To be able to preferentially control muscle regions, two conditions must be met. First, most of the fibres innervated by a single motoneurone must be confined in a small volume within the muscle. The existence of motor units with localized territories has been documented in most (Buchtal et al., 1959; Gootzen et al., 1992; Vieira et al., 2011; Gallina & Vieira, 2015), approximately 70% (Harris et al., 2005) or up to half (Héroux et al., 2015) of the motor units in human muscles. The variability observed across studies (Vieira et al., 2011; Héroux et al., 2015) may be due to differences in the technique used and the definition of motor unit territory. It should be noted that even motor units with large territories (i.e.: fibres spread across the whole 6  muscle volume) can be considered to have localized territories if most of their fibres are clustered in a muscle region and only few fibres can be observed in the remaining muscle volume (Vieira et al., 2016). With regard to the muscles of interest for this thesis, localized motor unit territories were observed for the VM both using intramuscular (Gootzen et al., 1992) and surface (Gallina & Vieira, 2015) electromyographic (EMG) techniques. In addition, in a previous study (Gallina & Vieira, 2015) it was shown that the fibre orientation of VM motor units depended on their location within the VM, factor contributing to the different mechanical function of proximal vs. distal regions (Lin et al., 2004). The second prerequisite for regional control is that the spinal cord must possess the neural circuitry to preferentially recruit motor units located in different muscle regions. For multifunctional muscles such as the first dorsal interosseous, there is evidence that motor units can be recruited in a different order depending on the direction force is applied (i.e.: flexion vs. abduction; Desmedt & Godaux, 1981). Similar results were reported for biceps brachii and deltoid muscles (Herrmann & Flanders, 1998). Although this could be considered an exception to the Henneman’s size principle (Heckman & Enoka, 2012), if muscles are considered as comprising sub-modules that can be controlled preferentially, motor units can still be recruited according to their motoneurone size within those sub-modules (Riek & Bawa, 1992). The possibility to independently recruit motor units of different pools within a muscle led some authors to suggest that the CNS controls the motoneuronal output according to mechanical advantage, possibly in addition to the size principle (Herrmann & Flanders, 1998; Butler & Gandevia, 2008). However, these studies provide only partial evidence for motor unit recruitment based on their mechanical action for a number of reasons: i) only muscles that can be voluntarily activated to produce force in different directions could be tested; ii) due to the 7  redundancy of motor solutions available to perform a movement, changing the direction of force exerted is an indirect way to manipulate the mechanical demand of muscle regions; iii) the intramuscular recordings used in most studies provide only partial information on the position of the motor unit within the muscle and no information on its mechanical action (e.g.: its insertion); iv) only muscles where compartments can be described anatomically were tested. Overall, the evidence that the spinal cord has the neural circuitry to preferentially recruit motor units located in different muscle regions is not convincing. 1.5 Spinal stretch reflex and neuromuscular control The possibility that regional muscle activation is achieved by the regionalization of the spinal stretch reflex has been discussed in the scientific community for years (Windhorst et al., 1989), although no direct evidence of this phenomenon has been reported in humans. Previous research showed that the stretch reflex is organized regionally in animals. When sudden stretch was applied to single fibres of the cat RF (Cohen, 1953), only the fibres where the stimulus was applied (and not the whole muscle) produced a twitch response. Similarly, Eng and Hoffer (1997) measured medial gastrocnemius EMG activity and fibres stretch velocity in two locations (proximally and distally) in cats during postural perturbations. The EMG reflex response measured in one location within the muscle was strongly associated with the stretch of the fibres in the same, but not in the other location. In the same study, neither the local fibre shortening, nor the EMG responses, were associated with the velocity of shortening of the whole muscle. Finally, proprioceptive reflexes were shown to be localized in the cat splenius muscle, as differences existed in how the EMG activity was modulated in different compartments (Bilotto et al., 1982). These experiments show that, in mammalian models, the stretch reflex pathways are organized locally. These observations in animals led to the formulation of the “partitioning 8  hypothesis”, in which the muscle activation is modulated regionally based on local feedback from muscle spindles (Stuart et al., 1988; Windhorst et al., 1989).  However, the localization of stretch reflexes has never been proven in humans. While taps applied to proximal or distal regions of the human tibialis anterior were shown to preferentially activate 1a afferents close to the location of the tap, an increased probability of firing was observed similarly for motor units located close to and far from the location of the tap (McKeon et al., 1984). However, it should be noted that the distal insertion of the tibialis anterior is a single long tendon, hence the muscle has a low potential for region-specific mechanical action. As the amount of common drive between motor units in a muscle was suggested to be inversely correlated with the density of its spindles (De Luca et al., 2009), muscles with larger afferent feedback may have more independent motor unit activation strategies. For this reason, investigation of the localization of the stretch reflex in muscles with a high potential for region-specific mechanical actions, such as the VM, warrants further research.  1.6 Pain and neuromuscular control The effect of pain on neuromuscular activation can be studied using experimental models. Although experimental pain does not replicate all aspects of clinical pain, it enables the characterization of the motor adaptation without the confounding factors found in natural pain conditions such as presence of injury and inflammation, differences in muscle composition, and variability in the source of pain. When investigating motor adaptation, a transient episode of acute pain can be experimentally elicited by injecting hypertonic saline solution in muscles. This model has been used in a variety of muscles such as trapezius (Madeleine et al., 2006; Falla et al., 2009; Dideriksen et al., 2016), extensor carpi radialis (Birch et al., 2000), vasti (Tucker et al., 2014a), gastrocnemii (Hug et al., 2013) and back muscles (Tucker et al., 2014b). 9  Experimental pain can also be elicited by injecting non-muscular tissues such as the medial infrapatellar fat pad (FP) (Bennell et al., 2004; Tucker & Hodges, 2009; Tucker et al., 2009; Hodges et al., 2009) and the interspinous ligament in the back (Tsao et al., 2011b). Interestingly, pain location and qualitative descriptors reported after infrapatellar pain injection are similar to those commonly found in patellofemoral pain (PFP) (Bennell et al., 2004). Although changes in neuromuscular control during experimental pain are, generally, variable across studies and across participants (Hodges et al., 2013; Gizzi et al., 2015), intramuscular recordings usually show a predominant reduction in motor unit discharge rate during pain (Farina et al., 2002a; Hodges et al., 2008; Tucker & Hodges, 2009; Dideriksen et al., 2016). However, this trend is not homogeneous, as the discharge rate of some motor units increases instead (Tucker & Hodges, 2009). Recruitment of additional populations of motor units was also observed, some of which were not those expected according to Henneman’s size principle (Tucker et al., 2009). When gross activation is estimated from a large area of muscle using surface electromyography, variable results reflecting increased (Poortvliet et al., 2014), decreased (Muceli et al., 2014) or no consistent changes (Hodges et al., 2008; Tucker & Hodges, 2009) in neuromuscular activation can be observed.  Surface EMG measures from different regions within a muscle showed that acutely-induced pain changes the regional distribution of activation within the upper trapezius (Madeleine et al., 2006), likely due to an uneven strength of synaptic projections of nociceptors located in different muscle regions to the spinal cord (Dideriksen et al., 2016). Interestingly, injections performed in the cranial or caudal region of the trapezius resulted in a similar caudal redistribution of activation (Falla et al., 2009; Dideriksen et al., 2016). Redistribution of activation within muscles was also observed in a number of musculoskeletal conditions such as 10  trigger points in the trapezius (Barbero et al., 2016), in the masseter in people with non-specific neck-pain (Testa et al., 2017) and in the low back muscles in people with low back pain (Falla et al., 2014). To my knowledge, regional activation within VM and VL in PFP has never been investigated. 1.7 Patellofemoral pain 1.7.1 Prevalence PFP is a condition characterized by retro- or peri-patellar pain, aggravated by activities such squatting, stair climbing or sitting with the knee bent for an extended period of time. PFP is a common complaint of patients assessed in sports clinics (Devereaux & Lachmann, 1984), and is especially prevalent in young active females (Witvrouw et al., 2000; Boling et al., 2010). While the prevalence in the general population is unknown, studies on military recruits found a prevalence of 12% (males) and 15% (females) and an incidence of 15/1000 (males) and 33/1000 person-years (females) (Boling et al., 2010). PFP alone accounts for more than 16% of running-related injuries seen in the Sport Medicine Clinic at the University of British Columbia, Vancouver (Taunton et al., 2002). In addition, the incidence of knee injuries can exceed 10% in individuals attending running clinics in preparation for the Vancouver Sun Run (Taunton et al., 2003), a major recreational running event with about 50,000 registered participants each year.  1.7.2 Clinical management PFP is diagnosed through clinical examination; the clinical diagnosis is not based on a single test but rather on the combination of symptoms referred by the patient, exclusion of other injuries or pathologies, and selected tests. According to recent guidelines (Crossley et al., 2016), diagnosis of PFP is given when the patient reports pain located behind or around the patella that is aggravated by activities that load the PF joint, such as squatting, stair ambulation, jogging, and 11  hopping. Additional signs may be pain when keeping the knee bent for extended periods of time, crepitus, and tenderness of the patellar facets. Commonly performed clinical tests include pain elicited by patellar palpation, patellar compression, resisted knee extension, quadriceps muscle contraction while applying pressure proximally to the patella (Clarke’s sign), and in specific tasks but reviews found overall low diagnostic accuracy of these tests for PFP (Cook et al., 2012; Nunes et al., 2013). Differential diagnoses considered are: meniscal and ligament injuries (self-report and imaging), tibiofemoral osteoarthritis (age, pain over the tibio-femoral joint line, imaging), patellar tendonitis (pain over the tendon distal to the patella, especially around its tibial insertion), iliotibial band syndrome (pain over the distal iliotibial band insertion, lateral side of the knee), patellar instability (self-report of subluxations), referred pain from hip/back (clinical tests). PF osteoarthritis is commonly excluded by considering age (> 40 years old) and potentially imaging (Hinman et al., 2002). PFP has traditionally been treated with exercise, manual therapy, bracing and pain reduction strategies such as modalities and medications. A meta-analysis revealed that multimodal physiotherapy and exercise improve pain (by an average of 2 points out of 10 on a visual analog scale) and function in the short term (Collins et al., 2012). However, outcomes are often unsatisfactory in the long term (Witvrouw et al., 2004; Blønd & Hansen, 1998; Kannus et al., 1999). For instance, a recent study found that 57% of individuals with PFP reported an unfavorable recovery after 5 - 8 years (Lankhorst et al., 2015). In addition, it was recently suggested that PFP could be a precursor of PF joint osteoarthritis (Crossley, 2014). 1.7.3 Etiopathogenesis of patellofemoral pain PFP is a complex, multifactorial syndrome whose pathogenesis has not been fully elucidated yet. While experts suggest a role of the FP in PFP, definitive evidence for a role of 12  this structure as the source of PFP is still limited (Crossley et al., 2016). Hence, it is currently not known which tissues are the source of PFP, if a single or multiple tissues are involved, or if the irritated tissue is associated with the clinical presentation of the pathology. Regardless, it is likely that the irritated tissue is located within or around the PF joint.  1.7.4 Neuromuscular dysfunction associated with patellofemoral pain If adaptation to pain represents a purposeful adaptation of the neuromuscular system to avoid stimulation of sensitized tissues (Hodges & Tucker, 2011), redistribution of the load within the PF joint should be observed, achieved by changes of activation within the quadriceps muscle. In line with this, two meta-analyses found that decreased knee extension strength (KES) is associated with (Lankhorst et al., 2013) and a risk factor for PFP (Lankhorst et al., 2012). Additionally, higher KES was protective against cartilage loss in the lateral PF joint (Amin et al., 2009), potentially preventing or limiting the progression of PF joint osteoarthritis (Thuillier et al., 2013). These studies provide evidence for a pivotal role of KES in PFP.  Deficits of KES can be due to two main mechanisms: 1) lower quadriceps muscle force production capacity; 2) altered quadriceps muscle activation. These mechanisms are not mutually exclusive and can co-exist in PFP. As individual muscle force production cannot be directly measured in humans, muscle size is usually considered as a surrogate measure of muscle force production capability. Decreased quadriceps muscle size was observed in PFP using magnetic resonance imaging (Pattyn et al., 2011) and ultrasound (Giles et al., 2013). On the motor control side, unbalanced activation of VM and VL has been considered one of the contributors to PFP (McConnell, 1986; Van Tiggelen et al., 2009); this alteration has been described in some (Cowan et al., 2001; Mellor & Hodges, 2005; Van Tiggelen et al., 2009), but not all (Karst & 13  Willett, 1995; Cavazzuti et al., 2010) studies, as summarized in a systematic review (Chester et al., 2008).  Part of this variability may be due to the fact that none of these studies evaluated regions within VM and VL. Because VM and VL have different actions on the patella (Lin et al., 2004; Wilson & Sheehan, 2010), changes in their relative activation may also be used as a strategy to redistribute the load within the PF joint. As neuromuscular dysfunction may be maintained even after pain resolution (Bennell et al., 2010), some of the unsatisfactory long-term outcomes of PFP (Witvrouw et al., 2004; Blønd & Hansen, 1998; Kannus et al., 1999; Lankhorst et al., 2015), and perhaps progression to PF joint osteoarthritis (Thuillier et al., 2013; Crossley, 2014) may be related to failure to resolve alterations of the neuromuscular control once the knee is not painful anymore. Besides local factors, there is evidence that poor motor control of joints other than the knee might also contribute to PFP. In clinical practice, physiotherapists classify individuals with PFP into subgroups with prevalent local (knee), proximal (hip) or distal (foot) dysfunction (Selfe et al., 2013; Witvrouw et al., 2014). Hip rotation/adduction or increased foot mobility may result in tibiofemoral rotation or knee valgus, potentially changing the load distribution within the PF joint. Dynamic foot mobility, defined as change in midfoot width and foot arch height when comparing weightbearing and non-weightbearing positions, was shown to be associated with PFP (McPoil et al., 2011) and to be a predictor of the effectiveness of foot orthoses treatment (Collins et al., 2009). Similarly, reduced hip abductors, extensors and external rotators strength was shown to be associated with PFP (Ireland et al., 2003; Souza & Powers, 2009). A hip muscle strengthening program has been shown to improve PFP symptoms in the short term (Khayambashi et al., 2012). Although a systematic review concluded that there is moderate 14  evidence that lower hip muscle strength is not a risk factor for PFP (Rathleff et al., 2014), a prospective study identified increased hip adduction while running as potential risk in female runners (Noehren et al., 2013). These studies suggest that non-local factors may also play a role in PFP. A detailed description of the neuromuscular mechanisms involved in PFP, taking into account regional activation within VM and VL as a means to redistribute loads within the PF joint, may provide new targets for rehabilitation, possibly reducing the long-term consequences of PFP (Hodges & Tucker, 2011) such as long-standing pain and PF joint degeneration. To my knowledge, the relative activation of regions within VM and VL in individuals with PFP in comparison to healthy controls has not been investigated in detail. 1.8 Methodological approach The investigations described in this thesis were carried out using a number of methodologies. This section will focus on electromyography, and specifically high-density electromyography (HDsEMG), which was the technique used for all chapters. In addition, it is described how factorization algorithms can be used to extract information from HDsEMG. 1.8.1 Electromyography Electromyography is a technique that records muscle activation. Following excitation of the motoneurone, action potentials are generated in the neuromuscular junction and propagate along the muscle fibres towards the musculotendinous junctions. The voltage of the action potential can be detected by electrodes placed close to the muscle fibres (intramuscular electromyography) or, as biological tissues conduct electricity, by electrodes placed on the skin (surface electromyography). Different parameters can be used to extract information from EMG signals. The most relevant in the context of muscle activation is the amplitude of the EMG 15  signal, which is associated with the number of action potentials generated in the muscle in a given time period (Suzuki et al., 2002). As increased neural drive to a muscle results in the recruitment of additional motor units and an increase in discharge rate, there will be an increased number of action potentials propagating in the muscle, and hence an increase in EMG amplitude.  However, several other factors must be considered, making this relationship between EMG amplitude and neural drive non-linear and potentially misleading, especially in between-subject comparisons (Merletti et al., 2001). Some of these factors are: distance between the recording electrodes and the active motor units, motor unit size, amplitude cancellation, muscle architecture, changes in the intracellular action potential shape (e.g.: due to fatigue), location of the electrodes with respect to the innervation zone, etc. The influence of these “confounding” factors increases in conditions where the muscle architecture changes during the recording (Farina, 2006), challenging the interpretation of EMG findings. While the effect of some of these factors can be mitigated by normalization to maximal voluntary contraction (MVC; Keenan et al., 2004; Beck et al., 2008), this can be problematic when comparing asymptomatic participants to populations with disorders that may not allow them to fully recruit the muscle of interest (e.g.: due to pain or other reasons). 1.8.2 Surface electromyography Surface electromyography consists of collecting EMG signals from two electrodes placed on the skin over the muscle of interest. The recording volume of surface electrodes is generally large, meaning that the action potentials collected by the electrodes may be generated by a source far from the recording site; this is commonly referred to as crosstalk (De Luca & Merletti, 1988). For instance, a study using electrical stimulation showed that 5 - 30% of the amplitude collected over VM was actually generated by the VL or vice versa (Farina et al., 2002b). As the surface 16  EMG amplitude decreases with the distance from its source (Roeleveld et al., 1997b), the use of EMG amplitude to investigate the neural drive to the muscle can be particularly problematic for between-subject comparisons, especially when the thickness of the tissues interposed between skin and muscle differs between participants. Amplitude estimates also depend on the position of the electrodes along the muscle fibre. For instance, electrodes placed on the VM bridging the innervation zone record absolute EMG amplitudes up to 75% lower than electrodes placed away from it (Gallina et al., 2013a; Rainoldi et al., 2004; Rainoldi & Nazzaro, 2000). This can especially be misleading in dynamic contractions, where muscle architecture changes during the recording (Farina, 2006); if the innervation zone shifts under the EMG electrodes as a result of muscle shortening, a large decrease in EMG amplitude will be observed despite there being possibly little change in neural drive to the muscle.  1.8.3 High-density surface electromyography Some of the limitations of conventional surface electromyography can be partially overcome by observing the same action potentials from many locations over the skin. This is commonly done using HDsEMG, which consists of a large number of small electrodes organized in grids and placed over the muscle(s) of interest (Merletti et al., 2010). Electrode grids are held in place using bi-adhesive foam, and optimal electrical contact between the grid electrode and the skin is obtained by applying conductive paste. When action potentials enter the sampling volume of one of the electrodes of the grid, the electrical potential is amplified and collected, referencing the intensity of the signal to a reference electrode that collects minimal EMG activity (usually placed on a bony reference). As the same action potential is usually recorded by many electrodes of the grid, HDsEMG provides highly detailed information on how the electrical potentials are distributed on the skin. Having high spatial (usually 8 mm, but can be as low as 0.5 17  mm) and temporal (> 2000 Hz) resolution, HDsEMG provides highly detailed information on temporal and spatial characteristics of surface EMG signals. In muscles with fibres that run parallel to the skin, electrodes placed along the fibre direction will detect where the action potential is generated (neuromuscular junction, also called innervation zone) and the propagation of the potential along the fibres until its extinction at the musculotendinous junction (Merletti et al., 2003). The observation of propagating action potentials ensures that the potentials are generated from the muscle of interest (and are not crosstalk or artefacts). Using HDsEMG it is possible to estimate the EMG indices commonly obtained using conventional surface electromyography, but with less variability associated with electrode position (Gallina et al., 2011, 2013a; Rojas-Martínez et al., 2012), and with reliability across days (Gallina et al., 2016c). Due to its higher spatial resolution than conventional techniques, HDsEMG is particularly well suited to describe regional activation within superficial muscles such as trapezius (Madeleine et al., 2006; Falla et al., 2009; Gallina et al., 2013b), first dorsal interosseus (Zhou et al., 2011), RF (Watanabe et al., 2012b, 2014a), gastrocnemius (Vieira et al., 2010b, 2015; Cronin et al., 2015) and forearm extensors (Gallina & Botter, 2013; Hu et al., 2015). Interestingly, altered regional activation observed with HDsEMG provided insights into the neurophysiology of adaptation of the neuromuscular system in conditions such as aging (Watanabe et al., 2012), delayed onset muscle soreness (Hedayatpour et al., 2008), diabetes (Watanabe et al., 2012), stroke (Rasool et al., 2015) and musculoskeletal disorders (Falla et al., 2014; Barbero et al., 2016; Testa et al., 2017). 18  1.8.4 Factorization algorithms and high-density surface electromyography Factorization algorithms can be used to reduce the dimensionality of an HDsEMG dataset by approximating the information into a smaller number of factors. As the same action potential is observed by many electrodes, in some cases HDsEMG may provide redundant data. For instance, if one investigates regional activation within a muscle, electrodes along the muscle fibre direction provide little information because they detect the same potential propagating along the fibres (Staudenmann et al., 2013). Factorization algorithms such as Principal Component Analysis (PCA, Jolliffe 1986) and Non-negative Matrix Factorization (NMF; Lee & Seung, 1999) have been used to study muscle coordination, i.e.: to extract activation patterns common across multiple muscles, indirectly providing information on neural control strategies (d’Avella et al., 2003; Bizzi & Cheung, 2013). Specifically, if PCA is applied to a dataset of EMG signals collected from, for example, 15 muscles during a dynamic activity, the algorithm will provide 15 principal components (PCs). Each will consist of 15 weights (describing how much each muscle contributes to that specific PC), one temporal coefficient (describing the activation in time of that specific PC), and the variance explained (describing how much of the signal is explained by that specific PC). Usually, > 90% of the variance is explained by the first 2 - 5 components; hence, a low number of PCs can be used to describe the EMG activation patterns of a high number of muscles. When applied to HDsEMG recordings, both PCA (Staudenmann et al., 2009, 2013) and NMF (Muceli et al., 2013; Gazzoni et al., 2014; Huang et al., 2015) identify clusters of electrodes that have similar profiles of temporal activation, approximating the information from 64 - 128 channels into 3 - 5 components. Similarly to that described earlier for individual muscles, in this thesis, factorization algorithms were applied to HDsEMG signals to provide 19  objective information on the spatiotemporal characteristics of the activation of regions within the VM. Although being based on different mathematical principles, factorization of HDsEMG using PCA and NMF yields almost identical results (Gallina et al., submitted), indicating that the factorization is a property of the signals rather than a mathematical artefact. 1.9 Thesis outline The overall aim of this thesis was to investigate whether pain leads to a redistribution of regional muscle activation within VM and VL. This thesis first explored if the human spinal cord has the neuromuscular circuitry to independently recruit motor units located in different regions of the VM. To accomplish this, localized stretch reflexes where evoked by mechanical stimulation of different muscle regions, and regional muscle activation was recorded using HDsEMG and intramuscular electromyography. Next, this thesis sought to determine how VM regional activation estimated with HDsEMG could be influenced by anatomical factors during dynamic contractions. An investigation of regional activation within the VM in a dynamic knee extension task, comparing different knee angles and concentric/eccentric phase of the movement, is explored in this chapter. The next two chapters focus on the redistribution of regional activation within the VM and VL in the presence of two different types of pain, clinical and experimental. The effect of PFP on regional activation within the VM and VL was explored using a cross-sectional design. Between- and within-muscle regional activation was extracted from HDsEMG recordings using PCA and compared between females with and without PFP. The co-existence of neuromuscular dysfunction and clinical measures was also investigated. Finally, the effect of the localization of the experimental pain on vasti regional activation was studied by injecting hypertonic saline solution in different muscular and non-muscular tissues around the knee. 20  1.10 Objectives and hypotheses Chapter 2  Objective: The purpose of this chapter was to determine whether motoneurones innervating muscle fibres located in different regions of the human VM can be independently recruited at the spinal level. Hypothesis: It was hypothesized that stretch reflexes within the VM are localized and vary systematically with the location of the mechanical stimulus. If localized stretch of muscle fibres results in regional activation within the muscle, this would demonstrate that motor units located in different regions can be independently recruited at the spinal level. Chapter 3 Objectives: The purpose of this chapter was: i) to examine the contribution of changes in muscle fibre orientation occurring at different knee angles to any regional activation identified with HDsEMG; ii) to examine whether regional activation, observed as shifts in HDsEMG distribution in the same direction as the regional activation induced by electrical stimulation, is observed during voluntary dynamic contractions. Hypotheses: It was hypothesized that: i) intramuscular stimulation of two regions within the VM would result in localized activation of muscle fibres in those regions and be observed as different EMG amplitude distributions along the columns of the EMG electrode grid. Furthermore, changes in knee angle would result in small shifts of each EMG amplitude distribution along the rows, and no shift along the columns of the EMG electrode grid; ii) in the dynamic contractions, changes in HDsEMG amplitude distribution would be observed along the columns of the EMG electrode grid.  21  Chapter 4 Objectives: The purpose of this chapter was to compare quadriceps muscle regional activation patterns between females with and without PFP. Hypotheses: It was hypothesized that participants with PFP would have altered neuromuscular activation strategies, defined by spatial and temporal features of the PCs extracted from quadriceps muscle EMG activity during a standardized dynamic task. It was also hypothesized that altered neuromuscular activation strategies would be most evident in participants with lower knee extension strength. Chapter 5 Objectives: The purpose of this chapter was to determine whether: i) quadriceps muscle activation and ii) knee force direction are modulated in a manner that is specific to the location of acute noxious input (experimental pain) at different locations within the quadriceps muscle and non-muscular tissue of the knee. Hypotheses: It was hypothesized that activation would be inhomogeneously reduced across regions of the quadriceps muscle and that adaptation of muscle activation and force direction would differ when experimental pain was induced in different locations. It was also hypothesized that adaptive motor strategies would persist after pain resolution.  22  Chapter 2: Regionalization of the stretch reflex in the human vastus medialis 2.1 Introduction  Motor units, consisting of a motoneurone and the muscle fibres it innervates, provide the conduit for the CNS to convert neural signals into forces. It has been known for years that motoneurone soma size is a main determinant of motor unit recruitment order, with smaller motoneurones activated before larger ones (Henneman, 1957), and that motor units within individual muscles and across synergistic muscles share a common command from the CNS (De Luca & Erim, 1994; Laine et al., 2015). However, the existence of motor units with localized territories in humans suggests that the CNS may have more independence in motor unit recruitment and control strategies than previously thought. For example, most (Buchtal et al., 1959; Gootzen et al., 1992; Vieira et al., 2011; Gallina & Vieira, 2015), 70% (Harris et al., 2005) or up to half (Héroux et al., 2015) of the motor units in human muscles have localized territories, meaning that motoneurones innervate muscle fibres clustered in limited muscle regions. While the functional implication of this anatomical structure is currently unknown, it constitutes a basis for the CNS to distribute neural inputs to motoneurones based on the location of the muscle fibres they innervate. This was suggested as a possible mechanism for the CNS to take advantage of heterogeneous muscle architecture (Windhorst et al., 1989; Vieira et al., 2011), and for changes in motor control strategies in the presence of pain (Tucker et al., 2009); yet, there is currently no evidence for the existence of such selective motor unit recruitment in humans.  The regionalization of the stretch reflex requires that motor units located in different muscle regions can be selectively recruited by the human spinal cord. Due to its complex architecture, the VM offers a good opportunity to study the regionalization of the stretch reflex in 23  humans. The VM has a distributed insertion along 40 - 60% of the medial side of the patella (Holt et al., 2008) and on the common quadriceps muscle tendon (Smith et al., 2009b). Proximal-to-distal differences were observed in the orientation of VM muscle fibres (Smith et al., 2009b; Gallina & Vieira, 2015), pennation angle (Blazevich et al., 2006; O’Brien et al., 2010), thickness (Blazevich et al., 2006; O’Brien et al., 2010) and fibre length (O’Brien et al., 2010). Because of these regional differences in muscle architecture and insertion, muscles fibres located in the proximal or distal VM exert forces directed proximally or medially on the patella (Lin et al., 2004) despite the absence of clear anatomical compartmentalization (Smith et al., 2009b). Also, motoneurones supplying the human VM innervate muscle fibres confined to limited regions within the muscle (Gootzen et al., 1992; Gallina & Vieira, 2015), which is a neuro-anatomical prerequisite for regional activation. Regionalization of the stretch reflex may be a mechanism the CNS uses to coordinate the activation of VM regions with different mechanical properties.  In this study, we investigated whether motoneurones innervating muscle fibres located in different regions of the human VM can be independently recruited at the spinal level. Regional activation of 1a afferents localized in different VM regions was produced through mechanical taps and the spatial location of the recruited motor units was investigated using a grid of surface EMG electrodes (Vieira et al., 2011) and confirmed with intramuscular fine-wire recordings. It was hypothesized that stretch reflexes within the VM are localized and vary systematically with the location of the tap. If localized stretch of muscle fibres results in regional activation within the muscle, this would demonstrate that motor units located in different regions can be independently recruited at the spinal level.  24  2.2 Methods 2.2.1 Participants The screening process consisted of applying taps along the VM insertion on the patella and visually assessing whether muscle twitches were elicited. Reflex contractions were observed in all screened individuals but some reported discomfort due to the large amount of force needed to elicit the reflex and were not further tested. Nine individuals participated in this study (1 female; 24 - 53 years old). All participants signed a written informed consent form. The study conformed to the standards set by the latest revision of the Declaration of Helsinki and was approved by the University of British Columbia Clinical Research Ethics Board. 2.2.2 High-density surface electromyography Placement of the HDsEMG grid was guided by anatomical references. The medial and lateral boundaries of the VM were identified with ultrasound imaging (LogicScan 64 LT-1T, Telemed, Vilnius, Lithuania) and were marked on the skin. The innervation zone of VM was located using a linear electrode array (16 silver bar electrodes, 10 mm inter electrode distance, OTBioelettronica, Torino, Italy) moved over different regions of the muscle along muscle fibres while the participants maintained a low-force isometric knee extension contraction. The innervation zone, identified by the bi-directional propagation of the bipolar action potentials observed in consecutive channels, was marked on the skin. Similar to a previous study (Gallina & Vieira, 2015), the innervation zone of fibre groups was found to be oriented diagonally across the VM (Figure 2.1).   25   Figure 2.1: Experimental setup. The innervation zone is illustrated as a dashed black line. Black crosses identify the average location of the tendon taps (mean and standard deviation) across participants. The surface EMG amplitude plot on top of the electrode grid represents the expected spatial distribution of the response to the tap of location 4. The target location for the intramuscular wires is indicated by the three black diamonds.  26  The HDsEMG grid (semi-disposable adhesive matrix; OTBioelettronica, Torino, Italy) consisted of 64 electrodes arranged in 5 columns and 13 rows (an electrode missing in one of the corners), spaced by 8 mm with a total area covered by the electrodes of 3072 mm2 (96×32 mm). The grid was placed proximally to the innervation zone with the long axis of the grid (columns) parallel to it. The distal column of electrodes (column 1) was placed approximately 5 mm from the estimated location of the innervation zone and the medial row of electrodes (row 1) was close to the medial border of the VM muscle (Figure 2.1). Bi-adhesive foam held the grid in place and conductive paste (Ten20, Weaver and Co., Aurora, CO, USA) ensured good electrical contact between the skin and electrodes. Two surface electrodes (20×35 mm; conductive hydrogel; Kendall, Covidien, Mansfield, MA, USA) were placed on the medial side of the knee as reference electrodes. 2.2.3 Intramuscular recordings As differences in motor unit territory estimates in the medial gastrocnemius were recently observed when assessed using HDsEMG versus intramuscular recordings (Vieira et al., 2011; Héroux et al., 2015), intramuscular multiunit EMG signals were recorded in three participants together with HDsEMG. Custom-made electrodes consisted of two 0.05 mm insulated stainless steel wires (California FineWire, Grover Beach, CA, USA) wound together and inserted via a 1.75 inch 25 gauge hypodermic needle (EXEL International Medical Products, St Petersburg, FL, USA). The threaded wires were folded back to create one 4 mm barb and one 10 mm barb, with the insulation removed from the distal 5 mm (longer barb) or 2 mm (shorter barb) to form the recording sites. The large exposed areas of the wires were chosen in order to favour multi-unit EMG recordings. Three wire electrodes were inserted under ultrasound guidance at a ~10 mm depth along the 3rd column of HDsEMG electrodes at rows 3, 7 and 11. A surface electrode 27  was placed over the lateral femoral epicondyle and served as the ground for fine-wire electrode recordings. 2.2.4 Protocol Participants sat in the chair of a Biodex dynamometer (System 4 Pro, Biodex Medical Systems, Shirley, NY, USA) with their lower leg strapped to the knee attachment at 80 degrees knee flexion. Taps were manually applied using a custom-made hammer with a load cell embedded (Force-Displacement Transducer FT 10, Grass Instrument Co., Quincy, Mass, USA). The head of the hammer was a plastic cone with a rounded tip (5 mm diameter). Taps were applied orthogonally to the skin over the VM muscle fibres by the same investigator in all participants while monitoring the tap force and the EMG response on a computer screen. Taps were first applied close to the distal insertion of the most distal fibres of the VM, and then moving proximally following the patellar edge until a clear response was observed (L1, Figure 2.1). The other locations (5 locations maximum) were identified by applying taps progressively more proximally along the edge of the patella in steps of 10 mm until no responses could be observed (Figure 2.1). For each location, taps were applied starting at the edge of the patella, and then moving away from the patella along the muscle fibres. The location that provided the largest EMG responses while minimizing artefacts was marked on the skin. Thirty taps were applied to each location with varying input force to obtain a range of reflex response amplitudes. Surface EMG activity was carefully monitored online to ensure that the VM was at rest when taps were applied. Following muscle taps, participants who took part in the validation with intramuscular EMG recordings were asked to perform three isometric MVCs with verbal encouragement. Tap locations were marked and measured on a coordinate system referenced to the centre of the patella. 28  Surface EMG signals were collected in monopolar configuration using an HDsEMG amplifier (128-channel EMG-USB; OTBioelettronica, Torino, Italy). Signals were amplified (×500 - 1000), filtered (band-pass 10 - 750 Hz) and digitized at 2048 Hz using a 12 bit A/D converter. Differential fine-wire EMG signals were filtered (band-pass 30–6000 Hz; NL 134 and NL 844, Digitimer, Garden City UK), amplified (×1000; NL 820 A and NL 844, Digitimer, Garden City UK) and then A/D converted at 20 kHz (Power 1401 with Spike2 software, Cambridge Electronic Design, Cambridge, UK). The force signal was amplified (×100), low-pass filtered (10 KHz) and simultaneously digitized by the two acquisition systems used for EMG recordings. The force signal collected at 20kHz was used for analysis. 2.2.5 Data analysis All data analysis was performed in Matlab R2013b (The MathWorks, Inc., Natick, MA, USA). EMG signals were band-pass filtered (dual-pass Butterworth, 4th order for each direction; surface: 20 - 400 Hz; intramuscular: 300 - 2000 Hz) before analyses. Tendon taps were identified using the force measured with the force transducer placed in the hammer. The timing of each tap was identified as the first data point after which the force signal reached 5% of the peak force amplitude (Figure 2.2).  29   Figure 2.2: Identification of a surface EMG response. The force signal used to determine the tap onset is depicted on top. EMG signals from five channels along the muscle fibre orientation are plotted in the channels below. Numbers on the right side of the plot are latency estimates of the negative peak of the action potential. Physiological action potential propagation latencies ensured the distinction of artefacts from EMG reflexes.  Epochs from 50 ms before to 450 ms after each tap were analyzed. Surface EMG channels showing artefacts or predominantly power line interference, as determined by visual inspection (less than 10% of the channels; range: 0 - 7 channels), were replaced by the linear interpolation of the four adjacent channels. The onset of the response and the occurrence of action potential propagation along the rows of the electrode grids were used to distinguish the presence of a spinal reflex from mechanical artefact. Only taps that resulted in clear negative peaks delayed by 1 - 3 ms in channels progressively more proximal along the VM fibres (further away from the neuromuscular junction) were included in the analysis (Figure 2.2). Because mechanical taps were applied in a range of forces (see below), no muscle activation in response 30  to the tap was observed in 22% of the trials across all locations. These trials were excluded from all analyses. For each channel, the magnitude of the surface EMG response was calculated as the amplitude of the largest negative peak occurring 15 - 45 ms after the tap. Artefacts due to the tap could sometimes be observed superimposed on the EMG response in columns 1 and 2 and these columns were excluded from the analysis for all participants. An example of surface EMG signals can be observed in Figure 2.3, top rows. For each of the taps where an EMG response was observed, the amplitude values of columns 3 - 5 of each row were averaged obtaining an array of 13 values (an example is shown in Figure 2.1 above the EMG grid). Thus, in each single tap location, up to 30 arrays of amplitude values representing the distribution of the reflex response along the columns of the grid were established.  31   Figure 2.3: Example of responses to taps of location 1 (left panels), 3 (middle panels), and 5 (right panels) for participant 3 (Table 1). Surface EMGs channels (top panels) are organized from distal (ch.1) to proximal (ch.13). Each row shows EMG signals from three channels placed along the approximate fibre orientation. For intramuscular EMG signals (bottom panels), the top signals were collected from the wire inserted in the distal region of the VM, the bottom ones from the most proximal.  Figure 2.4 shows the amplitude range of the responses to the manually-evoked taps from the surface recordings of a representative subject. A consistent spatial localization of the response is observed despite the difference in amplitude. As muscle thickness (Blazevich et al., 2006) and the amount of skin/fat tissues (Botter et al., 2011) change across the VM, applying the 32  same input force across the tap locations does not necessarily ensure similar muscle spindle activation. To ensure that the localization of the stretch reflex across tap locations was not influenced by input force or amplitude of the EMG responses, three separate analyses were conducted: all trials (all mechanical taps), trials matched for input (force-matched) or trials matched for output (EMG amplitude-matched). Force or EMG amplitude matching across different tap locations was done separately. For the force-matched analysis, the largest five taps in each location were selected and the input force values were averaged. The lowest of the average values among locations was chosen as reference. For each location, the 5 trials with input force closest to this reference value were selected and included in the analysis. For the EMG amplitude-matching analysis, selection of the trials was done following the same procedure, but trials were matched for amplitude of the EMG responses instead of force input. For the force-matched and amplitude-matched responses, the coefficient of variation (CoV) was calculated for each participant as the standard deviation divided by mean across tap locations and expressed as a percentage. This index was used to describe the variability of force and EMG amplitude across tap locations to verify that the matching was effective.  33   Figure 2.4: Responses to individual mechanical taps in three locations for a representative participant. The location of the proximal (P), middle (M) and distal (D) intramuscular wires are depicted. Black lines identify the five responses with highest amplitude. For each location, the spatial location of the response is similar across taps. The location of the intramuscular wire in relation to the HDsEMG grid is presented with dashed line in each panel.  In all analyses, the amplitude distributions of the selected trials (5 out of 30 for force-matched and amplitude-matched analyses) or of all trials were averaged for each location, resulting in one array of 13 amplitude values per tap location. Position, amplitude, size and latency of the responses were identified for each averaged distribution as follows: a cluster of channels with amplitude larger than 40% of the maximal value of the 13 channels were identified (threshold determination detailed below; Figure 2.5), and: i) the size of the active region within the VM was calculated as the number of channels included in the cluster; ii) the localization of the EMG response was described as the barycentre of the channels, calculated as  34  barycentre =  with ch being each channel in the cluster, ARV being their Average Rectified Value (measure of amplitude), POS being their position in the array. The channels in the cluster were used to estimate the latency of the response, calculated as the average timing between the onset of the tap and the negative peak of each of the channels.   Figure 2.5: Identification of size and location of EMG responses across the grid. Each gray line is the amplitude distribution calculated from five responses, matched for amplitude across locations. The arrows identify the barycentre (Bar) of the channels above the threshold (black circles) for taps of locations 1 to 5.  35  For intramuscular recordings, the same taps analyzed in the all-trials surface EMG analysis for locations 1, 3 and 5 were included in the analysis. The amplitude of the response for each wire was calculated as the root mean square value in the 10 - 50 ms window after tap onset. The amplitude of the baseline noise (root mean square value of an epoch 10 - 50 ms before the tap) was subtracted from the amplitude of the response. For each tap, the amplitude of the EMG response in each wire was expressed as a percentage of root mean square value measured during isometric maximal knee extension (maximal voluntary electrical activation, MVEA; maximal value of 50 ms epoch calculated with 45 ms overlapping windows). For each tap location, the normalized amplitude measured in each of the three intramuscular electrodes was averaged across the thirty taps resulting in a matrix of 3 participants × 3 tap locations × 3 fine-wire locations. The 40% threshold for the surface EMG analysis was chosen based on the concurrent analysis of the surface and intramuscular EMG signals in the subset of three experiments where EMG signals were collected with both techniques. For each intramuscular recording location, the surface EMG amplitude distribution averaged over all trials was compared to the intramuscular recordings. As seen in Figure 2.3 (bottom rows) no EMG responses were observed in the wires other than the wire corresponding to the location of the mechanical tap. This indicated that the low level EMG activity registered by the surface electrodes located above the intramuscular wires with no EMG activity was not a response to the mechanical tap but was due to volume conduction or crosstalk. Therefore, a series of thresholds from 5 to 95% of the peak value of the surface EMG amplitude distribution were tested for each tap location (3 participants with 3 tap locations each). The lowest threshold that excluded surface EMG channels placed above the intramuscular EMG locations that exhibited no activity was selected for each location. The 36  average threshold value across all 9 tap locations (40%; 38.8% rounded up to the closest 5%, range: 25 - 50%) was used to analyze all data from the HDsEMG. This threshold value is more conservative than the 70% used in other studies (Vieira et al., 2010) and will lead to larger estimated regions of active muscle fibres. 2.2.6 Statistical analysis Sample size calculation was performed using Gpower 3.1. Considering a power of 80%, alpha = 0.05, and standard large effect size of 0.4, 9 participants were required for the main analysis. Statistical analyses were performed using SPSS v. 22 (IBM Inc., Armonk, NY, USA). When data were not normally distributed (Shapiro-Wilk test), non-parametric statistics were used. To verify that the input force or amplitude of the response were effectively matched in the corresponding analyses, Friedman tests were run to assess the effect of Tap location on input force (force-matched) or EMG amplitude of the response (amplitude-matched). The variability of input force and EMG amplitude values across locations was also verified using the CoV.  To investigate the regionalization of the stretch reflexes within the VM, the number of channels in the cluster was used as a measure of size of the active area within the VM and the barycentre of the channels in the cluster was used as a measure of spatial localization of the active area within the VM. The effect of Tap location on the number of channels in the cluster was tested using the Friedman test. The effect of Tap location on the barycentre of the channels in the cluster was tested using analysis of variance (ANOVA) with repeated measures, performed separately for force-matched, amplitude-matched and all-trial analyses. Separate analyses were run to avoid violations of the assumption of independent observations for the ANOVA test. As reflexes from locations 5 and 4 were not observed in some participants, only locations 1, 2 and 3 were compared (additional locations are shown in Figure 2.7). Post-hoc decompositions of main 37  effects were performed using paired Student T-tests with Bonferroni correction. For each pair of locations, effect sizes were calculated as:  where mean and SD are the mean and standard deviation of the difference between the groups. Results from the validation with intramuscular electrodes are reported as the average across participants. Data are reported as mean and standard deviation unless specified otherwise. The statistical significance was set at p ≤ 0.05. 2.3 Results Localized muscle twitches could be visually observed in all participants (Figure 2.6).   Figure 2.6: Frames of a video showing muscle twitches in response to mechanical taps for a single participant. For each location, arrows identify the location of the muscle twitch (top) and are placed in the same position in the images at rest (bottom). 38   Reflexes were observed in the surface EMG signals as a single burst of activity (Figure 2.3), with a mean latency of the largest negative peak of approximately 29 ms (force-matched: 29.2 ± 3.6 ms; amplitude-matched: 28.9 ± 3.6 ms; all-trials: 28.9 ± 3.4 ms). No medium- or long-latency EMG responses to the taps were observed. No muscle activation was observed before applying the mechanical taps (ARV calculated on a 100 ms window 50 ms before the taps, mean: 3.6 ± 0.9 µV). There was no difference in force or EMG response amplitude across tap locations for the trials selected based on these measures, respectively, confirming that the matching was effective (force-matched: p = 0.89, CoV = 3.4 ± 1.8% across participants, 25th - 75th percentiles: 13.3 - 22.7 N; EMG amplitude-matched: p = 0.36; CoV = 9.6 ± 5.4%, 57 - 166 µV; N = 5 taps in each location).  The response to a tap always consisted in a single area of activity within the VM. The size of this active area spanned only few channels for all tap locations (Figure 2.5) with the median being 5 channels irrespective of the tap location for any analysis (all-trials: p = 0.14, 25th - 75th percentiles: 4.5 - 6; force-matched: p = 0.07, 5 - 6; amplitude-matched: p = 0.08, 4 - 6).  39   Figure 2.7: A) Effect of tap location on the localization of the response, force-matched condition. Lines depict the position of the responses on the grid for individual participants. * p < 0.001. B) Localization of the responses for force-matched, amplitude-matched and all-trial conditions.  All participants showed responses when taps were applied in the distal region of the muscle (locations 1 - 3 in Figure 2.1). Taps applied to locations 4 and 5 elicited EMG responses in 8 and 5 participants, respectively (Figure 2.7). The location of the tap influenced the localization of the EMG response on the grid (all-trials: p < 0.001, F(2,16) = 45.5; amplitude-matched: p < 0.001, F(2,16) = 51.5; force-matched: p < 0.001, F(2,16) = 60.5; Figure 2.7). For all analyses, post-hoc testing revealed that each location was different from the other two (all p < 0.01; t > 3.6), resulting in large effect sizes (d > 1.2). Taps applied more proximally along the 40  patella resulted in more proximal responses within the VM than taps applied more distally. This localization of the response was confirmed by the intramuscular EMG recordings (Figure 2.3). Taps applied to the distal location (location 1) resulted in larger EMG responses in the intramuscular wire placed distally in the VM (7.5% MVEA distal; 0.7% MVEA middle; 0.1% MVEA proximal). Similar patterns of localized responses were observed for taps applied to the middle (0.1% MVEA distal; 11.8% MVEA middle; 0.4% MVEA proximal) and proximal location (0.9% MVEA distal; 2.3% MVEA middle; 9.1% MVEA proximal). Responses for each participant are presented in Table 2.1.  Table 2.1: Intramuscular EMG activation (%Maximal Voluntary Electrical activation) in response to mechanical taps. Rows identify different tap locations (DISTAL – location 1; MIDDLE – location 3; PROXIMAL – location 5). Columns represent the three wires placed distally (D), middle (M) and proximally (P). Each row depicts the EMG responses in the three muscle locations for the same mechanical stimulation. For each participant, the wire with expected largest response (gray) recorded amplitude higher than the other two muscle regions in the same row.   PARTICIPANT 1 PARTICIPANT 2 PARTICIPANT 3  D M P D M P D M P DISTAL 3.7 2.0 0.2 1.0 0.0 0.0 17.8 0.1 0.0 MIDDLE 0.1 12.8 0.3 0.0 0.4 0.0 0.0 22.2 0.9 PROXIMAL 0.3 2.4 8.1 0.9 0.8 1.4 1.6 3.5 17.8  41  2.4 Discussion The regionalization of the stretch reflex observed in this study implies that the human spinal cord can independently recruit motoneurones innervating muscle fibres located in different regions within the VM. As regional recruitment was observed in response to the activation of 1a afferents localized in regions separated by only 10 mm, it follows that the human spinal cord has the circuitry to control motoneuronal output regionally based on motor unit location. Mechanical taps applied to the VM muscle fibres and HDsEMG enabled the characterization of the spatial relation between regional stimulation of 1a afferents and location of the motor units recruited by the spinal cord. Mechanical taps were used to activate muscle spindles located in different regions of the VM. Although techniques such as the Hoffman reflex enable a fine control of the input and consistency across trials and conditions (McNeil et al., 2013), mechanical taps can target muscle spindles located in different regions within large, flat muscles. HDsEMG was used to investigate the localization of the EMG response within the VM. As the surface EMG amplitude peaks above the active motor units and decreases with distance from the active muscle fibres (Roeleveld et al., 1997a), surface EMG amplitude distribution obtained with HDsEMG provides information on the position of the active motor units within a muscle (Roeleveld et al., 1997a; Vieira et al., 2011; Gallina & Vieira, 2015; Gallina et al., 2016b). Spatial localization was confirmed in a subset of participants using multiple intramuscular recordings, validating the findings of the HDsEMG as localized activation was observed using both HDsEMG and intramuscular electrodes. Using a threshold based on the intramuscular recordings, this study identified active areas spanning 5 channels of the grid (median value) in response to the mechanical taps. The 40% threshold we used is more 42  conservative than the threshold value previously utilized to identify regional activation in simulated EMG signals (Vieira et al., 2010a). Similar spatial localization for different tap location was also observed when data were analyzed using the 70% threshold (analysis not reported), although the active muscle region was smaller (3 channels, median value). Regardless of the threshold used, the present results imply that the active VM region in response to mechanical taps is not larger than 5 channels. This value, however, may be an overestimation and further experimental validation is needed to determine the threshold that accurately defines the contracting muscle region. Regardless, there are similarities between the regional activation observed in the current and a previous study employing selective, intramuscular stimulation (Gallina et al., 2016b). Previous research on the relationship between surface EMG amplitude distribution and active fibres (Roeleveld et al., 1997a) and the results of the intramuscular recordings in the present study strongly suggest that activation in response to the mechanical tap was regionalized within the VM.  Taps applied to muscle fibres in specific VM regions resulted in EMG reflex responses preferentially observed in some channels of the electrode grid and in a single intramuscular site. In addition, mechanical taps applied to muscle fibres in different VM locations resulted in EMG responses localized in different regions within the muscle. This indicates that the excitation of muscle spindles of a limited region of the muscle does not result in reflex activation of the whole muscle, but instead the reflex is confined to a specific region. Localized activation of the VM to mechanical taps was observed regardless of which taps were included in the analyses (force-matched, amplitude-matched, all-trials), strongly supporting the main results of this study. The localization of the stretch reflex implies regionalization at three levels of the spinal circuitry: preferential response of 1a afferents located in different regions of the target muscle, specific 43  connection of these afferents to motoneurones innervating the same muscle region in the spinal cord, and motoneurones innervating fibres confined in a region of the muscle (Windhorst et al., 1989). Regional response of 1a afferents was demonstrated in animals (Cameron et al., 1981) and humans (McKeon et al., 1984), where mechanical stimuli applied to regions of a muscle were shown to result in discharges of 1a afferents from those regions only. Our results support previous observations suggesting that motoneurones innervate VM muscle fibres confined to limited regions of the muscle (Gootzen et al., 1992; Gallina & Vieira, 2015). The localization of the stretch reflex was shown in cats (Cohen, 1953; Bilotto et al., 1982; Eng & Hoffer, 1997) but not in humans (McKeon et al., 1984). Differences in the results between our study and the one by McKeon and colleagues (McKeon et al., 1984) may be related to differences in the architecture between the tibialis anterior and the VM, e.g.: single long tendon vs. flat insertion along the patellar edge. To our knowledge, this is the first evidence for the regionalization of the stretch reflex in humans. Our results further reveal that this regionalization is distributed quite finely, as clearly separated responses could be observed for locations as close as 10 mm apart. Similarly to the observations for directional preference of motor unit activation in biceps brachii and deltoid (Herrmann & Flanders, 1998), our results indicate that stretch reflexes can be elicited continuously across the VM rather than clustering in anatomically-defined neuromuscular compartments (e.g.: VM longus or obliquus, Smith et al., 2009b).  The regionalization of the stretch reflex may potentially modulate the motor output of large, structurally complex muscles such as the VM. For instance, selective stretch reflexes may be useful in the case of perturbations that result in preferential stretch of a muscle region, such as sudden directional translation of the patella or tibio-femoral rotation which may occur especially in certain activities or sports. The current study indicates that the human spinal cord has the 44  neuromuscular circuitry to modulate spatially the motoneuronal output to VM regions based on regional afferent feedback. It has been suggested that regionalization of afferents and efferents may be used by the CNS to shape patterns of activation in order to optimize muscle performance (Windhorst et al., 1989). The abundance and distribution of muscle spindles in human muscles was suggested to be functionally useful to detect regional changes in length within the muscle and locally regulate the motoneuronal output (Windhorst et al., 1989). Indeed, while the synaptic input is largely shared across motor units both within single muscles and between synergists (De Luca & Erim, 1994; Laine et al., 2015), the common drive between motor units tends to be lower in muscles with a higher density of spindles (De Luca et al., 2009). This suggests that afferent proprioceptive information from muscle spindles may promote more independent motor unit firing patterns. The current study adds that the spinal cord has the circuitry to spatially organize the 1a stretch response within a muscle. Furthermore, this study shows that the human spinal cord has the neuromuscular circuitry to preferentially drive motor units localized in different muscle regions. This constitutes a neuroanatomical substrate for reports of region-specific motor unit recruitment (Herrmann & Flanders, 1998; Butler & Gandevia, 2008) and inhomogeneous alteration of motor unit recruitment and firing rate in the condition of experimental pain (Tucker & Hodges, 2009; Tucker et al., 2009). Future studies should investigate whether regional activation of afferents during voluntary contractions can alter motor unit recruitment strategies. Overall, our results showed that mechanical stimulation of 1a afferents localized as close as 10 mm apart within the human VM resulted in regional recruitment of motor units whose location was organized topographically with respect to the stimulus location. This indicates that the human spinal cord has the neuromuscular circuitry to preferentially modulate the neural drive 45  directed to motor units residing in different muscle regions, which is a neuroanatomical prerequisite for regional activation of skeletal muscles. 46  Chapter 3: Regional activation within the vastus medialis in stimulated and voluntary contractions 3.1 Introduction The activation of different regions within the VM (VM) is known to affect knee extension (proximal VM) and patellar tracking (distal VM) (Lin et al., 2004). For this reason, proximal and distal VM are often considered as separate muscles (Hedayatpour et al., 2009; Pattyn et al., 2013; Tenan et al., 2013). Surface electromyography is one of the techniques commonly used to characterize quadriceps muscle activation patterns during functional activities and exercise. However, the interpretation of the EMG signals can be complicated by factors unrelated to the level of muscle activation (Merletti et al., 2001). For instance, crosstalk (i.e. EMG activity originating far from the target muscle) between quadriceps muscle heads has been previously described (Farina et al., 2002b). However, no detailed information about recordings from specific regions within the muscle has been reported to our knowledge. Extraction of activation patterns is also complicated in dynamic contractions, as the orientation of the muscle fibres in quadriceps muscle changes with the knee angle (Farina, 2006), and this was acknowledged as a limitation in studies that analyzed regional activation within the RF during cycling (Watanabe et al., 2015) and walking (Watanabe et al., 2014b). Thus, more information on how changes in muscle fibre orientation influence surface EMG signals is needed. HDsEMG may be helpful to improve the estimation of VM regional activation in dynamic contractions. HDsEMG comprises several small electrodes arranged in a two-dimensional grid placed over the studied muscles. As the electrodes are closely spaced, it is possible to identify where the main source of EMG activity is localized (Gallina & Botter, 2013) 47  and to describe how the EMG amplitude decreases in space (Vieira et al., 2011). In addition, when electrodes along the muscle fibres are considered, information on the position of the innervation zone can be obtained (Beck et al., 2007; Gallina et al., 2013a), and it was suggested that the innervation zone can be used to track changes in muscle fibre position (Farina, 2006). In isometric contractions of the trapezius muscle, changes in amplitude distribution along the columns and the rows of the EMG electrode grid were previously suggested to be related to anatomical factors and regional activation, respectively (Gallina et al., 2013b). A similar approach may be used to describe whether shifts in spatial amplitude distribution can be observed in different directions when they occur as a result of regional activation versus when the same region is activated but appears in a different location on the grid because of changes in the muscle fibre orientation with respect to the surface electrodes. The purpose of this study was to examine the contribution of changes in muscle fibre orientation occurring at different knee angles to any regional activation identified with HDsEMG. To do so, two regions within the VM were selectively activated at different knee angles using intramuscular electrical stimulation. Intramuscular stimulation has the advantage of activating a consistent, selective area of the muscle. This enables the separation of the effect of regional activation (stimulation of different regions within the muscle) from changes in fibre orientation (stimulation of the same muscle region at different knee angles) on the HDsEMG recordings. Shifts in HDsEMG distribution during voluntary dynamic contractions would then need to be observed in the same direction as the regional activation induced by electrical stimulation to be considered as true regional activation. We anticipated that: i) intramuscular stimulation of two regions within the VM would result in localized activation of muscle fibres in those regions and be observed as different EMG 48  amplitude distributions along the columns of the EMG electrode grid; ii) changes in knee angle would result in small shifts of each EMG amplitude distribution along the rows, and no shift along the columns of the EMG electrode grid. In the dynamic contractions, we hypothesized that changes in HDsEMG amplitude distribution would be observed along the columns of the EMG electrode grid.  3.2 Methods 3.2.1 Participants Twenty-two healthy individuals participated in this study, ten in the intramuscular stimulation protocol and twelve for the dynamic contractions protocol. Participants were included if they were older than 19 years old and if they reported no known neuromuscular disorders or recent injury. Each individual signed a written informed consent form. The study conformed to the standards set by the latest revision of the Declaration of Helsinki and was approved by the University of British Columbia Clinical Research Ethics Board (H14-02035). 3.2.2 Experimental setup Participants sat in the Biodex dynamometer chair (System 4 Pro, Biodex Medical Systems, Shirley, NY, USA) with the lower leg (randomly selected) strapped to the knee attachment for the duration of data collection. Placement of the HDsEMG grid and the stimulation electrodes occurred with the lower limb in a standard position (sitting with hip flexed at 100 degrees and knee flexed at 90 degrees) and was guided by anatomical references. The fibre orientation of two muscle regions within the VM and its medial and lateral edges were identified using an ultrasound imaging system (LogicScan 64 LT-1T, Telemed, Vilnius, Lithuania) and were marked on the skin.  49  The innervation zone was located using a linear electrode array (16 silver bar electrodes, 10 mm inter electrode distance, OTBioelettronica, Torino, Italy) positioned over different regions of the VM while the participants maintained a low-force isometric knee extension. The innervation zone, identifiable as phase opposition of the propagating action potentials, was oriented diagonally across the VM (i.e: aligned from medial-distal to proximal-lateral, Figure 3.1; see also (Gallina et al., 2013a; Gallina & Vieira, 2015) and was marked on the skin. The HDsEMG grid (semi-disposable adhesive matrix; OTBioelettronica, Torino, Italy) consisted of 64 electrodes arranged in 5 columns and 13 rows (an electrode missing in one of the corners), spaced 8 mm with a total area covered by the electrodes: 96x32 mm. The position of the electrode grid was determined according to the following anatomical references: i) innervation zone aligned between the 2nd and 3rd column; ii) center of the grid placed approximately halfway between the medial and distal edge of the VM. The grid was held in place using bi-adhesive foam, and conductive paste (Ten20, Weaver and Co., Aurora, CO, USA) ensured an optimal electrical contact between the skin and the electrodes. Two reference electrodes (2x3.5cm; conductive hydrogel; Kendall, Covidien, Mansfield, MA, USA) were placed on the patella and on the medial side of the knee. Motor unit action potentials are generated in the innervation zone and then propagate along the muscle fibres, hence the monopolar EMG activity detected over the innervation zone represents the activation of the whole fibre. As VM fibres are parallel to the skin, the EMG signals collected with HDsEMG in this study are representative of the activation of a VM region larger than the grid itself.  Intramuscular stimulation of VM was performed using two insulated Teflon-coated stainless steel fine wire electrodes. Each electrode consisted of a 50 µm diameter wire (California Fine Wire Company, CA, USA) with 3 mm of the insulation stripped at the tip and 50  threaded through a disposable 3.5 cm long, 25 gauge hypodermic needle (Becton Dickinson and Company, Franklin Lakes, NJ, USA) for intramuscular insertion. A small hook was made at the terminal end of the fine wire electrode that held the electrode in place after the needle was removed. All wire electrodes and hypodermic needles were sealed and autoclaved in an AMSCO Sterilizer (STERIS Corporation, Mentor, Ohio, USA) for 45 min at 120º C prior to use. Because electrical stimulation performed close to the neuromuscular junction requires lower intensities than stimulation applied to other locations within the muscle (Popovic et al., 1991), the insulated tip of the wire was placed close to the innervation zone (as identified above) by puncturing the skin approximately 25 mm distal to the innervation zone, and advancing the needle at a constant angle of 45 degrees to the skin surface following the muscle fibre orientation previously determined. As shown in Figure 3.1, the electrodes were inserted approximately 40 mm apart, close to electrode rows 5 - 6 (defined: proximal stimulation site) and rows 10 - 11 (defined: distal stimulation site). The stimulation was applied in monopolar modality with single square pulses of 10 µs duration, using the wire as the cathode and a large carbon stimulating electrode (5x10 cm), placed on the back of the thigh as the anode. Stimulation was conducted through a constant-voltage stimulator (Grass S88, Natus Neurology Inc. - Grass Products, Warwick, RI, USA) with a stimulus isolation unit triggered by a digital interface (Power 1401 with Spike2 software, Cambridge Electronic Design, Cambridge, UK).  3.2.3 Anatomical reference Anatomical references are illustrated in Figure 3.1. Shifts in EMG amplitude distribution will be described as “along the columns”, when aligned with the long dimension of the HDsEMG grid with high row numbers being distal and low row numbers being proximal; or 51  “along the rows”, when aligned with the short dimension of the HDsEMG grid, with electrodes closer to the kneecap being lateral and away from the kneecap being medial.    Figure 3.1: Experimental setup. The 64-electrode grid was placed so that the VM innervation zone (dashed line) was located between the 2nd and 3rd column of electrodes. The wires used for stimulation (black triangles) were inserted close to electrode rows number 5 - 6 (proximal) and 10 - 11 (distal). Orientation of the VM fibre was illustrated based on the data from Smith et al. (2009b).  52  It should be noted that the grid used in this experiment did not cover all the innervation zones of the VM motor units. For this reason, throughout the manuscript the terms “proximal” and “distal” do not refer to the VM as a whole but rather as relative to the area of the muscle examined with the HDsEMG system. 3.2.4 Intramuscular stimulation protocol Intramuscular electrical stimulation of two regions within the VM was delivered at 5, 30, 60, or 90 degrees of knee flexion (0 equals to full extension). In addition, to examine the effect of tendon slack on the EMG amplitude distribution, the stimulation was applied at rest or while the participant was maintaining a low-force background contraction. The stimulation intensity was set just above motor threshold for all knee angles (as determined by the presence of a visible muscle twitch). However, the stimulation amplitude could differ between the wires and when the stimulation was applied at rest or with a background contraction. The pulses were delivered approximately 2 s apart. Ten stimuli were delivered in each stimulation location at 4 different knee angles, repeated with/without background contraction. The testing order of the four knee angles was randomized. The knee joint angle was set using the Biodex dynamometer in isometric mode. The amplitude of the background contraction was standardized based on the surface EMG activity. As knee extension torque may be produced by activating different synergists, feedback based on EMG amplitude rather than force was used to standardize VM activation. Participants were asked to maintain a contraction level of approximately 50 µV, which resulted in a knee extension torque between 5% and 10% of the MVCs. The MVC was measured three times with the knee at 90 degrees of flexion at the end of the protocol. 53  3.2.5 Dynamic contractions protocol Participants sat in the Biodex chair with their hip flexed 100 degrees and performed three knee extension isometric MVCs with their knee flexed at 45 degrees. Then the Biodex was set to provide an isotonic resistance equal to 10% of the isometric MVC during knee flexion-extension movements. The participants practiced performing smooth knee flexion-extensions from 90 to 10 degrees of knee flexion for 5 - 8 trials. The protocol consisted of ten cycles of 3s concentric knee extension, 3s eccentric knee flexion, 3s rest. The timing was standardized using a metronome. 3.2.6 Data analysis Electromyographic signals were collected in monopolar modality using an HDsEMG amplifier (128-channel EMG-USB; OTBioelettronica, Torino, Italy). Signals were amplified 200 - 500 times and digitized at 2048 samples/s using a 12 bit A/D converter. Before data processing, signals were band-pass filtered (10 - 400 Hz) using a 4th order Butterworth filter. Channels with noise or artifacts due to bad skin-electrode contact (approximately 1 - 2 channels per participant) were identified through visual inspection and replaced with the linear interpolation of the adjacent channels. Stimulation onset (intramuscular stimulation protocol) or knee angle (dynamic contraction protocol) signals were also digitized with the same EMG acquisition system. All data analysis was performed in Matlab R2013b (The MathWorks, Inc., Natick, MA, USA). In the intramuscular stimulation protocol, the onset of the stimulation artifact was determined by visual inspection for each participant. The stimulation artifact is easily distinguished from the compound muscle action potential as it occurs simultaneously on all the channels, whereas the first negative peak of the M-wave can be observed at different latencies in the adjacent channels along the rows because of its propagation along the muscle fibre. Samples 54  contaminated by the stimulation artifact were excluded from the analysis. The M-waves were similar both for repeated stimulation at a single knee angle (Figure 3.2) and at different knee angles (Figure 3.3).    Figure 3.2: Example of M-waves elicited by stimulation of the proximal (A) and distal (B) muscle regions in a representative participant (knee angle: 90 degrees). Channels are organized as in the electrode grid in fig. 1. Each channel shows 10 responses superimposed. When the stimulation is applied proximally, almost no EMG activity can be observed in the distal muscle region and vice-versa.  55    Figure 3.3: Example of action potentials along columns (A) and rows (B) of the grid. The stimulation artifact is highlighted (dotted lines). M-waves elicited at the four knee angles tested are superimposed. The thicker, grey lines identify stimulations at 90 degrees. In A, action potential propagation can be observed as longer latencies of the negative peak in channels far (1, 4 and 5) than channels close to the innervation zone (IZ - 2 and 3). The negative peak in column 5 occurs 4 ms later than in column 3 (detailed in panel C). This confirms that the rows of the grid were placed along the approximate VM fibre orientation. In B, the difference in latencies across channels is minimal. The negative peak in row 7 occurs 1 ms later than in row 9 (detailed in panel D). Differences in latency between positive and negative peak of each M-wave at different knee angles are likely related to changes in muscle fibre length.  For each stimulus, the amplitude of the response in each channel was defined as the peak-to-peak value of the M-wave, resulting in 64 amplitude values, which were averaged across the 10 stimuli. Afterwards, the spatial resolution of the amplitude distribution was improved by 56  interpolation (spline, factor 8) (Lapatki et al., 2006; Gallina & Vieira, 2015). With this method, a virtual grid with channels spaced 1 mm in both directions was obtained. All the amplitude distributions had a single EMG amplitude peak and values monotonically decreasing along both columns and rows of the grid. The location of this peak on the grid was determined by extracting both coordinates. The analysis of the two coordinates separately enables us to characterize changes in amplitude distribution occurring along the rows from those occurring along the columns of the electrode grid. In the dynamic contractions protocol, each of the eccentric and concentric phases was split into 4 intervals of 20 degrees to provide a total of 8 intervals per trial. For each interval, an amplitude distribution was obtained by calculating the ARV for each channel of the grid. As regional activation is expected to be represented by changes along the columns of the electrode grid, only the maximum of each row of 5 electrodes was considered. This created a virtual array of 13 values, describing the amplitude distribution along the columns of the grid. Regional activation was described as the barycentre of the 3 channels with the highest amplitude. The activation level was quantified as the average amplitude values of the 3 channels, expressed as a percentage of the EMG activity measured during the MVC (amplitude of the 4 highest amplitude values). Also, to characterize the amount of EMG activity measured by the top three electrodes in comparison to the 13-channel distribution, the difference between amplitude of the peak and that of the other channels was quantified as: 100*(M3 - M13)/M13, where M3 is the average amplitude of the three highest channels and M13 is the average of the 13 values. In a secondary analysis, we determined whether isometric contractions performed at different knee angles resulted in different EMG amplitude distributions. The data collected when intramuscular stimulation was applied while keeping a background activation was analyzed. M-57  waves (intervals from 50 ms before to 150 ms after each stimulus) were excluded from the analysis. The ARV was calculated for each channel of the grid in a 15 s epoch moving in steps of 250 ms. The epoch in which the average of the five largest amplitude values was closest to 50 µV was used for the analysis. The EMG amplitude distribution along the columns of the grid was quantified in the same way as for the dynamic contractions.  3.2.7 Statistical analysis Sample size calculation was performed using Gpower 3.1. Considering a power of 80%, alpha = 0.05, and standard medium effect size of 0.25, 9 participants were required for the main analysis. For the intramuscular stimulation protocol, the decrease in amplitude with distance from the peak is reported as the mean and standard deviation at different distances from the peak. Two separate 3-way ANOVA tests were used to test the effect of knee angle, stimulation location and background contraction on the position of the peak of the EMG amplitude distribution along the columns and the rows of the EMG electrode grid. Amplitude data were reported as the peak-to-peak value (positive and negative deflections). To ensure that the stimulation artifact that periodically covered part of the onset of the M-wave did not influence the results of these analyses, 3-way ANOVA tests were applied also on the coordinates extracted from the second, positive peak only. As the main results of the two analyses are comparable, only data for peak-to-peak amplitudes are reported.  For the dynamic contractions protocol, the effect of knee angle and contraction type (concentric, eccentric) on the position of EMG amplitude distribution along the columns of the grid was tested using 2-way ANOVA. Because the amplitude distribution may be influenced by the activation level, the same test was also run on the normalized EMG amplitude to test whether knee angle and contraction type influence the amount of muscle activation. 58  For both experiments, statistical analyses were performed using SPSS v. 22 (IBM Inc., Armonk, NY, USA). All factors were considered as within-subject. The assumption of normally-distributed data (Shapiro-Wilk test) was met for all tests. When sphericity (Mauchly’s test) was not assumed, a Greenhouse-Geisser correction was applied. Bonferroni corrections were applied to post-hoc pairwise comparisons. Post-hoc analysis of the factor angle in the dynamic contractions protocol was performed across the 4 levels using a contrast analysis (Laija, 1997).  For the isometric contractions, the effect of knee angle on the position of EMG amplitude distribution along the columns of the grid was tested using the Friedman test because the data were not normally distributed. Post-hoc pairwise comparisons were run using paired Wilcoxon tests with Bonferroni correction. Comparisons that identified shifts in the amplitude distribution smaller than half an interelectrode distance (< 3.5 mm) are not reported as such small shifts are below the spatial resolution of the electrode grids. The statistical significance was set at p ≤ 0.05. 3.3 Results 3.3.1 Participants Ten participants took part in the intramuscular stimulation protocol (3 female; 30 ± 11 years old; height: 180.7 ± 8.2 cm; weight: 75.9 ± 12.6 kg) and twelve in the dynamic contractions protocol (10 female; 27 ± 4 years old; height: 172.2 ± 9.5 cm; weight: 63.0 ± 12.4 kg). 3.3.2 Intramuscular stimulation protocol Each M-wave was biphasic with an initial negative deflection, partially covered by the stimulation artifact, and a second positive peak. Different latencies across channels due to propagation of the action potential could be observed for the negative but not for the positive peak (Figure 3.2 and Figure 3.3), and this helped identify M-waves from artifacts (highlighted in 59  Figure 3.3). In the channel with the largest response, the latency was on average 14.7 ± 0.9 ms and 28.3 ± 3.1 ms for the negative and positive peaks, respectively. The M-wave peak-to-peak value was a median of 3581 µV (25th - 75th percentiles: 1586 – 7868 µV). The variability of the responses across the ten stimuli was negligible (Figure 3.2). Figure 3.3 illustrates M-waves at all the angles tested. 3.3.3 Changes in M-wave amplitude with distance Each stimulation resulted in a single, well-defined peak of activity in the EMG amplitude distribution (Figure 3.4). From this peak activity, the amplitude decreased monotonically both along rows and columns of the grid. The average decrease in EMG amplitude with distance from the peak averaged across participants is shown in Figure 3.4. When normalized to the peak amplitude, monopolar EMG signals recorded one interelectrode distance (8 mm) away from the peak were on average 90.1 ± 6.2% of the maximum amplitude. Amplitude further decreased with distance to 71.2 ± 11.0% at 16 mm, 51.9 ± 12.5% at 24 mm and became less than 10% (9.9 ± 4.9%) of the peak value at a distance of 64 mm from the peak.  60   Figure 3.4: A) EMG amplitude distributions at different knee angles in a representative participant. The gray scale of each map is normalized between 0 and its maximal value (reported on top of each panel). The peak is localized more proximally (along the columns) when the stimulation is applied proximally (top vs. bottom panels). A lateral-medial shift (along the rows) shift can be observed as the knee is moved to a more extended position (panels on the right). B) Decrease of EMG amplitude with distance. For all participants and conditions (N = 160), amplitude values were normalized to the peak, pooled together and plotted as a function of the distance from the peak in steps of 8 mm (inter-electrode distance). Bars indicate standard deviation.  3.3.4 Intramuscular stimulation The position of the maximum EMG peak-to-peak amplitude was analyzed for the “along the columns” and the “along the rows” coordinates separately (Figure 3.5). For the direction “along the columns” (Figure 3.5), stimulation applied to the proximal region resulted in EMG 61  activity localized on average 47.3 mm more proximal than that resulting from stimulation of the distal region (F (1, 9) = 364.1, p < 0.001). There was an interaction effect of knee angle and background contraction (F (3, 27) = 5.6, p < 0.01). Post-hoc testing revealed that stimulations applied while holding a background contraction resulted in a more proximal coordinate of the EMG peak than at rest at all the knee angles (p < 0.01; mean difference: 3.9 mm), but no differences amongst knee angles were identified (p > 0.06, shift smaller than half an interelectrode distance).   Figure 3.5: Shift of EMG amplitude distribution along columns (A) and rows (B) of the electrode grid. Effects of stimulation site (proximal, grey; distal, black) and background contraction (rest, square; background contraction, circle) can be clearly observed along the columns (A). Effects of knee joint angle can mainly be observed along the rows (B). Data are mean ± standard error; * p < 0.05; ** p < 0.01. 62  For the “along the rows” direction (Figure 3.5), the position of the maximum EMG peak-to-peak amplitude was influenced by knee angle and background contraction (interactive effect, F (3, 27) = 6.4, p < 0.01). The EMG peak shifted medially at each knee angle (post-hoc test, p < 0.05) except when the knee was moved from 30 to 5 degrees while holding a background contraction (p = 0.59). The total shift (from 90 to 5 degrees) was 6.2 mm at rest and 7.5 mm while holding a background contraction. Furthermore, the position of the EMG peak was influenced by stimulation location and background contraction (interactive effect, F (1, 9) = 6.1, p < 0.05). Post-hoc analysis revealed that the EMG peak amplitude was more medial for the proximal than for the distal stimulation site, more so when stimulation was applied at rest (p < 0.05; difference: 4.9 mm) than while holding a contraction (p = 0.08; difference: 3.8 mm). 3.3.5 Dynamic contractions Data from representative participants are shown in Figure 3.6, and data pooled across participants are shown in Figure 3.7. On average, the region of maximal EMG amplitude was localized more distally in eccentric compared to concentric contractions (F(3,33) = 7.62; p < 0.05) and moved proximally (F(3,33) = 4.52; p < 0.05) with knee extension (linear trend, p < 0.05). However, this was not uniformly found in all participants (Figure 3.6, right panel), as indicated by the large standard errors in Figure 3.7.   63   Figure 3.6: Examples of EMG amplitude distributions at different knee angles in the concentric/eccentric phase of the dynamic contraction. The gray scale of each map is normalized between 0 and its maximal value (reported on top of each panel). Black circles identify the peak channels. The peak EMG activity was more proximal in the concentric phase than the eccentric phase at each knee angle for participant A. Participant B shows an opposite pattern (a single knee angle is shown for illustrative purposes).  A proximal shift of the region of maximal EMG amplitude with more extended knee positions was observed in 9 participants out of 12 both in the concentric and in the eccentric 64  phases. A distal shift of maximal EMG amplitude in the eccentric compared to the concentric phase of the movement was observed in 10 participants out of 12 (at 70 - 50 degrees). A trend for an interactive effect between knee angle and contraction type was identified (F(3,33) = 2.47; p = 0.07), and post-hoc comparisons showed that the region of maximal amplitude within the VM was more distal in the eccentric than in the concentric contraction between 30 and 70 degrees (p < 0.05), but not when the knee was more extended or more flexed (p > 0.13). The amplitude of the three highest channels was on average 14.2 ± 6.8% higher than the average of the array of 13 amplitude values. Amplitude increased linearly with the knee angle (linear trend, p < 0.001). An interaction effect between knee angle and contraction type was identified for the peak amplitude (Figure 3.7; F(3,33) = 6.03; p < 0.01) with the amplitude being higher in the concentric than in the eccentric phase of the knee extension between 30 and 10 degrees only (p = 0.05).   Figure 3.7: Peak EMG amplitude distribution along the columns of the electrode grid (A) and changes in muscle activation (B) during the dynamic contractions. Phase of the movement (squares, concentric; circles, eccentric) and knee angles are shown. Data are mean ± standard error; * p < 0.05; ** p < 0.01. 65  3.3.6 Isometric contractions The region of maximal amplitude within the muscle differed across joint angles (F(3,27) = 6.88; p < 0.05). The amplitude distribution was localized more proximally at 30 than at 90 degrees (p < 0.017) and there was a trend for a difference between 60 and 90 degrees (p = 0.022) and 5 and 90 degrees (p = 0.047). No differences were identified across the other angles. 3.4 Discussion This study demonstrated that shifts in EMG amplitude distribution associated with regional activation versus changes in muscle fibre orientation were differentially observed on the two dimensions of the electrode grid. These findings were used to interpret data on regional activation within the VM in dynamic voluntary contractions. In this study, the activation of regions within the VM could be described in space by a single amplitude peak. Across the muscle fibre direction, both monopolar and differential surface EMG signals peak above the location of the active motor units (Roeleveld et al., 1997a). Along the muscle fibre direction, monopolar EMG signals peak above the innervation zone while differentials peak twice, between the innervation zone and each tendon (Rodriguez-Falces et al., 2013). In the current study, the columns of the EMG electrode grid were aligned to the VM innervation zone, and therefore each electrode was placed on a different group of fibres. For this reason, shifts in EMG amplitude observed along the direction of the columns indicate regional activation. Instead, the rows of the grid were placed along the approximate VM fibre direction, hence changes in EMG amplitude distribution along the direction of the rows would be related to changes in fibre orientation. Intramuscular stimulation has two main advantages over surface stimulation. First, intramuscular stimulation can be applied at lower intensities, likely reducing the stimulation 66  artifact. Second, because the wire was placed in the muscle, it increases the likelihood that the same groups of muscle fibres were targeted at different knee angles (Figure 3.3). Surface stimulation electrodes would be more affected by movement of the muscle fibres with respect to the skin which could result in stimulation of different muscle fibre groups with changing knee angle. Of note, the placement of the intramuscular wire in this study ensured that most of the electrical stimulation was applied in close proximity to the fibre innervation zone, likely resulting in a direct stimulation of the neuromuscular junction or of the terminal nerve branches at low stimulus intensities. The combination of intramuscular stimulation technique and HDsEMG may also be used to study the effects of changes in muscle architecture and regional activation on the EMG signals from other muscles, as long as the innervation zone can be properly located. Another application may be for the study of localized myoelectric manifestations of fatigue in response to repetitive electrical stimulation of muscle regions. Electrical stimulation selectively applied through intramuscular fine wire electrodes resulted in a localized peak of EMG activity. Approximately 25 mm away from the peak, the EMG amplitude was decreased by half. At more than 60 mm from the peak, the M-wave amplitude was only 10% of the largest response detected by the grid. Given that VM motor unit action potentials were shown to have localized representation on the skin, suggesting that muscle fibres innervated by individual VM motoneurones are confined to limited muscle regions (Gallina & Vieira, 2015), the amplitude distribution of the M-waves observed in this study suggests that the intramuscular electrical stimulation focally recruited motor units with muscle fibres confined to the VM region of interest. Stimulation applied through the proximal or distal intramuscular electrode had a large effect on the amplitude distribution along the columns. The peak of EMG activity observed on 67  the grid was close to the electrode column where the wire was inserted. This confirms that changes in regional activation within the muscle are reflected in shifts in the EMG amplitude along the columns of EMG electrode grid. The location of the intramuscular stimulation also had a small effect on the position of the EMG amplitude peak along the rows of the electrode grid, resulting in a more medial position of the EMG peak amplitude when the stimulation was applied proximally, especially at rest. Minor differences in the alignment of the electrode grid with respect to the proximal and distal innervation zones might explain this difference. In the stimulation protocol, changes in knee angle mainly influenced the EMG amplitude distribution along the rows of the electrode grid. The total excursion of the EMG amplitude peak with the tested knee angles was close to one interelectrode distance, and it was not uniform across the knee angles. Extending the knee from 90 to 60 degrees of flexion resulted in the largest shift, followed by 60 to 30 and lastly at the final degrees of knee extension some comparisons failed to reach statistical significance. Two studies investigated changes in the position of the innervation zone along the fibre direction at similar knee joint angles. A shift of up to 10 mm in two individuals out of three (Farina et al., 2001) and inconsistent shifts across VM regions (Gallina et al., 2013a) were reported. Our findings were in the same order of magnitude, suggesting that the shift in EMG amplitude distribution resulted from a proximal movement of the innervation zone with increasing knee extension. Changes in knee angle had a minor influence on the EMG amplitude distribution along the columns of the electrode grid, with a non-significant proximal shift of less than half an interelectrode distance. Notably, this shift was much smaller than that related to regional activation (six interelectrode distances), hence shifts of the EMG amplitude along the columns of the electrode grid related to changes in knee angle are unlikely to be a confounding factor in the estimation of regional activation within VM. 68  Holding a background contraction mainly influenced the EMG amplitude distribution along the columns of the electrode grid. The largest M-waves were localized one half of an interelectrode distance more proximally when elicited during a background contraction than at rest. A small shift was identified along the rows of the electrode grid. The inclusion of the background contraction in the testing conditions was necessary to analyze the effect of muscle and tendon slack on the position of the EMG amplitude distribution. Similar to the changes observed with altering knee angle, the relatively small shift (one half of an interelectrode distance) of the EMG amplitude distribution along the columns of the electrode grid related to holding a background contraction is unlikely to impact the estimation of changes in regional activation within the VM. Changes in the barycenter of the EMG amplitude distribution along the columns of the electrode grid were observed during dynamic knee flexion-extension contractions and in the isometric contractions at different knee angles. We suggest that in the dynamic task changes in muscle fibre orientation are unlikely to explain this redistribution. Given the results of the intramuscular stimulation protocol, these changes in amplitude distribution may be explained by preferential regional activation within the VM. This interpretation is further strengthened by the comparison of concentric and eccentric contractions, as significant differences in the amplitude distribution along the electrode grid were observed for the same knee angles (Figure 3.7). While inter-subject variability in muscle anatomy (Holt et al., 2008) and in motor strategies to accomplish the task may have resulted in inconsistent patterns across individuals, pooled data showed that the amplitude distribution along the columns of the grid was localized more distally in eccentric contractions and at more flexed knee joint angles. A similar pattern was observed in the isometric contractions. Differences between these findings and the non-significant ones 69  reported in a previous study (Gallina & Gazzoni, 2013) are likely due to difference in protocols and data analysis. In this study, because the activation level was not different for 7 out of 8 angles tested, the difference in amplitude distribution observed between concentric and eccentric contractions cannot be attributed to differences in the intensity of muscle activation. We cannot exclude the possibility that the intensity of muscle activation had an impact on the proximal shift of the EMG amplitude with increasing knee angle in both concentric and eccentric contractions. The difference in EMG amplitude distribution between concentric and eccentric contractions provides preliminary evidence that regions within the VM can be preferentially recruited according to the mechanical demand of the task. This may reflect strategies to compensate for mechanical efficiency (e.g. regional differences in fibre shortening at different knee angles) and/or to take advantage of the different force vectors produced by fibres residing in different regions of the muscle (Lin et al., 2004). Future studies are needed to investigate the mechanisms and functional consequences of regional muscle activation, as well as the consistency of these strategies across healthy individuals and in clinical populations. This study has some limitations. Some of the measures described in the intramuscular stimulation protocol were on average close to or less than half of an interelectrode distance. While these measures were calculated on signals interpolated from electrodes spaced 8 mm apart, the shift of the peak EMG amplitude was consistent across participants and conditions. Even if the effects that identified shifts smaller than one interelectrode distance were simplified as “no shift”, the main findings of this study would not change. There was an imbalance in the number of male and female participants in the two experiments. While differences in subcutaneous adipose tissue might affect some parameters of the EMG signal, a recent study (Rodriguez-Falces et al., 2013) showed that motor unit depth influences how widely its surface 70  amplitude distributes on the skin but not the position of its peak amplitude, which is the main parameter of interest in this study. Furthermore, the analysis of isometric and dynamic voluntary contractions revealed similar results although being performed on two different groups of participants. High-density surface electromyography revealed consistent changes in the EMG amplitude distribution associated with regional activation and changes in muscle fibre orientation. The placement of the electrode grid according to anatomical references enabled us to distinguish regional activation (shifts of EMG amplitude distribution along the columns of the grid) from changes in muscle fibre orientation associated with changes in knee angle (shifts of EMG amplitude distribution along the rows of the grid). Shift of the EMG amplitude distribution along the columns of the electrode grid related to changes in knee angle and muscle contraction were on average negligible (less than one half of an interelectrode distance) and much smaller than the shift observed for regional activation (six interelectrode distances in this study). When used to investigate within-muscle changes in a voluntary dynamic task, this technique showed a proximal shift of the amplitude distribution in more extended knee positions and in concentric compared to eccentric contractions that likely reflects preferential activation of muscle regions in the VM. 71  Chapter 4: Regional activation within vastus medialis and lateralis in a dynamic task is altered in patellofemoral pain 4.1 Introduction PFP is a disorder common in young individuals engaged in sports. PFP is a complex, multifactorial syndrome whose pathogenesis has not yet been fully elucidated. Historically, poor patellar tracking due to unbalanced activation of the VM and VL muscles has been considered a main contributor to PFP (McConnell, 1986). While some studies support this hypothesis (Cowan et al., 2001; Mellor & Hodges, 2005; Van Tiggelen et al., 2009), others did not find any difference in muscle timing or activation between symptomatic individuals and controls (Karst & Willett, 1995; Cavazzuti et al., 2010). Indeed, a systematic review of the literature on this topic pointed out a ‘substantial and unexplained heterogeneity of results’ (Chester et al., 2008). This variability in the results is likely related to both physiological and methodological reasons. One factor that is likely to underlie this variability is how the muscle activity is measured. The technique most commonly used is surface electromyography (EMG), which consists of a pair of electrodes placed on the belly of each vastus (Karst & Willett, 1995; Cowan et al., 2001; Van Tiggelen et al., 2009). Surface electromyography has a number of limitations: for instance, due to the location of the innervation zone, surface EMG amplitude differences up to 75% can be observed in the VM for electrodes positioned only 15 mm apart (Gallina et al., 2013a). While this effect can be mitigated with normalization during isometric contractions (Beck et al., 2008), the VM innervation zone has been shown to shift under the electrodes at different knee angles (Gallina et al., 2016b), complicating the recording of representative surface EMG activation in dynamic contractions. In addition, as VM motoneurones innervate muscle fibres clustered within 72  the muscle (Gootzen et al., 1992; Gallina & Vieira, 2015), VM activation can be observed regionally both in reflex (Gallina et al., 2017) and voluntary (Hedayatpour & Falla, 2013; Gallina et al., 2016b) contractions. Recent advances in EMG technology allow for the placement of several tens of electrodes on single muscles (HDsEMG), overcoming these limitations and obtaining muscle activity estimates that are less influenced by anatomical factors (innervation zone) or regional activation than conventional surface EMG recordings. Another contributor to the unexplained variability of altered muscle activation in PFP is the potential influence of different clinical presentations on muscle activation. Common concurrent clinical findings in people with PFP include lower KES (Lankhorst et al., 2013), lower hip muscle strength (Souza & Powers, 2009), and higher dynamic foot mobility (McPoil et al., 2011). In addition, interventions focused on different sites such as knee muscle strengthening (Herrington & Al-Sherhi, 2007), hip muscle strengthening (Khayambashi et al., 2012) and foot orthoses (Collins et al., 2009) have all been shown to improve PFP symptoms in the short term. It is possible that altered quadriceps muscle activation is more common in people with PFP who also have weak knee extensors. To our knowledge, no studies have tested whether altered quadriceps muscle activation is associated with PFP clinical presentation.  The purpose of this exploratory study was to compare quadriceps muscle regional activation patterns between females with and without PFP. We hypothesized that participants with PFP would have altered neuromuscular activation strategies, defined by spatial and temporal features of the PCs extracted from quadriceps muscle EMG activity during a standardized dynamic task. We also hypothesized that altered neuromuscular activation strategies would be most evident in participants with lower KES. The research hypotheses were framed within contemporary theories on neuromuscular adaptations to pain, which predict that 73  altered neuromuscular control is one of the factors that could sustain pain and function loss (Hodges & Tucker, 2011). 4.2 Methods 4.2.1 Participants Thirty-six females with symptomatic PFP and 20 healthy, age- and sex-matched control participants were recruited for the study from the community and from local physiotherapy clinics. To be included in the PFP group, participants had to be: female, 19 - 35 years old, with retro- or peri-patellar knee pain of an intensity equal to or greater than 3/10 for at least 1 month aggravated by any of the following activities: sitting for long periods of time, stairs, squatting, running, kneeling or jumping. They also needed to report pain or discomfort during at least one of the following tests: patellar palpation, patellar compression, resisted knee extension with knee close to full extension, isometric knee extension while applying pressure proximally to the patella. Control participants must not have had any knee pain in the last 12 months. For both groups, exclusion criteria were: previous lower-limb surgery, or chronic neuromuscular disorders. All participants provided written informed consent before the start of the experimental session. The study that was approved by the institution's Clinical Research Ethics Board. Age, body mass, height, duration of pain (self-reported), average pain intensity in the previous week (11-point numerical rating scale; 0 meaning: ‘no pain’; 11 meaning: ‘worst pain possibly imaginable’) were obtained for each participant. Physical activity (General Physical Activity Questionnaire; Armstrong & Bull, 2006) and functional limitation (Anterior Knee Pain score; Kujala et al., 1993) were estimated using validated questionnaires. The test leg was the most painful knee (if both were painful) or predetermined for controls before study commencement to ensure equal representation between legs. 74  4.2.2 Clinical tests Dynamic foot mobility was assessed using the ‘foot mobility’ test. Using a validated and reliable procedure (McPoil et al., 2009), foot arch height and midfoot width were measured using a caliper twice; while sitting and standing. The difference between measures taken in non-weightbearing and weightbearing positions was recorded on a sheet and used to describe dynamic foot mobility.  Isometric KES was measured using a Biodex System 4 Pro (Biodex Medical Systems, Shirley, NY). The hip and knee angles were standardized at 85 and 45 degrees, respectively, and the participants were secured firmly to the chair. The resistance was applied approximately 2 cm proximal to the medial malleolus. The participants were asked to contract maximally the quadriceps muscle of the leg being tested, reaching a maximal contraction in approximately 1-2 s and to hold it for at least 3 s. This procedure was repeated 3 times with at least 1 minute of rest between trials. Verbal encouragement was provided during each trial. The highest peak torque of the three trials was used as maximal KES. KES was also normalized to body mass (nKES). 4.2.3 Protocol After a few repetitions to warm-up and familiarize with the task, participants performed 10 repetitive knee flexion-extension movements on the dynamometer from approximately 100 degrees to 5 degrees of knee flexion against a constant resistance of 10% of their maximal KES. A metronome standardized the pace for performing 3s concentric knee extension, 3s eccentric knee extension, 3s rest.  4.2.4 Data collection Similar to a previous study (Gallina et al., 2016b), the HDsEMG grids were placed according to anatomical references (Figure 4.1).  75    Figure 4.1: Experimental setup. Left: placement of the electrode grids on VM and VL. Gray diamonds identify the innervation zones. Right: example of knee joint angle (thick gray line) and monopolar surface EMG signals collected from proximal (P) and distal (D) locations within VM and VL.  The medial and lateral edges of VM and VL were identified using ultrasound imaging (LogicScan 64 LT-1T; Telemed, Vilnius, Lithuania) and were marked on the skin. As thickness of the interposed tissues between the electrodes and the muscle may influence EMG recordings, a single ultrasound image was also taken in the proximal and distal region of both muscles, to correspond with the proximal and distal third of the grid array. VM and VL innervation zones were located using a linear electrode array (16 silver bar electrodes, 10- mm interelectrode distance; OTBioelettronica, Torino, Italy) and marked on the skin. Two HDsEMG grids 76  (semidisposable adhesive matrix; OTBioelettronica) were placed on the skin so that the innervation zone aligned between the second and third column, and all the electrodes were placed over the muscle of interest. Each grid comprised 64 electrodes with an 8 mm inter-electrode distance, arranged in 5 columns and 13 rows with a single electrode missing in one of the corners; the grid was held in place using bi-adhesive foam. Reference electrodes (2x3.5 cm; conductive hydrogel; Kendall, Covidien, Mansfield, MA) were placed on the patella and on the medial and lateral epicondyles. HDsEMG signals were collected in monopolar modality using an EMG amplifier (128-channel EMG-USB; OTBioelettronica, Torino, Italy). Signals were amplified 500 - 1000 times, filtered (band-pass 10 - 750 Hz) and digitized at 2048 Hz using a 12 bit A/D converter. The knee position signal from the dynamometer was acquired simultaneously using the same amplifier. 4.2.5 Data analysis Ultrasound images were analysed using ImageJ (National Institutes of Health, Bethesda, Maryland, USA). The thickness of the interposed tissues was measured as the distance between the skin and the most superficial edge of each muscle. All EMG analyses were run in Matlab 2016B (The MathWorks, Inc., Natick, MA, USA). A Butterworth filter (4th order, 10 - 400 Hz) was applied to the EMG signals before processing. EMG envelopes were calculated for each channel of both HDsEMG grids by full-wave rectifying and low-pass filtering at 8 Hz (Butterworth filter, 4th order). For each participant, the EMG values corresponding to 10 - 90 degrees of the knee flexion-extension repetitions were extracted, and envelopes were normalized to the maximal envelope value across all channels. EMG envelopes were concatenated in two matrices of 128 EMG channels by N samples (N = 36 or N = 20 participants, multiplied by time samples), one for the PFP group and one for the control group, respectively.  77  As the analysis aimed at identifying regional activation within the vasti, PCA (Joliffe 1986) was applied to the HDsEMG dataset. In line with previous studies that ran separate factorization analyses for different conditions or groups (Muceli et al., 2014; van den Hoorn et al., 2015; Diamond et al., 2016), PCA was run separately for PFP and controls. As opposed to running the PCA by pooling all participants together, this approach enables the identification of between-group differences in the spatial weights. However, temporal coefficients could only be compared between-groups only for the PCs that have similar spatial weights (R > 0.95 in this study).  In brief, PCA identifies clusters of channels with a large covariance in time, factorizing the signal into PCs. Each PC can be described by three indices (Figure 4.2; Figure 4.3): 1) spatial weights: the location of the channels where the PC is most represented; 2) temporal coefficients: the time profile of the activation of the PC; 3) the variance explained: how much of the variance of the signal is accounted for by the PC. Each EMG envelope matrix M was factorized into 128 PCs, each consisting of 128 weights and N coefficients. Spatial weights were calculated as the eigenvectors (ζ) of the covariance matrix of M. Temporal coefficients were calculated as ζ T * M, which is the matrix product between the transposed eigenvectors and the EMG envelope matrix. PCs were sorted according to their eigenvalues. Spatial weights, temporal coefficients and variance explained of the PCs were compared between groups. For each participant, the temporal coefficients of the first 4 PCs corresponding to the concentric and the eccentric phase of each repetition were identified and averaged across knee angles and repetitions.   78   Figure 4.2: Example of PCA analysis of HDsEMG signals for a control participant. Left: 8 of the 128 EMG envelopes used for the PCA from 3 repetitions of a control participant. Light and dark gray boxes identify the concentric and eccentric phase of the movement (10 - 90 degrees). Middle: spatial weights of PC1 - 4 (from PCA ran on all controls). Light and dark shades identify positive and negative weights respectively. Right: temporal coefficients calculated from the 3 repetitions on the left, and average temporal coefficients calculated over 10 repetitions, separately for concentric and eccentric phase.  79   Figure 4.3: Example of PCA analysis of HDsEMG signals for a participant with PFP. Left: 8 of the 128 EMG envelopes used for the PCA from 3 repetitions of a participant with PFP. Light and dark gray boxes identify the concentric and eccentric phase of the movement (10 - 90 degrees). Middle: spatial weights of PC1 - 4 (from PCA ran on all participants with PFP). Light and dark shades identify positive and negative weights respectively. Right: temporal coefficients calculated from the 3 repetitions on the left, and average temporal coefficients calculated over 10 repetitions, separately for concentric and eccentric phase.  The coefficient of determination (CD = 1 – SSE / SST, where SSE is the sum of squared residuals, and SST is the total variance of the original signal) was used to calculate the variance explained for the first 4 PCs, separately for the concentric and eccentric phase of each 80  participant. The mean total variance explained was calculated separately for the concentric and eccentric phase of the movement for each participant by varying the number of PCs between one and ten. The minimum number of PCs that accounted for at least 90% of the variance was identified for each participant, separately for the concentric and eccentric phase of the movement.  4.2.6 Statistical analysis Sample size calculation was performed using Gpower 3.1. Considering a power of 80%, alpha = 0.05, an effect size of 0.91 estimated from Cowan et al., (2002) and Mellor & Hodges (2005), 16 participants each group were required for the main analysis. As the secondary analysis involved a correlation within the PFP group, the sample size was increased. All statistical analyses were run using SPSS v. 22 (IBM Inc., Armonk, NY, USA). Parametric or non-parametric tests were run according to normal data distribution and equality of variance. Anthropometric parameters and clinical measures were compared between groups using independent T-tests.  To investigate whether the thickness of interposed tissues differed between females with and without PFP, the effect of group (PFP or control, between-subject factor), muscle (VM or VL, within-subject factor) and location (proximal or distal, within-subject factor) on the thickness of the interposed tissues was analyzed using 3-way mixed model (ANOVA). The three descriptors of muscle activation identified with PCA were compared. To identify which muscle activation patterns contributed the most in each group, the number of PCs that accounted for at least 90% of the variance was compared between groups using Wilcoxon tests, separately for the concentric and the eccentric phase of the movement. To describe whether the spatial localization of the PCs was similar between groups, Pearson correlations were 81  performed on the spatial weights of the first four PCs. PCs whose spatial weights correlated with R > 0.95 were considered similar between groups. To identify between-group differences in the temporal activation of the vasti, the effect of group and phase (concentric or eccentric, within-subject factor) on the temporal coefficients was tested using 2-way mixed model ANOVA, separately for PCs with a similar spatial structure (R > 0.95) between groups. Student T-tests with Bonferroni corrections for multiple comparisons were used for post-hoc comparisons. For PCs whose spatial structure differed between groups, temporal coefficients were compared between the concentric and eccentric phase of the movement using paired Student T-tests. To identify any relation between clinical measures and neuromuscular activation patterns, Pearson correlation analysis was used to test the association between the EMG indices significant in the between-group comparisons and KES, nKES, dynamic midfoot width and dynamic foot height. If the ANOVA identified a significant interaction between group and phase, the correlation was run on a “redistribution index”, calculated as: RI = tecc – tconc where tconc is the temporal score of a PC during the concentric phase and tecc is the temporal score of a PC during the eccentric phase of the movement. Statistical significance was set at p ≤ 0.05. 4.3 Results Anthropometric parameters and clinical measures are reported in Table 4.1. Participants with PFP had knee pain for 12 - 60 (interquartile range) months and did not differ from controls for age, body mass, height, or physical activity level. A significant difference was identified for body mass index (BMI) (p < 0.01), although the average value for both groups fell within the normal range. 82  Ultrasound measurement of the thickness of interposed tissues did not differ between groups (PFP: 9.2 ± 3.5 mm; controls: 8.6 ± 3.5 mm; F(1,54) = 0.67; p = 0.42). Interposed tissues were larger over the VL than VM (9.1 ± 3.4 vs 8.0 ± 3.4 mm; main effects of muscle; F(1,54) = 10.19; p < 0.01), and proximally than distally (9.4 ± 3.8 vs 7.7 ± 2.8 mm; main effect of location; F(1,54) = 20.09; p < 0.001). No interactions were observed (p > 0.25; F(3,33) < 1.36).  Table 4.1: Anthropometric and clinical measures, differences between groups.  CONTROL PFP p value Age [years] 25.6 ± 4.3 26.7 ± 4.1 0.38 Height [cm] 168 ± 9 166 ± 8 0.59 Body mass [kg] 58.2 ± 8.5 62.3 ± 8.9 0.10 BMI [kg/m2] 20.6 ± 1.7 22.5 ± 5.2 < 0.01 Physical Activity  [GPAQ questionnaire, MET] 3153 ± 2034 4018 ± 2961 0.20 Pain Intensity [out of 10] / 4.1 ± 1.3 / Pain Duration [months, interquartile range] / 12 - 60 / Function [Kujala score] 100 ± 0 74 ± 8 / Midfoot Mobility [cm] 0.82 ± 0.17 0.88 ± 0.40 0.44 Foot Height Mobility [cm] 1.18 ± 0.29 1.43 ± 0.17 < 0.01 KES [Nm] 135.3 ± 32.9 116.5 ± 30.6 < 0.05 nKES [Nm/kg] 2.31 ± 0.41 1.88 ± 0.54 < 0.01  A lower number of PCs was needed to reconstruct 90% of the variance of the signal for participants with PFP (median: 2; 25th - 75th percentiles: 2 - 3; Figure 4.4) than for controls (3; 2 83  - 4.5) in the concentric phase of the movement (p < 0.05). No differences were observed in the eccentric phase of the movement (p = 0.20). Given that four PCs explained 92.2 ± 4.0% and 94.7 ± 2.4% for controls and PFP respectively, all remaining analyses have been performed on the first four PCs.   Figure 4.4: Comparison between the minimum number of PCs that explains at least 90% of the variance in the concentric (left) or eccentric (right) phase of the movement. * p < 0.05  Visual assessment of the spatial location of the PCs enabled the determination of which regional activation pattern was being described by each PC. The PCs other than PC1 had both positive and negative values in their spatial weights and temporal coefficients. For instance, in controls (Figure 4.2) the PC3 had positive spatial weights in the distal region of both VM and 84  VL, and negative values proximally. When the temporal coefficients are positive, muscle activation increases in the channels with positive spatial weights (distally) and decreases where they are negative (proximally). By contrast, when the temporal coefficients are negative, muscle activation increases proximally (channels with negative spatial weights) and decreases distally (channels with negative spatial weights). For this reason, the PC3 in Figure 4.2 describes co-activation of the distal region of VM and VL (when the temporal coefficients are positive) and of the proximal regions (when the temporal coefficients are negative). PC1 had positive spatial weights for all the channels, describing simultaneous activation of both vasti, and was similar between groups (R = 0.96). The spatial weight values for PC2 were positive for VM and negative for VL, hence describing the relative contribution of VM and VL. The localization of PC2 was also similar between groups (R = 0.99). PC3 and PC4 differed between groups (R = 0.75 and R = 0.73, respectively). In control participants, PC3 described a co-activation of proximal or distal regions within the vasti, and PC4 described the co-activation of proximal VL and distal VM or vice versa (Figure 4.2). In PFP, PC3 described regional activation within the VL and no concomitant regional activation in VM, and PC4 identified regional activation within the VM (Figure 4.3). As the spatial weights of PC3 and PC4 differed between groups in their spatial location, only the temporal coefficients of PC1 and PC2 were compared between groups. The temporal coefficients revealed that PC1 was more active in the concentric than the eccentric phase of the movement (main effect of phase; F(1,54) = 92.50; p < 0.001; Figure 4.5) and did not differ between groups (main effect; F(1,54) = 2.33; p = 0.14, interactions; F(1,54) < 0.01; p = 0.99). A significant interaction effect for the temporal coefficient of PC2 (F(1,54) = 4.52; p < 0.05) identified differences between groups in the relative activation of VM and VL during the 85  concentric and eccentric phase of the movement. Both groups showed negative PC2 temporal coefficients (indicating prevalent VL activation) in the concentric phase of the movement and positive PC2 temporal coefficients (indicating prevalent VM activation) in the eccentric phase of the movement. This resulted in significantly lower temporal coefficients during the concentric phase than the eccentric phase of the movement (F(1,54) = 64.98; p < 0.001). As an interaction was found for PC2, the “VM/VL redistribution index” was used (Figure 4.4). A Student T-test confirmed that participants with PFP had a significantly lower (0.29 ± 0.34) VM/VL redistribution index than control participants (0.50 ± 0.36; p < 0.05). In controls, the temporal coefficients of PC3 or PC4 did not differ between concentric and eccentric phase of the movement (PC3: 0.02 ± 0.26 and 0.03 ± 0.23, p = 0.67; PC4: -0.01 ± 0.20 and 0.02 ± 0.17, p = 0.36). In PFP, the PC3 was lower in the concentric than the eccentric phase of the movement (-0.02 ± 0.26 and 0.03 ± 0.17, p < 0.05). A similar trend was observed for PC4 (0.00 ± 0.15 and 0.03 ± 0.17, p = 0.06). 86   Figure 4.5: Comparison of temporal coefficients of PC1 (left) and PC2 (right). The contribution of PC1 was larger during the concentric than the eccentric phase of the movement, regardless of the group. For both groups, PC2 was negative (prevalent VL activation) in the concentric and positive (prevalent VM activation) in the eccentric phase of the movement. However, this redistribution was smaller in the PFP than in asymptomatic controls (interaction effect identified by the arrows). * p < 0.05; ** p < 0.01  Correlations were run for the participants with PFP to investigate associations between the EMG indices that were significantly different between groups (VM/VL redistribution index; number of PCs necessary to reconstruct 90% of the variance in the concentric phase of the movement) and clinical measures (Table 4.2). One individual was identified as a potential outlier through visual inspection; to confirm the existence of an outlier, the data were inputted to a linear regression model. For that participant, Cook’s distance measures of 1.05 (VM/VL redistribution index and nKES) and 0.95 (VM/VL redistribution index and KES) were much 87  higher than the cut-off value for outliers (4 / N = 0.11). With the exclusion of that individual, participants with a higher VM/VL redistribution index had lower KES (correlation between VM/VL redistribution index and KES: p < 0.05, R = -0.39; nKES: p < 0.05, R = -0.41; Figure 4.6). No other associations were identified.   Figure 4.6: Scatter plot of KES and VM/VL redistribution index in females with PFP. The data point of the participant excluded from this analysis was crossed. Pearson R identified a moderate inverse correlation between the two variables.   88  Table 4.2: Pearson R correlation values between neuromuscular activation indicators and clinical measures. * p < 0.05; ** p < 0.01.  VM/VL redistribution index Number of Components, concentric Midfoot Mobility [mm] 0.06 -0.18 Foot Height Mobility [mm] 0.07 0.29 KES [Nm] 0.39* -0.08 nKES [Nm/kg] 0.41* -0.02  4.4 Discussion  The regional activation within the quadriceps muscle during a low-force dynamic knee extension task differs between females with and without PFP. The lower number of PCs needed to reconstruct 90% of the variance for those with PFP compared to controls suggests a simpler control strategy in PFP. Further, differences in the spatial weights suggest a lower co-activation between VM and VL in PFP than in controls. While the muscle activation was redistributed from VL to VM between the concentric and eccentric phase of the movement in controls, participants with PFP did so to a lesser extent. The inverse association between VM/VL redistribution and maximal KES in PFP shows that lower KES was associated with higher redistribution of muscle activity between VM and VL.  Altered quadriceps muscle activation patterns have been observed in PFP in this study. During the concentric phase of the knee extension, vasti activation of females with PFP can be explained by two main activation patterns; global activation (PC1), and redistribution between VM and VL (PC2). To reconstruct the signal to a similar extent, control participants needed additional activation patterns consisting in co-activation of regions between VM and VL. Our 89  results of a simpler motor strategy in PFP compared to controls is similar to a previous report for deep hip muscles in participants with femoroacetabular impingement (Diamond et al., 2016). During the eccentric phase, the number of PCs did not differ between groups. However, the additional activation patterns in PFP represented regional activation within each head of the quadriceps muscle instead of co-activation of regions between VM and VL. Altogether, our results suggest that co-activation of the distal (or proximal) regions of VM and VL was one of the four PCs that described the largest variance in controls but not in PFP.  The co-activation patterns between VM and VL identified in this study may represent coordination strategies for optimal patellar tracking in controls. Females with PFP instead mainly relied on general activation and redistribution between vasti, and to a lesser extent on regional activation within each vastus. As the load applied to each vastus in isolation is known to influence the distribution of forces applied to the patella (Lin et al., 2004; Sheehan et al., 2012), the patterns of redistribution between regions of VM and VL in PFP may be associated with the altered patellar kinematics and pressure distribution within the PF joint observed in PFP (Wilson et al., 2009; Pal et al., 2012; Chen & Powers, 2014). Our results of more independent vasti activation are consistent with previous research that identified lower motor unit synchronization between VM and VL in PFP (Mellor & Hodges, 2005) and other studies that identified altered timing and amplitude of vasti activation using surface electromyography (Cowan et al., 2001; Owings & Grabiner, 2002; Pal et al., 2012).  The relative activation of VM and VL during dynamic knee extension is altered in PFP. Participants without PFP showed a prevalent VL activation during the concentric phase and prevalent VM activation during the eccentric phase of the knee extension movement, potentially helpful in ensuring adequate medial patellar tracking. This between-muscle redistribution was 90  limited in PFP. This loss of muscle activation specificity is consistent with that observed in other musculoskeletal pathologies such as low back pain (Hodges & Richardson, 1999). Interestingly, Schabrun and colleagues (Te et al., 2017) recently showed that the cortical representations of the individual heads of the quadriceps muscle are closer together in the cortex of individuals with PFP than in healthy controls. Similar to what was suggested in other studies (Tsao et al., 2011a; Schabrun et al., 2014), this merging of the muscle representations at the cortical level may underlie a lower ability to differentiate muscle activation, i.e.: the lower redistribution of vasti activation between the concentric and eccentric phase of the movement in this study. However, altered excitability at other levels of the nervous system (e.g.: spinal level) cannot be discounted. Future studies should investigate the neurophysiological origin of this dysfunction.  The temporal coefficients of PC1 (global activation pattern) showed that PC1 contributed more during the concentric than in the eccentric phase of the movement, without a difference between groups. This finding suggests a lower muscle activation in the eccentric versus the concentric phase of movement (Christou et al., 2003) is not altered in PFP. The co-activation patterns (PC3 and PC4) in controls were equally observed in the concentric and eccentric phase of the movement. The within-muscle regional activation patterns (PC3 and PC4) in PFP are indicative of preferential activation of the distal VL and VM (trend) in the eccentric phase of the movement, similar to preliminary observations in the VM (Gallina et al., 2016b).  Contrary to our hypothesis, altered neuromuscular control (lower redistribution between VM and VL) was not preferentially observed in participants with lower KES. Instead, an inverse association was observed. A recent classification identified two categories of adaptation to pain: major “movement avoidance” patterns and subtle “redistribution within and between muscle” (Hodges & Smeets, 2014). The current study suggests that the adaptations in females with PFP 91  distribute along a continuum between these two, some favoring a “reduced force output” strategy, some showing more subtle alterations of muscle coordination, and others showing a combination between the two. The lower force output may be associated not only with neuromuscular factors, but also with changes in muscle architecture (Pattyn et al., 2011). It should be noted that, due to the cross-sectional design of the study, it is not possible to identify whether changes in force output and neuromuscular activation are a result of PFP or if they were already present before the pain onset. Regardless, this finding has potential clinical implications as it may be helpful to identify subgroups of participants that respond differently to interventions. Specifically, females with PFP and lower KES may benefit from interventions that focus on quadriceps muscle strengthening, whereas motor control exercises may be beneficial for females with PFP and higher KES. Future studies should investigate whether individualized interventions for females with PFP and higher or lower KES result in improved clinical outcomes.  In conclusion, females with PFP have a lower co-activation of regions between VM and VL and a lower redistribution of activation from VL to VM when the concentric and eccentric phases of the knee extension are compared. As VM/VL redistribution was inversely correlated to maximal KES, it is suggested that our study identified two different presentations of PFP: prevalent lower KES or prevalent lower redistribution between VM and VL. These dysfunctions may be preferentially targeted by different interventions, potentially resulting in improved clinical outcomes. 92  Chapter 5: Location specific responses to nociceptive input support the purposeful nature of pain adaptation 5.1 Introduction Pain-adaptation theories propose that motor strategies change purposefully during pain, to reduce stress on irritated tissues (Murray & Peck, 2007; Hodges & Tucker, 2011). Adaptation of muscles that act on the knee can be investigated using transient acute pain elicited by hypertonic saline injection into the muscles (Tucker et al., 2014a) or the FP (Bennell et al., 2004). Adaptations to unload these muscular or non-muscular tissues to reduce nociceptive activity might be expected to depend on the location of noxious stimulus. Consistent with this proposal, the knee extension force angle changes during noxious stimulation of the medial FP, presumably to unload the painful tissue (Tucker & Hodges, 2010). Likewise, noxious stimulation of specific muscle regions would be expected to change muscle activation to protect the irritated muscle tissue. Yet when vasti muscle activation (using electromyography) or muscle stress (shear wave elastography) have been investigated in the region of noxious input, consistent decreases have only been observed when the contralateral limb could produce compensatory force in a bilateral task (Hug et al., 2014b; Tucker et al., 2014a) but not during unilateral, isometric knee extension when redistribution of activity between muscles/muscle regions, within a limb, would be required to unload the painful part (Hug et al., 2014b; Tucker et al., 2014a). Failure to observe such changes might be explained by the limitations of single-channel electromyography, which assess net changes in a broad muscle area, precluding observation of changes in individual motor units or small areas of the muscle. 93  The complex anatomy of the quadriceps muscle (Blazevich et al., 2006) and its potential to produce forces in different directions (Lin et al., 2004) makes interpretation of changes in muscle activation difficult when measured from a single region. Further, as motoneurones innervate muscle fibres that are spatially localized to regions within the VM (Gootzen et al., 1992; Gallina & Vieira, 2015) and RF (Buchtal et al., 1959), variation of motor unit recruitment between regions within each quadriceps muscle may be a strategy to adapt to pain. The potential for regional variation of vasti muscle activity with intramuscular saline injection has been observed when measured with intramuscular electrodes in 8 locations within VM and vastus VL (Tucker & Hodges, 2009). However, the spatial interpretation of this finding is unclear, as the location of the intramuscular electrode within the motor unit territory is unknown. HDsEMG has revealed regional re-distribution of activation within other large, multifunctional muscles such as the trapezius (Madeleine et al., 2006; Falla et al., 2009), and the masseter (Castroflorio et al., 2012). The HDsEMG technique provides a powerful tool to assess regional variation of muscle activation in response to experimental pain. As some data also show the persistence of modified motor unit recruitment strategy after pain resolution (Tucker et al., 2012), HDsEMG could identify whether this redistribution of activation between regions of the muscle is maintained. This study aimed to determine whether: i) quadriceps muscle activation and ii) knee force direction are modulated in a manner that is specific to the location of acute noxious input (experimental pain) at different locations within the quadriceps muscle and non-muscular tissue of the knee. We hypothesized that activation would be inhomogeneously reduced across regions of the quadriceps muscle and that adaptation of muscle activation and force direction would differ when experimental pain was induced in different locations. We also investigated whether adaptive motor strategies persisted after pain resolution.  94  5.2 Methods 5.2.1 Participants Fourteen individuals with no current knee pain, history of lower limb surgery, or neuromuscular disorders participated in this study (7 females; 18 - 47 years old). Participants provided written informed consent prior to the experimental session. The study conformed to the standards of the latest revision of the Declaration of Helsinki (2013) and was approved by the institutional Human Research Ethics committee. 5.2.2 Experimental protocol The dominant leg (leg used to kick a ball) was tested for all participants. Sitting in a custom-built chair (Figure 5.1), participants performed isometric knee extension contractions with the hip flexed to 100° and the knee flexed to 60° from full extension. The ankle was strapped to a 3-D force sensor (Sensix, France) positioned 2 cm proximal to the medial malleolus. At the beginning of the session, participants performed three MVCs of the knee extensor muscles for 5 s with strong verbal encouragement. 95   Figure 5.1: Experimental setup. The setup was the same used in Salomoni et al., (2016). Image used under the Creative Commons Attribution license CC BY.  5.2.3 Electromyography and force recordings The skin was cleaned with abrasive gel (Neuprep, Weaver and Company, USA) and surface EMG signals from VM and VL were collected using two HDsEMG grids of 64 electrodes (semi-disposable adhesive matrix; OTBioelettronica, Torino, Italy) arranged in 5 columns and 13 rows spaced by 8 mm (Figure 5.2).  96   Figure 5.2: Position of the electrode grids and pain areas for the four locations. Black stars identify the location of the hypertonic saline injections, areas outline by grey lines identify the painful region for each participant. Rows 2 (R2) and 12 (R12; reference for hypertonic saline injection) are identified for the electrode grid placed on the VM in the far right panel.   Each grid was placed with its long axis aligned to the muscle innervation zone. The innervation zone was localized prior to electrode placement using a linear electrode array placed along the approximate muscle fibre direction (16 silver bar electrodes, 10- mm inter-electrode distance, OTBioelettronica, Torino, Italy) that was moved over different regions of the muscle while the participants maintained a low-force isometric knee extension contraction. The medial and lateral boundaries of VM and VL were identified prior to grid placement using ultrasound imaging (Logiq e, GE Medical Systems, China) to ensure that all the electrodes were placed over the muscle of interest. The electrode grid was held in place using bi-adhesive foam. Conductive paste (Ten20, Weaver and Co., Aurora, CO, USA) facilitated good electrical contact between the skin and electrodes. The reference electrodes were placed over the patella and the medial and 97  lateral femoral condyles. The HDsEMG signals were collected in monopolar mode using a HDsEMG amplifier (128-channel EMG-USB; OTBioelettronica, Torino, Italy) and were amplified x500 and digitized at 2048 Hz using a 12-bit A/D converter.  Three-dimensional components of the isometric knee extension force were recorded using a triaxial force sensor (Sensix, France), amplified (×500), filtered (band-pass 0 - 750 Hz) and collected using the HDsEMG amplifier. Surface EMG signals of the RF muscle were collected using a pair of electrodes (Ambu Blue Sensor N, Denmark – interelectrode distance: 20 mm) placed over the rectus femoris muscle belly, ~5 cm proximal to the VM grid. RF EMG signals were pre-amplified (x1000, Wave Wireless EMG, Cometa, Italy), filtered 10 - 500 Hz, digitized at 2000 Hz using a data acquisition and analysis system (Cambridge Electronic Design, UK: 1401 and Spike2 v7. To synchronize the force and EMG recordings, the force signal was split and simultaneously digitized by the two acquisition systems. 5.2.4 Experimental knee pain Acute pain was induced by single bolus injections of hypertonic saline (medial FP: 0.25 ml, concentration 5%; muscle: 0.5 ml, concentration 7%) using a 1 ml syringe and 25G needle (25 mm). The order of the injections was randomized before each experiment. Injection into the FP was performed according to previous studies (Hodges et al., 2009). Muscle injections were performed ~5 mm medially to row 12 (vastus medialis distal; VMD), laterally to row 2 (vastus medialis proximal; VMP) or medially to row 4 (VL, see Figure 5.2). The pain intensity was assessed using a numeric rating scale from 0 to 10, anchored with “no pain” at 0 and “worst pain imaginable” at 10. Participants rated the level of pain that they experienced during each contraction immediately after the contraction was completed. Participants also marked the outline of the painful area on their leg at the end of each condition; photographs were taken and 98  scanned to trace the painful regions. As muscle twitches and cramping were reported after saline injection during pilot testing, participants were asked to report the occurrence of any involuntary muscle activation following the injections. Eight out of 14 participants were naïve to hypertonic saline injections. 5.2.5 Voluntary contractions The experimental protocol consisted of five repetitions of a 10% MVC isometric knee extension ramp contraction with visual feedback of the total knee extension force (5 s ramping up, 10 s hold, 5 s ramping down, 10 s rest). This task was performed before the first injection (baseline), after each of four injections of hypertonic saline solution in different locations beginning when the participant reported a pain of at least 3/10 (pain), and ~1 min after resolution of the pain (0/10; post-pain) for each of the pain locations (Figure 5.3:). Prior to the experiment, participants practiced the submaximal task for 5 repetitions.   Figure 5.3: Experimental protocol. The experimental task (five isometric knee extension contractions) was repeated to warm up, at baseline, and during pain and after pain for each of the four injections of hypertonic saline solution, randomized. The task started when pain was rated 3 or higher in a numeric rating scale (Pain) or 1 minute after pain had completely ceased (post-pain). 99  5.2.6 Data analysis Signals were imported and analyzed using Matlab (v.2016a, The MathWorks, USA). Pain ratings were averaged across the 5 ramps for each pain condition. Force signals were low-pass filtered at 10 Hz (Butterworth, 2nd order). In line with a previous study (Salomoni et al., 2013), changes in the direction of force were quantified as the angle between the knee extension and the tangential force vectors (FX: medio-lateral; FY: proximo-distal). The average knee extension force amplitude and the force direction angles were averaged across the five ramps. For each trial, the force offset (calculated 500 ms before the beginning of the first ramp) was subtracted from the average value to account for subtle changes in lower limb position between trials. EMG signals were band-pass filtered at 10 - 400 Hz (Butterworth, 2nd order). The quality of the HDsEMG signals was carefully assessed and channels that showed noise or large 50 Hz interference due to poor skin-electrode contact were excluded and replaced by the linear interpolation of four adjacent channels (on average, 3 channels each HDsEMG grid). For each ramp, a 5 s epoch between the 4th and 9th second of the 10 s hold phase was extracted and used for the analyses of both EMG signals and force. The intensity of muscle activation was calculated as the ARV from each channel of the grid (VM and VL) and the electrode pair (RF) within the same 5 s epoch of each ramp. This resulted in two ARV amplitude spatial distributions (13x5 channels, VM and VL) and a single ARV value for RF for each condition. The overall activation level of VM and VL was quantified as the average of the 5 highest amplitude values across the electrode grid. To characterize spatial variations of the EMG distribution in the pain and post-pain conditions, distributions of normalized change scores (NCS) were calculated using the following formula for each channel of the grid:  NCSi = 100*(TRIALi-BASi)/BASMEAN 100  where i is the channel of the grid, TRIAL is the ARV distribution of VM or VL during individual pain and post-pain conditions, BAS is the ARV distribution of VM or VL at baseline (before any injections), BASMEAN is the average across all the channels of the grid during baseline trials.  Hence, NCS describes the percentage change of EMG amplitude for each channel of the grid relative to baseline. To identify any regional reduction in muscle activation across conditions, the 5 channels with the lowest NCS were identified (i.e. towards negative values); this procedure identified channels with the largest decrease (or the smallest increase in instances where there was a positive change) of EMG amplitude during the pain or post-pain conditions compared to baseline. The identified channels were characterized in terms of the position (barycenter of their coordinates) and the intensity (average value). The proximal-distal coordinate of the barycenter, which is related to the position of the active muscle fibres within the VM (Gallina et al., 2016b) was used to represent regional activation within the VM and VL. 5.2.7 Statistical analysis Sample size calculation was performed using Gpower 3.1. Considering a power of 80%, alpha = 0.05, an effect size of 1.1 based on Falla et al., (2009), 8 participants were required for the main analysis. To account for interindividual variability, the sample size was increased to 14. Statistical analyses were performed using SPSS v. 22 (IBM Inc., Armonk, NY, USA). Parametric or non-parametric tests were used according to the normality of the distribution of the data as determined with the Shapiro-Wilk test. When Mauchly’s test identified a violation of the assumption of sphericity, a Greenhouse-Geisser correction was applied. All factors were considered within-subject. Pain scores were averaged across the 5 ramps, and compared across the four injection locations (FP, VMD, VMP, VL) using the Friedman test. Separate tests were performed for force and EMG amplitude analyses to compare baseline with pain, and baseline 101  with post-pain. One-way ANOVAs were used to identify differences between baseline and pain (all four locations) and between baseline and post-pain (four locations) for normalized knee extension force and force direction angles; when significant, planned contrasts were used as post-hoc analyses to identify which locations significantly differed from baseline. Similarly, Friedman tests were used to test whether VM, VL and RF EMG amplitudes differed between baseline and the four injection locations, separately for pain and post-pain; post-hoc analyses involved paired Wilcoxon tests to compare each individual location to baseline.  For VM and VL, additional analyses were undertaken to quantify the EMG amplitude changes accounting for regional redistribution within the muscle. To identify whether muscle activity was consistently reduced in a specific region during pain, two-way ANOVAs were used to test the effect of location and pain condition (pain/post-pain) on the position of the channels with lowest NCS. Post-hoc tests identified which conditions were significantly different from 7 (i.e. the midpoint of the electrode grid); if the channels showing the largest decrease were scattered across the grid, or inconsistent across participants, their average position would be close to the middle of the grid. Instead, if they were consistently clustered in the proximal or distal region, a significant difference from baseline would be observed. Bonferroni corrections for multiple comparisons were applied to all post-hoc tests by multiplying the p value by the number of tests. Statistical significance was set at p ≤ 0.05. 5.3 Results Participants generally reported pain in the proximity of the injection site (Figure 5.2). Pain intensity did not differ across locations (p = 0.2; FP: 2.9 ± 1.1; VMD: 3.4 ± 1.2; VMP: 3.3 ± 1.1; VL: 3.1 ± 1.3). Injections were performed 21.0 ± 5.2 minutes apart. The tasks for the post-pain condition started ~1 minute after pain returned to 0/10 which was 13.8 ± 4.8 minutes after 102  the participant reported a pain level of at least 3/10. Nine (7 men) of the 14 participants reported cramping or repetitive muscle twitches following the hypertonic saline injection. These were confirmed by the EMG data (Figure 5.4).    Figure 5.4: Involuntary muscle activation recorded during pain at rest. Left panels: monopolar ARV amplitude distributions that illustrate the location of cramps and repetitive muscle twitching of VMP, VMD or VL for individual participants (P1, P2, etc.). Dark and white boxes identify low and high amplitude values, respectively. Crosses indicate the location of the hypertonic saline injection. Middle panel: Single differential surface EMG signals of a cramp in VMP of one participant after hypertonic saline injection into VMP. High activation can be observed in rows 4 and 5. Right panel: Details of rows 4 and 5 from same recording as middle panel; the innervation zone (intersection of hashed lines) and action potential propagation (angle of hashed lines) can be observed across columns.  The involuntary muscle activation was observed following injection in VMP (5 participants), VMD (3 participants), VL (2 participants), but never after FP injection. One participant reported twitching and cramps in more than one location. The cramping and twitches 103  started before, during, or after the submaximal force task. Visual observation of the EMG distribution revealed that the muscle fibre twitching/cramping was spatially localized in proximity of the site of injection (Figure 5.4). Because of the cramping and twitches observed during the submaximal contractions, 7 ramps from 4 participants were excluded from the subsequent analysis.   Figure 5.5: Effect of experimental pain on medio-lateral (A) and proximo-distal (B) knee extension force direction. Boxplots (median, 25th - 75th quartiles, range) show the changes in knee force angle from baseline (BAS) for each hypertonic saline injection location (FP; VMD; VMP; VL), during (Pain) and after pain (Post-pain). ** p = 0.01; * p < 0.05.  104  During pain, Friedman tests revealed that the surface EMG amplitude significantly decreased for VM (average of 5 recording sites with highest EMG amplitude; p < 0.05) and increased for RF (single electrode pair; p = 0.05), but no consistent changes across participants were observed for VL (average of 5 recording sites with highest EMG amplitude; p = 0.43). Although responses generally differed between individuals, a tendency for the activation of VM and VL to change in the same direction was observed (Figure 5.6). After Bonferroni corrections, post-hoc analyses with paired Wilcoxon tests identified only a tendency for decreased VM EMG amplitude when VMD was injected (median decrease: 10.4%; Bonferroni-adjusted p = 0.08); and no significant change for the other locations (p > 0.76). When the FP was injected, there was a tendency for RF surface EMG amplitude to increase by 12.6%, however this was not significant (Bonferroni-adjusted p = 0.09). A large variability was observed across participants such that no consistent changes were observed when the noxious stimulus was localized in the other muscle locations (p > 0.9). Surface EMG amplitude did not differ from baseline for any of the muscles tested post-pain (p > 0.84). 105   Figure 5.6: Effect of experimental pain on EMG amplitude during pain for individual participants. For each muscle (VM; VL; RF), and hypertonic saline injection location (FP; VMD; VMP, VL), black and gray bars represent increased and decreased activation, respectively, from baseline for each participant. Changes are shown as percentages of the baseline score. Friedman tests identified main effects for VM and RF. Post-hoc comparisons are displayed, † p < 0.09  The 5 channels with the lowest NCS (i.e. larger decrease or smallest increase of EMG amplitude from baseline) were adjacent to each other in most trials (see representative participants in Figure 5.7 and Figure 5.8). For VM, a 2-way ANOVA identified a main effect of injection location (F(3,39) = 2.89, p < 0.05), but no difference between pain and post-pain (F(1,13) = 1.39, p = 0.26) or interaction between these factors (F(3,39) = 0.32, p = 0.81). As no interactions were observed, the values during and post-pain were averaged before post-hoc 106  comparisons. The channels with the lowest NCS were significantly more distal than the midpoint of the electrode grid when VMD was injected (3.1 ± 1.2 inter-electrode distances, t = 9.7, Bonferroni-adjusted p < 0.001) and a similar change in VMD was also observed when VL was injected (2.1 ± 2.3 inter-electrode distances, t = 3.4, Bonferroni-adjusted p < 0.05). No significant effects were observed for the distribution of activation within VL (p > 0.11) or on the medial-lateral coordinate of the location for either VM (p > 0.19) or VL (p > 0.09).   Figure 5.7: Regional activation in response to experimental pain. Left panel: Surface EMG signals collected from proximal (P) or distal (D) monopolar channel at baseline or when VMD was injected with hypertonic saline. Middle panels: Surface EMG amplitude distribution for the two conditions, averaged over the five contractions; P and D identify channels plotted in the left panel. Right panel: Normalized change scores that describe the amplitude change from baseline to VMD, expressed as a percentage of baseline, for each channel of the electrode grid. White circles identify the five channels with lowest normalized change score, and the white cross identifies their barycenter.  107   Figure 5.8: Changes in surface EMG amplitude distribution. Left panels: Normalized change score distributions for VM of a representative participant. White circles identify the channels with lowest change scores, and white crosses indicate their barycenter. Surface EMG amplitude is consistently decreased in the distal region of the VM when VMD, but not VMP, VL or FP are injected with hypertonic saline, both during (Pain) and after pain (Post-pain). Right panel: Group mean (standard deviation) position of the barycenter of the channels with lowest normalized change scores. ** p < 0.001, * p < 0.05.  5.4 Discussion The results of this study reveal adaptations in force and muscle activation that depend on the location of noxious stimulus. Three key observations were made. First, the regional modulation of the quadriceps muscle was observed in response to acute noxious stimulation of the distal region of the VM or VL. Second, the force direction was modified after medial FP 108  injections, but without consistent regional variation in muscle activation. Third, the motor adaptations were maintained even after pain resolution. The specificity of the adaptation of force direction and muscle activation to different locations of the noxious stimulus suggests a purposeful adaptation to protect the painful region (Hodges & Tucker, 2011). Early changes in force direction were consistently observed when the medial FP was injected, whereas changes following VM muscle injections reached statistical significance only post-pain. It is possible that the medial FP may be unloaded by posterior translation of the medial tibial condyle, which when combined with knee extension, could underlie a more medially-directed force vector at the ankle. The only trend for a consistent change in muscle activation after injection of the FP was an increase in RF activation. Although it was not possible to resolve from this study, RF produces moments around the hip joint, which may contribute to the change in knee extension force vector. The absence of systematic changes in the vasti activation with FP injections suggests that the changes in knee extension force direction may be due to changes in the activation of muscles other than the vasti.  Consistent changes of VM activation were observed in response to VMD injection, which led to changes in spatial distribution and decreased amplitude. The location of the EMG channels with the greatest decrease (or smallest increase in instances where there was a positive change) identifies that, regardless of whether activation of the VM muscle as a whole increased, decreased, or did not change with pain, the activation of the distal region of the muscle was relatively smaller than the remainder of the muscle during distal VM pain. This suggests a preferential unloading/less loading of the region in which nociceptive stimulation was applied. Interestingly, similar changes in the amplitude distribution within the VM were observed when VL was injected with hypertonic saline, although to a lesser extent and less consistently across 109  participants. This stereotypical redistribution of activation appears to be, in part, consistent with that observed when noxious stimuli were applied to different regions of the upper trapezius (Falla et al., 2009; Dideriksen et al., 2016). Although speculative, the absence of changes in the spatial distribution when VMP was injected may be related to the different function of the distal (patellar stabilization) vs. proximal (knee extension) regions of the vasti. Taken together, the results provide some support for the hypothesis that adaptation depends on the location of pain, as noxious stimulation of non-muscular (FP) tissues mainly resulted in changes in force direction, whereas noxious stimulation of VMD or VL mainly resulted in regional changes in VM regional activation.  The results of this study show preferential modulation of individual quadriceps muscle heads as well as regions within an individual head instead of uniform inhibition of the agonist muscles in the presence of pain. Across participants, the changes in VM and VL activation were generally observed in the same direction (i.e. both muscles either increased or decreased activation); however, VM activation reduced in response to noxious input of VM more consistently than VL (Figure 5.6). The VM and VL share synaptic input (Laine et al., 2015), and their activity is usually co-modulated in response to pain (Hug et al., 2014b), voluntary activation (Hug et al., 2014a), and fatigue (Kouzaki et al., 2010). In contrast, experimental joint effusion shows that afferent input affects the quadriceps muscle motoneurone pool inhomogeneously, as spinal excitability of VM was reduced with lower volumes of effusion and to a greater extent than VL or RF (Spencer et al., 1984). Differences in the relative timing and amplitude of activation of VM and VL in association with experimental pain (Hodges et al., 2009), delayed onset muscle soreness (Hedayatpour & Falla, 2013) and PFP (Cowan et al., 2001) provide further evidence of partly inhomogeneous effects of pain on the drive to vasti muscles.  110  Although several previous experiments found no evidence for a systematic decrease in the VM EMG amplitude during isometric contractions when noxious input was applied to VM (Hug et al., 2014a, 2014b), here we demonstrated decreased activation. In the present study, EMG amplitude was significantly decreased in the distal part of the VM when the hypertonic saline was injected into the same muscle region. Similar to Hug et al. (2014a), we observed no consistent significant change in EMG activity in a more proximal (less distal) muscle region when that region was injected. The greater impact on the distal region of VM concurs with localized alteration of myoelectric manifestations of fatigue after delayed onset muscle soreness (Hedayatpour et al., 2008).  In this study, some adaptations were observed during pain and persisted after pain resolution, whereas other adaptations were only observed either during pain or after pain resolution. Our findings of an altered distribution of activation within the muscle but an unchanged overall activation post-pain agrees with the observation that the motor unit discharge rate, but not the population of recruited motor units, returns to baseline after pain resolution (Tucker et al., 2012). The inhomogeneous time-course concurs with other studies that showed adaptation to corticospinal excitability both during and after pain resolution (Svensson et al., 2003; Schabrun et al., 2015a), but adaptation of spinal excitability only after pain resolution (Le Pera et al., 2001). Although speculative, this may provide some indication of the underlying mechanisms. That is, the immediate effect of the muscle injection on the distribution of muscle activity may depend on adaptations to the descending command from the motor cortex, whereas the late impact of the muscle injection on the force direction may depend on a spinal mechanism. The time-course of the effect of FP injection on the knee force direction similarly implies cortical involvement, yet why this would differ with the mechanism for muscle injection is 111  unclear. Although not explaining the apparent difference in the time course of some changes, one possibility for the persistence of the changes after pain resolution may be that pain is a motivator to adapt (presumably to protect the tissues), but the resolution of pain is not necessarily a motivator to adapt back to the pre-pain state (Hodges & Tucker, 2011). The highly localized involuntary activation in the HDsEMG recording supports three features of neuromuscular organization that are important for the interpretation of the broader findings of this study. Involuntary muscle activation, in the form of localized cramp or repetitive muscle twitching, was observed after 10/42 muscle injections (in 9/14 participants) but never after FP injections. Visual analysis revealed that the action potentials were highly localized in close proximity to the injection location. When single differential signals were calculated from the monopolar recordings, the innervation zone and action potential propagation could be observed in all cases (Figure 5.4). This strongly suggests that the involuntary muscle activation is generated from the motor end-plate. The higher incidence of this involuntary muscle activity in this study than previous reports may relate to the location of the injections, which were performed close to the innervation zone. Although previous research reported no changes in muscle fibre membrane properties following hypertonic saline injection (indirectly measured as action potential conduction velocity, Farina et al., 2003), a direct peripheral effect of the hypertonic saline on the neuromuscular junction cannot be excluded, especially when the injection is performed in the proximity of the innervation zone. The observation of this localized involuntary muscle activation confirmed that: 1) sensory input from the hypertonic saline injection is highly localized within the muscle, as implied by imaging studies (Graven-Nielsen et al., 1997); 2) VM (and VL) motor units have most of their muscle fibres clustered within small regions of the muscle (Gootzen et al., 1992; Gallina & Vieira, 2015); and 3) regional activation 112  within the vasti can be observed as changes of the surface EMG amplitude distribution in the proximal-distal direction in the HDsEMG recordings (Gallina et al., 2016b, 2017). A possible limitation of this study is that the adaptation to the injections in a single location may have been influenced by the after-effect of a previous injection. However, the order of injection location was randomized, and location-specific changes were observed for all extracted indices which suggests that a cumulative effect, if any, had a minor effect on the results. Adaptation to acute experimental knee pain involved a systematic reduction of activation of the distal region of the VM, changes in the muscle activation across the quadriceps muscle heads, and modification of the knee force direction (more medial during pain) observed to a different extent depending on whether the noxious stimulation was applied to specific locations within the quadriceps muscle tissue or to the FP. Changes in the spatial distribution of activity provide evidence of inhomogeneous distribution of drive to the motoneurone pool of VM in the presence of noxious input. Regional changes in muscle activation and changes to knee force direction persisted even after pain resolution, whereas changes in the overall amplitude of muscle activation returned to baseline. These findings provide new insight on the short-term adaptation to pain. 113  Chapter 6: General discussion 6.1 Summary The results of thesis provide a better understanding of the physiology of regional muscle activation of the human quadriceps muscle and how it is altered by pain. Specifically, regional activation of VM (and VL) in electrically stimulated, reflex and voluntary contractions, as well as in response to location-specific experimental pain and clinical PFP, were investigated. Altogether, this thesis provides strong evidence for the existence of regional activation within the VM both in reflex and voluntary contractions. Regional responses to acute pain and regional modulation of quadriceps muscle heads in PFP compared to controls support the foundational model of this thesis (Hodges & Tucker, 2011). Although the focus of this thesis was on basic mechanisms, some findings may be relevant for clinical application in the rehabilitation of knee musculoskeletal disorders. For example, the association between altered regional activation and KES in PFP and the regional activation within the VM in a dynamic task may be relevant for exercise prescription. Methodological, anatomical, neurophysiological and clinical advancements of this thesis are discussed below. 6.2 Methodological advancements 6.2.1 Combining intramuscular stimulation and high-density surface electromyography to infer muscle architecture This thesis describes the combined use of intramuscular stimulation and HDsEMG, which enabled robust interpretations of quadriceps muscle anatomy. Previous studies showed the possibility to selectively activate quadriceps muscle regions using electrical stimulation (Botter et al., 2011) and recorded the mechanical output (Lin et al., 2004). However, these studies could only provide general information on the localization of activity, i.e.: proximal or distal, based on 114  visual observation of muscle contraction or conventional EMG detection. Muscle anatomy was previously studied using HDsEMG by averaging action potentials triggered from motor unit recordings (Lapatki et al., 2006; Vieira et al., 2011; Gallina & Vieira, 2015). However, this technique does not enable the study of muscle architecture at rest, or to track fibres of the same muscle region with large variations of joint angle due to the difficulty in following the same motor unit (Altenburg et al., 2009).  The combination of selective electrical stimulation and HDsEMG described in this thesis provides a novel tool to address questions related to muscle anatomy. In addition to the spatial location of active muscle fibres, other information such as orientation (Lapatki et al., 2006; Gallina & Vieira, 2015), length (Schulte et al., 2005) and conduction velocity (Falla & Farina, 2005) of fibres located in different muscle regions may be obtained by integrating existing computational methods to signals collected with selective electrical stimulation and HDsEMG. 6.2.2 Influence of crosstalk Findings from individual studies in this thesis provide evidence that crosstalk influenced HDsEMG signals to a small extent. In Chapter 1, highly localized VM activity was observed in response to mechanical stimulation both in HDsEMG and intramuscular EMG signals. Chapter 2 showed that recording monopolar signals 60mm away from the location of the highest muscle response to intramuscular stimulation resulted in M-waves with amplitude ~10% of the peak. In Chapter 3, crosstalk on the HDsEMG would have been seen as a PC describing similar variation of EMG activation in the proximal region of both grids. This was observed in PC3 in controls, but not in females with PFP. If crosstalk from RF influenced the HDsEMG signals to a large extent, it would have been observed in both groups, and especially in the PFP group who had larger BMI. Instead, the absence of a PC describing co-activation of proximal VM and VL 115  regions suggests that crosstalk from the RF was minimal in participants with PFP and that PC3 in controls rather represents co-activation of proximal or distal regions within VM and VL. Finally, in Chapter 4 are described highly localized cramp-like contractions, and RF had activation patterns different from VM and VL in response to FP injections. These data suggest that crosstalk influenced the findings of this thesis to a small extent. 6.2.3 Interpretation of surface electromyographic amplitude in dynamic contractions By tracking muscle fibre location at different knee joint positions, this thesis also provided valuable information with respect to surface EMG detection and interpretation. In dynamic contractions, shifts of the innervation zone under or away from the electrodes due to changes in joint angle (Gallina et al., 2013a) and contraction intensity (biceps; Piitulainen et al., 2009) can result in large changes of EMG amplitude unrelated to changes in the neural drive (Farina, 2006). In Chapter 3, relatively small changes in the location of the VM innervation zone were identified for different knee joint angles (on average < 10 mm along the rows, < 5 mm along the columns of the grid). When rest and muscle contraction conditions were compared, shifts were only observed along the columns of the grid (< 10 mm). Based on these data, it is suggested that conventional electrodes should be placed at least 10 mm away from the innervation zone of the VM. Moreover, interpretation of the EMG findings should take into account that the sampling volume of the EMG detection system may change by approximately 10 mm during a dynamic contraction. 6.2.4 Stretch reflexes elicited by mechanical taps applied to the muscle In Chapter 2, a methodology is used to selectively stretch muscle fibres located in different VM regions, as opposed to stretching the entire muscle. The use of a carefully positioned HDsEMG system allowed the observation of action potential propagation, 116  differentiating spinal reflexes from artifacts due to the mechanical taps (Figure 2.2), and the existence of reflexes was simultaneously validated with intramuscular electrodes in a subset of participants.  A potential confounding factor of a muscle tap compared to a tendon tap is the elicitation of an “idiomuscular response”, i.e.: a direct activation of the muscle fibres in response to the mechanical tap, described in previous publications (Brody & Rozear, 1970; Magistris & Kohler, 1996). Compared to spinal reflexes, the idiomuscular response is observed at a much shorter latency (3 - 5 ms, Magistris & Kohler, 1996; or immediately after the tap, Brody & Rozear, 1970). Also, the idiomuscular response is more easily elicited when the percussion is applied close to the motor point of the muscle (Magistris & Kohler, 1996), which in the VM is close to the center of the muscle belly (Botter et al., 2011) and far from its insertion on the patella. Finally, muscle percussion was shown to result in both idiomuscular response and stretch reflex, observed as two separate peaks of muscle activation in the EMG signal, not an idiomuscular response alone (Brody & Rozear, 1970). Importantly for Chapter 3, the latency of the HDsEMG responses (negative peak: 29 ms) and intramuscular (Figure 2.3) EMG signals, the existence of a single response and the application of the mechanical tap far from the motor point indicate that the EMG responses are stretch reflexes of spinal origin rather than idiomuscular responses. A similar methodology may be used in future studies to test the regionalization of stretch reflexes in different muscles or changes in spinal excitability in different muscle regions. 6.2.5 Twitching and cramping in response to hypertonic saline solution injection In Chapter 5, hypertonic saline solution was injected in four different locations to investigate neuromuscular adaptations to experimental pain localized in different muscular and non-muscular tissues around the knee. The incidental finding of highly localized 117  twitching/cramping (Figure 5.4), as well as the region-specific adaptation observed for the distal VM injection, support previous imaging studies showing that hypertonic saline solution remains locally confined in the fibres close to the injection site (Graven-Nielsen et al., 1997). The unusually high prevalence of twitching and cramping described in Chapter 5 may be related to the fact that hypertonic saline solution was consistently injected in the proximity of the innervation zone, whereas in other studies the location of the injection was typically far away from it (Falla & Farina, 2005; Falla et al., 2009). Although more research is needed to understand the origin of the twitching/cramping response, the data from Chapter 5 suggest that injections of hypertonic saline solution should be performed away from the innervation zone to avoid twitching and cramping that may influence the EMG recordings. Localizing the innervation zone, using HDsEMG or published literature (Barbero et al., 2012), before the injection of hypertonic saline solution may help avoiding cramping and twitching. 6.3 Anatomical advancements 6.3.1 Localized muscle fibre distribution in vastus medialis motor units The spatial localization of muscle fibres innervated by individual motoneurones is currently debated in the literature (Héroux et al., 2015; Blouin et al., 2016; Vieira et al., 2016). Evidence supporting spatial localization of muscle fibres in the human VM includes: i) intramuscular EMG data show that the in-depth territory of VM motor units is 2 - 8 mm (Gootzen et al., 1992); ii) the motor unit amplitude distribution obtained from HDsEMG combined with model simulation showed that most fibres are confined to less than 38 mm (median value) on the plane of the skin (Gallina & Vieira, 2015); iii) studies collecting motor 118  units from proximal and distal regions reported observing potentials limited to one of the two sites (Tenan et al., 2013, 2016; Cabral et al., 2017). In this thesis, the results of Chapters 2, 3, and 5 also support the fact that muscle fibres belonging to individual VM motor units are not spread across the whole muscle volume, but are instead clustered in muscle sub-volumes. In Chapter 2, mechanical taps elicited stretch reflexes that were highly localized within the VM. Simultaneous intramuscular recordings provided strong evidence of the localization of such reflex responses (Figure 2.3). In Chapter 3, localized intramuscular stimulation applied close to the innervation zone of fibres in the proximal or distal VM resulted in M-waves that could be observed only locally in proximity of the stimulation site. In Chapter 5, cramps and twitching were observed as localized EMG activity in the HDsEMG (Figure 5.4). In all these studies, the localized EMG amplitude distributions on the HDsEMG were accompanied by highly localized, strip-like twitches on the skin (Figure 2.6). These data strongly support the existence of spatial localization of muscle fibres innervated by individual VM motoneurones. 6.3.2 Localized spinal circuitry in vastus medialis Experimental evidence that the human spinal cord has the neuromuscular circuitry to independently drive motor units located in different muscle regions was previously missing. In Chapter 2, mechanical stimulations applied close to the VM insertion on the patella resulted in recruitment of motor units located topographically with respect to the location of the stimulus. While the main point of this chapter was the investigation of localized efferent projections from the spinal cord to the muscle, two other spinal mechanisms were shown in this chapter: i) localized afferents from muscle regions to the spinal cord, in line with previous studies 119  (Cameron et al., 1981; McKeon et al., 1984); and ii) specific connections in the spinal cord between localized afferents from one muscle region and efferents to the same region. Chapter 5 also provides some evidence of regional topographical organization of nociceptive projections to motoneurones, as hypertonic saline solution injected in the distal VM resulted in a systematic decrease in activation of the same muscle region only. The inconsistent responses across participants observed with acute experimental pain induced in the proximal VM and VL may be a reflection of the complex adaptation of the neuromuscular system in the presence of pain (Hodges & Tucker, 2011). Finally, redistribution of activation within the VM was observed in the dynamic flexion/extension task in Chapter 3. This redistribution would not be possible if the spinal cord could not preferentially drive motor units located in different muscle regions. Data from these studies support the notion that the human spinal cord has the neuromuscular circuitry to preferentially drive motor units localized in different VM regions. Due to the known association between the location of the muscle fibres within the VM and their mechanical action (Lin et al., 2004), the regionalized motor unit recruitment observed in Chapter 2 may also be considered to support the principle of “neuromechanical” control of motoneuronal output (Butler & Gandevia, 2008). By showing that motor units were recruited regionally based on their location within the VM, Chapter 2 also showed that motor units possessing the potential to produce forces in different directions may be independently recruited by the spinal cord. With respect to the principle of neuromechanical advantage, Chapter 2 integrates previous studies (Desmedt & Godaux, 1981; ter Haar Romeny et al., 1982; Herrmann & Flanders, 1998; Gandevia et al., 2006; Butler & Gandevia, 2008; Hudson et al., 2009) showing that region-specific recruitment applies to a muscle without clear anatomical compartmentalization (Smith et al., 2009b). Similarly, the regional activation observed in 120  Chapter 3 may be used by the CNS to take advantage of different mechanical efficiency of proximal and distal VM in the concentric and eccentric phases of the knee extension movement. Overall, the localization of muscle fibres innervated by individual VM motoneurones and the regionalization of the spinal circuitry shown in this thesis suggest a neuromuscular structure designed to activate the VM regionally. This would allow the CNS to take advantage of the different force vectors produced by VM region, supporting the principle of “neuromechanical” control of motoneuronal output. 6.4 Neurophysiological advancements 6.4.1 Relevance of localized stretch reflexes The finding of localized stretch reflexes described in Chapter 2 may be relevant for human movement in three ways: i) to induce selective regional activation in response to sudden perturbations; ii) to modulate regional activation in response to changes in muscle architecture during movement (Windhorst et al., 1989); iii) to determine motor unit recruitment strategies within the “neuromechanical” control of neuronal output (Butler & Gandevia, 2008) when muscle regions also have different mechanical actions. It is possible that selective VM regional activation may be observed, for instance, in response to sudden patellar translation due to direct contact or potentially through femoro-tibial rotations.  The observation that motor units located in different regions of the muscle can be recruited independently does not necessarily mean that sub-groups of motor units are controlled independently in every-day movements. Motor units within a muscle are thought to be controlled in synergy due to synaptic input being largely shared across motor units both within single muscles and between synergists (De Luca & Erim, 1994; Laine et al., 2015). Synergistic activation of motor units is considered to help the CNS decrease the degrees of freedom that 121  need to be controlled during movements. However, the extent to which motor units are controlled in synergy may depend on the muscle function. De Luca et al. (2009) reported an inverse association between common drive between motor units and spindle density, showing that the common fluctuations of motor units tends to be lower in muscles with a higher density of spindles. While the distribution of the spindles in relation to the muscle volume is likely to be more important than the density alone, this suggest that the CNS may rely on feedback from muscle regions to control regional motoneuronal output in multifunctional muscles. Recent evidence (Cabral et al., 2017) further supports this idea, showing that the common drive is higher for motor units located within the same compartment (proximal or distal) than between proximal and distal pairs. Due to the abundance of muscle spindles in large, structurally complex muscles such as the VM, regionalized stretch reflexes may be useful to detect movement-related changes in length of regions within the muscle and locally regulate the motoneuronal output (Windhorst et al., 1989). For instance, regionalized reflexes may be useful to adjust motoneuronal output on the basis of changes in regional muscle length in dynamic contractions.  Overall, the novel finding of regionalized stretch reflexes in humans is an important first step to revisit current theories of motor control and neuromuscular models, as suggested in a recent editorial (Héroux, 2017).  6.4.2 Neuromuscular adaptation to acute pain is location-specific Chapter 5 tested three parts of the foundational “Pain Adaptation” model (Figure 1.1) of this thesis: i) redistribution of activation between- and within muscles; ii) short-term benefit: adaptation is purposeful to protect the painful/injured tissue; iii) potential long-term consequences: failure to return to pre-pain motor strategies in the absence of experimental pain. By showing that the largest decrease in the EMG activation of the distal VM was preferentially 122  observed when the same muscle region was injected, and that changes in the knee extension force direction were mainly observed when the FP was injected, this chapter showed that the location of the painful stimulus influences the adaptation of the neuromuscular system. If the changes in knee extension force direction are considered to be a way to unload the FP, this chapter supports the purposeful nature of the adaptation to pain to protect painful/injured tissues. However, adaptations unrelated to the site of injection (regional redistribution within VM when VL was injected, changes in force direction after VM injections after pain resolution) were also observed, although their magnitude was smaller than those observed for location-specific injections. Differences between the current and previous studies that show similar adaptations regardless of the site of noxious stimulation (Falla et al., 2009; Dideriksen et al., 2016) may be due to differences in the target muscle (vasti vs. trapezius) or in the experimental protocol (trapezius used synergistically as a scapular stabilizer as opposed to the vasti acting as main agonists).  A further point raised in Chapter 5 is the persistence of the altered motor strategies after resolution of the nociceptive stimulus, i.e.: when the participants were pain-free. This is in line with previous findings of altered motor unit recruitment strategies (Tucker et al., 2012) and corticospinal excitability (Svensson et al., 2003; Schabrun et al., 2015a) after pain resolution. Although this adaptation was tested only in the very short-term in this chapter, sustained alteration of motor strategies in the absence of nociceptive input may be a mechanism for sustained pain in the long term, recurrence of pain, progression to osteoarthritis or injury of other sites (Hodges & Tucker, 2011).  123  6.4.3 Vasti neuromuscular activation in females with and without patellofemoral pain The use of PCA to compare the regional activation within and between VM and VL identified three main differences between participants with PFP and controls: i) patterns of co-activation between regions of VM and VL in control participants as opposed to independent regional activation within VM or within VL in people with PFP; ii) a simpler motor control strategy in the concentric knee extension in the PFP group; and iii) reduced redistribution of EMG activity between VL (concentric phase) and VM (eccentric phase) in females with PFP. These differences confirm the existence of altered activation of VM and VL in PFP (Cowan et al., 2001; Owings & Grabiner, 2002; Pal et al., 2011, 2012) and provide novel information on the differences in regional activation patterns. While the spatial structure of the PC1 (overall activation) and the PC2 (alternation between VM and VL) were similar between groups, PCs 3 and 4 differed. PCs 3 and 4 in controls represented synergistic activation of VM and VL that may be finalized to control PF kinematics and pressure distribution. The same PCs in PFP instead represented independent VM or VL activation. Similar to a previous study that found lower motor unit synchronization between VM and VL in people with PFP compared to controls (Mellor & Hodges, 2005), it is speculated that more independent VM and VL regional activation may be associated with different patellar kinetics and kinematics (which could be an adaptation to or a risk factor for PFP). Indeed, previous cross-sectional studies identified an association between VM/VL activation and patellar position (Pal et al., 2011, 2012).  The complexity of motor strategies was measured as the number of PCs necessary to reconstruct 90% of the signal. In line with a previous study on femoroacetabular impingement (Diamond et al., 2016), control strategies were simpler in people with PFP than controls during the concentric phase of the knee extension. Specifically, participants with PFP relied on global 124  VM and VL activation and the redistribution of activity between the two, whereas the control participants also had a third PC describing co-activation between proximal or distal regions between VM and VL. The smaller contribution of co-activation patterns in the PFP group may influence and be associated with altered patellar kinematics or pressure distribution within the PF joint in PFP (Wilson et al., 2009; Chen & Powers, 2014). While control participants demonstrated a redistribution of activation between the VL and the VM in concentric versus eccentric contractions, participants with PFP did so to a lesser extent. This lack of phase-specific patterns of activation is consistent with the loss of task-specific muscle activation observed in other pathologies such as back pain (Hodges & Richardson, 1999), and potentially is associated with the merging of the cortical representation of individual heads of the quadriceps muscle in PFP (Te et al., 2017). A similar loss of muscle-specific activation was observed when the firing rate of single motor units from VM and VL was compared between PFP and control groups (Gallina et al., 2018b). Overall, the differences described in Chapter 4 point to an altered co-activation and a redistribution of activation between VM and VL in PFP that may contribute to different distribution of force around the patella in PFP. 6.5 Clinical advancements 6.5.1 Exercise prescription to target regions within the vasti While researchers in physiotherapy have been searching for a way to preferentially activate the distal region of the VM during exercise for a long time, a systematic review found no evidence that altering lower-limb position or adding co-contraction resulted in preferential activation of distal VM over VL (Smith et al., 2009a). This thesis provides novel findings that during slow, low-force dynamic contractions in an isotonic dynamometer: i) eccentric 125  contraction resulted in preferential VM activation when compared to VL (Chapter 4); and ii) eccentric contraction and more flexed knee position resulted in preferential activation of the distal compared to proximal VM (Chapter 3). Eccentric contractions in a reduced range of motion can be implemented in clinical practice, although it is not known whether the neuromuscular activation patterns are similar using free weights versus a Biodex dynamometer. However, it should be noted that Chapter 4 did not find a consistent reduction in the activation of the distal VM in the PFP participants compared to controls. While eccentric quadriceps muscle contractions at more flexed knee positions may preferentially target the distal VM, their effectiveness in treating PFP or other knee musculoskeletal disorders should be properly tested before their use can be recommended in clinical practice. 6.5.2 Link between pain location and presentations of musculoskeletal disorders In Chapter 5, the motor adaptation of the vasti was shown to depend on the location of the experimental pain. Following this model, different presentations of musculoskeletal syndromes such as PFP may be associated with pain originating from different sources. The detection of these individual-specific altered motor control patterns may be difficult in cross-sectional designs or in clinical practice, due to a lack of knowledge of the pre-pathology motor control patterns and to the large variations in motor strategies in individuals without pain. Whether identifying and selectively treating individual neuromuscular impairments results in improved clinical outcomes should be tested in future clinical studies. 6.5.3 Association between knee extension strength and neuromuscular dysfunction Traditionally, changes in the neuromuscular activation strategies of the knee extensor muscles were generally considered to be associated with decreased maximal KES in PFP. A study using acute experimental pain model supports this hypothesis (Salomoni et al., 2016). 126  However, in Chapter 4, females with PFP had both lower maximal KES and different neuromuscular activation strategies (lower redistribution between VM and VL) than controls. In fact, there was an inverse association between the two variables, meaning that females with PFP who had low KES had neuromuscular activation strategies comparable to controls, while those with strength similar to controls had larger differences in neuromuscular activation. This appears to be consistent with current theories that classify the motor adaptation to pain (Hodges & Smeets, 2014) in “major avoidance behaviour” (e.g.: avoid producing high knee extension forces) as opposed to “subtle redistribution” (e.g.: changes in VM/VL redistribution). A possible explanation for this inverse association is that participants adapted to PFP in an individual-specific manner, either decreasing KES or changing how they activated their vasti. These clinical presentations of PFP could match the subgroup ‘strong’ and ‘weak’ identified recently (Selfe et al., 2016). The potential clinical implication of these findings is that females with PFP and high KES may benefit from motor control exercise and techniques that change neuromuscular activation patterns (e.g.: biofeedback, Bennell et al., 2010; Gallina et al., 2016a; taping, Cowan et al., 2002; McCarthy Persson et al., 2009). Instead, females with PFP and low KES may already have adequate neuromuscular activation patterns and may benefit from techniques that promote muscle size and power such as strengthening exercise and electrical stimulation. Future clinical studies should investigate whether interventions targeting motor control in females with PFP and high KES, or targeting KES in females with PFP and low KES, result in improve clinical outcomes compared to a-specific exercise. 6.6 Limitations There are several important limitations in the interpretation of the results of these studies: 127  1) This thesis mainly examined VM (all studies) and VL (Chapters 4 and 5) due to their potential to apply medial-lateral forces to the patella; RF was included only in Chapter 5. Given the architectural and functional differences between the quadriceps muscle heads, the findings of this thesis may be muscle-specific and not generalizable to the other heads of the quadriceps muscle. Also, an examination of the activation patterns of RF and vastus intermedius in Chapter 4 may have provided a better understanding of the differences in quadriceps muscle activation patterns between females with and without PFP. 2) The electrode grids covered an area of approximately 10x4 cm. While this area is smaller than the area of VM and VL, placement of the EMG electrodes above the innervation zone enabled the collection of EMG activity that was representative of a large portion of the muscle (see Chapter 3 for details). However, it should be noted that the electrode grids did not cover the most proximal VM and VL, so the activation patterns of these regions could not be characterized in this thesis. 3) The physiological and clinical relevance of changes in EMG activation described in this thesis are based on the assumption that vasti activation is a major contributor to the amount and distribution of force applied to the patella. While studies on cadavers (Elias et al., 2009; Wünschel et al., 2011; Lorenz et al., 2012) and in-vivo (Lin et al., 2004; Sheehan et al., 2012) support this hypothesis, the changes in load applied to the muscle in these experimental conditions are probably much larger than those observed in this thesis. For this reason, the differences in vasti activation observed in this thesis may result in minimal differences in force produced, and hence patellar kinematics or pressure distribution. 128  4) Due to the cross-sectional nature of Chapter 4, it is not possible to define whether differences in neuromuscular activation patterns between females with and without PFP are an adaptation to pain or a precursor to PFP. 5) Due to sex-differences in biomechanical features (Rathleff et al., 2014) and prevalence (Boling et al., 2010), only females with PFP were recruited for Chapter 4. For this reason, findings are not necessarily generalizable to males with PFP. 6) In Chapter 5, an acute experimental pain model was used. It is well known that features of chronic pain differ from those of acute pain (Schabrun et al., 2015b). Adaptation to long-term pain, which is likely relevant for clinical musculoskeletal conditions, may differ from that described in this thesis. 6.7 Implications and future directions 6.7.1 Representativeness of surface electromyographic recordings This thesis highlights the importance of considering regional activation in the investigation of quadriceps muscle activation patterns, especially when comparing individuals with and without clinical or experimental pain. The selective activation of regions 10 - 15 mm apart in the VM was observed in response to localized mechanical stimuli, meaning that the human spinal cord can selectively recruit motor units with a high degree of spatial selectivity. This highlights the importance of collecting EMG signals that are representative of multiple muscle regions. For instance, Figure 5.7 shows that, in response to experimental pain, the EMG amplitude decreases up to 15% in a highly localized region in the distal VM, and increases up to 10% more proximally. If EMG signals were only collected distally, it could have been concluded that neural drive to VM decreased. Instead, electrodes placed only proximally would have led to the conclusion that experimental pain results in an increased neural drive to the muscle. 129  Importantly, as conventional EMG techniques have low selectivity, EMG amplitude measured with a pair of large electrodes on the VM would have likely shown no change as a result of the averaging of two activation responses and suggested no change in neural drive to the VM. These results indicate that caution should be used when interpreting results from surface EMG collected in a single muscle region. Future studies should investigate the relevance of regional activation in different muscles and tasks and the electrode configuration that yields accurate estimates while minimizing the number of electrodes. 6.7.2 Regionalization of reflexes The finding of a regionalized stretch reflex and partly regionalized responses to noxious stimulation in humans opens the possibility to investigate the regionalization of other afferent/efferent systems, such as Golgi tendon organs, cutaneous receptors, and others. Also, building on the evidence in this thesis, future studies should investigate whether any regional variation of the sensory feedback from muscle spindles during voluntary contractions influences motor unit recruitment strategies. 6.7.3 Implications of location-specific adaptation to pain The evidence for location-dependent changes in neuromuscular activation supports the purposeful nature of adaptation to pain, with the aim of protecting the painful/injured part. While speculative at this stage, the variability of clinical presentations of individuals with musculoskeletal disorders may be in part be related to different sources of pain, although individual-specific adaptations have also been shown for pain standardized in a single location (Hodges et al., 2013). Future studies should find new methodologies to identify which tissue(s) mainly contribute to pain in musculoskeletal conditions, and analyze the relation between the source of pain and the clinical presentation.  130  The maintenance of altered motor strategies even in the absence of nociceptive input is of potential interest in rehabilitation. This is consistent with clinical observations that altered motor strategies are maintained even after symptom remission in PFP (Bennell et al., 2010). Importantly, this also suggests that treatments deemed “effective” because they reduce pain may not be effective in resolving altered neuromuscular activation patterns. Future studies should determine whether the restoration of optimal motor strategies results in better long-term clinical outcomes in the rehabilitation of musculoskeletal disorders, limiting re-occurrence, progression to osteoarthritis and the development of musculoskeletal complaints in other joints. 6.7.4 Unbalanced vasti activation in patellofemoral pain: clinical implications Chapter 4 confirmed the existence of unbalanced vasti activation in PFP. Future studies should investigate whether and how altered neuromuscular activation is associated with PF kinetics and kinematics, and whether it is a potential target for rehabilitative interventions. In line with studies that identified an association between VM/VL activation and patellar position (Pal et al., 2011, 2012) and an association between patellar orientation and early signs of degeneration of the lateral patellar facet (Thuillier et al., 2013), it could be interesting to determine whether the females with PFP and the lowest redistribution between VM and VL are more likely to develop PF osteoarthritis. The inverse association between VM/VL redistribution and KES showed that low KES is not necessarily indicative of unbalanced vasti activation, but rather that females with PFP tended to show either one or the other. Future intervention studies should consider classifying females with PFP and high or low KES, and see whether dysfunction-specific treatment results in better clinical outcomes. 131  6.8 Conclusions This thesis provides novel insight into the role of regional vasti activation in clinical and experimental pain. Regionalized stretch reflexes indicate that the human spinal cord has the neuromuscular circuitry to preferentially activate motoneurones innervating muscle fibres located in different regions of the VM, which is a necessary prerequisite for regional activation. Redistribution of activation between VM and VL was observed in dynamic contractions for controls, but to a lesser extent in females with PFP, especially those with higher KES. Acute experimental pain resulted in location-specific adaptations, including the redistribution of regional activation, which persisted after pain resolution. This thesis supports the importance of within- and between-muscle redistribution of activation in the adaptation to pain and in musculoskeletal disorders, as described in the foundational “Pain Adaptation” model (Hodges & Tucker, 2011). Future research should focus on understanding whether the differences in regional activation in response to pain are associated with different biomechanics and are a potential target for clinically-effective rehabilitative interventions. 132  References Altenburg TM, Haan A De, Verdijk PWL, Mechelen W Van, De Ruiter CJ (2009). Vastus lateralis single motor unit EMG at the same absolute torque production at different knee angles. J Appl Physiol 107, 80–89. Amin S, Baker K, Niu J, Clancy M, Goggins J, Guermazi A, Grigoryan M, Hunter DJ & Felson DT (2009). Quadriceps strength and the risk of cartilage loss and symptom progression in knee osteoarthritis. Arthritis Rheum 60, 189–198. Armstrong T & Bull F (2006). Development of the World Health Organization Global Physical Activity Questionnaire (GPAQ). J Public Health 14, 66–70. Barbero M, Falla D & Cescon C (2016). The Location of Peak Upper Trapezius Muscle Activity During Submaximal Contractions is Not Associated with the Location of Myofascial Trigger Points. Clin J Pain 32, 1044-1052. Barbero M, Merletti R & Rainoldi A (2012). Atlas of Muscle Innervation Zones: understanding surface electromyography and its applications. New York: Springer Milan. Beck TW, Housh TJ, Cramer JT, Malek MH, Mielke M, Hendrix R & Weir JP (2008). Electrode shift and normalization reduce the innervation zone’s influence on EMG. Med Sci Sports Exerc 40, 1314–1322. Beck TW, Housh TJ, Cramer JT & Weir JP (2007). The effect of the estimated innervation zone on EMG amplitude and center frequency. Med Sci Sports Exerc 39, 1282–1290. Becker I, Baxter GD & Woodley SJ (2010). The vastus lateralis muscle: an anatomical investigation. Clin Anat 23, 575–585. Becker I, Woodley SJ & Baxter GD (2009). Gross morphology of the vastus lateralis muscle: An anatomical review. Clin Anat 22, 436–450. 133  Bennell K, Duncan M, Cowan S, McConnell J, Hodges P & Crossley K (2010). Effects of vastus medialis oblique retraining versus general quadriceps strengthening on vasti onset. Med Sci Sports Exerc 42, 856–864. Bennell K, Hodges P, Mellor R, Bexander C & Souvlis T (2004). The nature of anterior knee pain following injection of hypertonic saline into the infrapatellar fat pad. J Orthop Res 22, 116–121. Bilotto G, Schor RH, Uchino Y & Wilson VJ (1982). Localization of proprioceptive reflexes in the splenius muscle of the cat. Brain Res 238, 217–221. Birch L, Christensen H, Arendt-Nielsen L, Graven-Nielsen T & Søgaard K (2000). The influence of experimental muscle pain on motor unit activity during low-level contraction. Eur J Appl Physiol 83, 200–206. Bizzi E & Cheung VCK (2013). The neural origin of muscle synergies. Front Comput Neurosci 7, 51. Blazevich AJ, Gill ND & Zhou S (2006). Intra- and intermuscular variation in human quadriceps femoris architecture assessed in vivo. J Anat 209, 289–310. Blønd L & Hansen L (1998). Patellofemoral pain syndrome in athletes: A 5.7-Year retrospective follow-up study of 250 athletes. Acta Orthop Belg 64, 393–400. Blouin J-S, Brown HJ, Héroux ME, Inglis TJ & Siegmund GP (2016). Reply from J. S. Blouin, H. J. Brown, M. E. Héroux, J. T. Inglis and G. P. Siegmund. J Physiol 594, 1955. Boling M, Padua D, Marshall S, Guskiewicz K, Pyne S & Beutler A (2010). Gender differences in the incidence and prevalence of patellofemoral pain syndrome. Scand J Med Sci Sports 20, 725–730. Botter A, Oprandi G, Lanfranco F, Allasia S, Maffiuletti N a & Minetto MA (2011). Atlas of the 134  muscle motor points for the lower limb: implications for electrical stimulation procedures and electrode positioning. Eur J Appl Physiol 111, 2461–2471. Brody I & Rozear M (1970). Contraction response to muscle percussion. Arch Neurol 23, 259–265. Buchtal F, Erminio F & Rosenfalk P (1959). Motor Unit Territory in Different Human Muscles. Acta Physiol Scand 45, 72–87. Butler JE & Gandevia SC (2008). The output from human inspiratory motoneurone pools. J Physiol 586, 1257–1264. Cabral H V., de Souza LML, Mello RGT, Gallina A, de Oliveira LF & Vieira TM (2017). Is the firing rate of motor units in different vastus medialis regions modulated similarly during isometric contractions? Muscle Nerve; in press. Cameron WE, Binder MD, Botterman BR, Reinking RM & Stuart DG (1981). “Sensory partitioning” of cat medial gastrocnemius muscle by its muscle spindles and tendon organs. J Neurophysiol 46, 32–47. Castroflorio T, Falla D, Wang K, Svensson P & Farina D (2012). Effect of experimental jaw-muscle pain on the spatial distribution of surface EMG activity of the human masseter muscle during tooth clenching. J Oral Rehabil 39, 81–92. Cavazzuti L, Merlo A., Orlandi F & Campanini I (2010). Delayed onset of electromyographic activity of vastus medialis obliquus relative to vastus lateralis in subjects with patellofemoral pain syndrome. Gait Posture 32, 290–295. Chen YJ & Powers CM (2014). Comparison of Three Dimensional Patellofemoral Joint Reaction Forces in Persons With and Without Patellofemoral Pain. J Appl Biomech 30, 493–500. Chester R, Smith TO, Sweeting D, Dixon J, Wood S & Song F (2008). The relative timing of 135  VMO and VL in the aetiology of anterior knee pain: a systematic review and meta-analysis. BMC Musculoskelet Disord 9, 64. Christou EA, Shinohara M & Enoka RM (2003). Fluctuations in acceleration during voluntary contractions lead to greater impairment of movement accuracy in old adults. J Appl Physiol 95, 373–384. Cohen L (1953). Localization of stretch reflex. J Neurophysiol 16, 272–285. Collins N, Crossley K, Beller E, Darnell R, McPoil T & Vicenzino B (2009). Foot orthoses and physiotherapy in the treatment of patellofemoral pain syndrome: randomised clinical trial. Br J Sports Med 43, 169–171. Collins NJ, Bisset LM, Crossley KM & Vicenzino B (2012). Efficacy of nonsurgical interventions for anterior knee pain: systematic review and meta-analysis of randomized trials. Sport Med 42, 31–49. Cook C, Mabry L, Reiman MP & Hegedus EJ (2012). Best tests/clinical findings for screening and diagnosis of patellofemoral pain syndrome: a systematic review. Physiotherapy 98, 93–100. Cowan SM, Bennell KL & Hodges PW (2002). Therapeutic patellar taping changes the timing of vasti muscle activation in people with patellofemoral pain syndrome. Clin J Sport Med 12, 339–347. Cowan SM, Bennell KL, Hodges PW, Crossley KM & McConnell J (2001). Delayed onset of electromyographic activity of vastus medialis obliquus relative to vastus lateralis in subjects with patellofemoral pain syndrome. Arch Phys Med Rehabil 82, 183–189. Cronin NJ, Kumpulainen S, Joutjärvi T, Finni T & Piitulainen H (2015). Spatial variability of muscle activity during human walking: The effects of different EMG normalization 136  approaches. Neuroscience 300, 19–28. Crossley K (2014). Is patellofemoral osteoarthritis a common sequela of patellofemoral pain? Br J Sports Med 48, 409–411. Crossley KM, Stefanik JJ, Selfe J, Collins NJ, Davis IS, Powers CM, McConnell J, Vicenzino B, Bazett-Jones DM, Esculier J-F, Morrissey D & Callaghan MJ (2016). 2016 Patellofemoral pain consensus statement from the 4th International Patellofemoral Pain Research Retreat, Manchester. Part 1: Terminology, definitions, clinical examination, natural history, patellofemoral osteoarthritis and patient-reported outcome. Br J Sports Med 50, 839–843. d’Avella A, Saltiel P & Bizzi E (2003). Combinations of muscle synergies in the construction of a natural motor behavior. Nat Neurosci 6, 300–308. De Luca CJ & Erim Z (1994). Common drive of motor units in regulation of muscle force. Trends Neurosci 17, 299–305. De Luca CJ & Erim Z (2002). Common drive in motor units of a synergistic muscle pair. J Neurophysiol 87, 2200–2204. De Luca CJ, Gonzalez-Cueto JA, Bonato P & Adam A (2009). Motor unit recruitment and proprioceptive feedback decrease the common drive. J Neurophysiol 101, 1620–1628. De Luca CJ & Merletti R (1988). Surface myoelectric signal cross-talk among muscles of the leg. Electroencephalogr Clin Neurophysiol 69, 568–575. Desmedt HE & Godaux E (1981). Spinal motoneuron recruitment in man: rank deordering with direction but not with speed of voluntary movement. Science 214, 933–936. Devereaux MD & Lachmann SM (1984). Patello-femoral arthralgia in athletes attending a Sports Injury Clinic. Br J Sports Med 18, 18–21. Diamond LE, Van den Hoorn W, Bennell KL, Wrigley T V., Hinman RS, O’Donnell J & 137  Hodges PW (2017). Coordination of deep hip muscle activity is altered in symptomatic femoroacetabular impingement. J Orthop Res 35, 1494-1504. Dideriksen JL, Holobar A & Falla D (2016). Preferential distribution of nociceptive input to motoneurons with muscle units in the cranial portion of the upper trapezius muscle. J Neurophysiol 116, 611–618. Elias JJ, Kilambi S, Goerke DR & Cosgarea AJ (2009). Improving vastus medialis obliquus function reduces pressure applied to lateral patellofemoral cartilage. J Orthop Res 27, 578–583. Eng JJ & Hoffer JA (1997). Regional variability of stretch reflex amplitude in the cat medial gastrocnemius muscle during a postural task. J Neurophysiol 78, 1150–1154. Falla D, Arendt-Nielsen L & Farina D (2009). The pain-induced change in relative activation of upper trapezius muscle regions is independent of the site of noxious stimulation. Clin Neurophysiol 120, 150–157. Falla D & Farina D (2005). Muscle fiber conduction velocity of the upper trapezius muscle during dynamic contraction of the upper limb in patients with chronic neck pain. Pain 116, 138–145. Falla D, Gizzi L, Tschapek M, Erlenwein J & Petzke F (2014). Reduced task-induced variations in the distribution of activity across back muscle regions in individuals with low back pain. Pain 155, 944–953. Farina D (2006). Interpretation of the surface electromyogram in dynamic contractions. Exerc Sport Sci Rev 34, 121–127. Farina D, Arendt-Nielsen L, Merletti R & Graven-Nielsen T (2002a). Assessment of single motor unit conduction velocity during sustained contractions of the tibialis anterior muscle 138  with advanced spike triggered averaging. J Neurosci Methods 115, 1–12. Farina D, Arendt-Nielsen L, Merletti R & Graven-Nielsen T (2003). Effect of Experimental Muscle Pain on Motor Unit Firing Rate and Conduction Velocity. J Neurophysiol 91, 1250–1259. Farina D, Merletti R, Indino B, Nazzaro M & Pozzo M (2002b). Surface EMG crosstalk between knee extensor muscles: experimental and model results. Muscle Nerve 26, 681–695. Farina D, Merletti R, Nazzaro M & Caruso I (2001). Effect of joint angle on EMG variables in leg and thigh muscles. IEEE Eng Med Biol Mag 20, 62–71. Gallina A, Render J, Santos J, Shah H, Taylor D, Tomlin T, Garland SJ (2018a). Influence of knee joint position and sex on vastus medialis regional architecture. Appl Physiol Nutr Metab, accepted. Gallina A, Garland SJ, Wakeling JM. Local fluctuations of the EMG envelope identified using factorization algorithms reveal regional activation within the vastus medialis. J Electromyogr Kinesiol, submitted. Gallina A, Hunt MA, Hodges PW, Garland SJ (2018b). Vastus lateralis motor unit firing rate is higher in women with patellofemoral pain. Arch Phys Med Rehab, accepted. Gallina A, Blouin J-S, Ivanova TD & Garland SJ (2017). Regionalization of the stretch reflex in the human vastus medialis. J Physiol 595, 4991–5001. Gallina A & Botter A (2013). Spatial localization of electromyographic amplitude distributions associated to the activation of dorsal forearm muscles. Front Physiol 4, 1–8. Gallina A & Gazzoni M (2013). Selective activation of muscle sub-portions within the vastus medialis: effect of gender, knee angle and force level. Ital J Physiother 3, 20–29. Gallina A, Gazzoni M, Falla D & Merletti R (2016a). Surface EMG biofeedback, in Surface 139  Electromyography: Physiology, Engineering, and Applications. Wiley, pp 488-500. Gallina A, Ivanova TD & Garland SJ (2016b). Regional activation within the vastus medialis in stimulated and voluntary contractions. J Appl Physiol 121, 466–474. Gallina A, Merletti R & Gazzoni M (2013a). Innervation zone of the vastus medialis muscle: position and effect on surface EMG variables. Physiol Meas 34, 1411–1422. Gallina A, Merletti R & Gazzoni M (2013b). Uneven spatial distribution of surface EMG: what does it mean? Eur J Appl Physiol 113, 887–894. Gallina A, Merletti R & Vieira TMM (2011). Are the myoelectric manifestations of fatigue distributed regionally in the human medial gastrocnemius muscle? J Electromyogr Kinesiol 21, 929–938. Gallina A, Pollock CL, Vieira TM, Ivanova TD & Garland SJ (2016c). Between-day reliability of triceps surae responses to standing perturbations in people post-stroke and healthy controls: A high-density surface EMG investigation. Gait Posture 44, 103–109. Gallina A & Vieira T (2015). Territory and fiber orientation of vastus medialis motor units: a surface electromyography investigation. Muscle Nerve 52, 1057-65. Gandevia SC, Hudson AL, Gorman RB, Butler JE & De Troyer A (2006). Spatial distribution of inspiratory drive to the parasternal intercostal muscles in humans. J Physiol 573, 263–275. Gazzoni M, Celadon N, Mastrapasqua D, Paleari M, Margaria V & Ariano P (2014). Quantifying Forearm Muscle Activity during Wrist and Finger Movements by Means of Multi-Channel Electromyography. PLoS One 9, e109943. Giles LS, Webster KE, McClelland JA & Cook J (2013). Does quadriceps atrophy exist in individuals with patellofemoral pain? A systematic literature review with meta-analysis. J Orthop Sports Phys Ther 43, 766–776. 140  Gizzi L, Muceli S, Petzke F & Falla D (2015). Experimental Muscle Pain Impairs the Synergistic Modular Control of Neck Muscles. PLoS One 10, e0137844. Gootzen T, Vingerhoets D & Stegeman DF (1992). A study of motor unit structure by means of scanning EMG. Muscle Nerve 15, 349–357. Graven-Nielsen T, McArdle A, Phoenix J, Arendt-Nielsen L, Jensen TS, Jackson MJ & Edwards RHT (1997). In vivo model of muscle pain: Quantification of intramuscular chemical, electrical, and pressure changes associated with saline-induced muscle pain in humans. Pain 69, 137–143. ter Haar Romeny BM, Denier van der Gon JJ & Gielen CC (1982). Changes in recruitment order of motor units in the human biceps muscle. Exp Neurol 78, 360–368. Harris AJ, Duxson MJ, Butler JE, Hodges PW, Taylor JL & Gandevia SC (2005). Muscle Fiber and Motor Unit Behavior in the Longest Human Skeletal Muscle. J Neurosci 25, 8528–8533. Heckman CJ & Enoka RM (2012). Motor unit. Compr Physiol 2, 2629–2682. Hedayatpour N & Falla D (2013). Delayed onset of vastii muscle activity in response to rapid postural perturbations following eccentric exercise: a mechanism that underpins knee pain after eccentric exercise? Br J Sports Med 48, 429-434. Hedayatpour N, Falla D, Arendt-Nielsen L & Farina D (2008). Sensory and electromyographic mapping during delayed-onset muscle soreness. Med Sci Sports Exerc 40, 326–334. Hedayatpour N, Falla D, Arendt-Nielsen L, Vila-Chã C & Farina D (2009). Motor unit conduction velocity during sustained contraction after eccentric exercise. Med Sci Sports Exerc 41, 1927–1933. Henneman E (1957). Relation between size of neuron and their susceptibility to discharge. 141  Science 126, 1345–1347. Héroux ME (2017). Tap, tap, who’s there? It’s localized muscle activity elicited by the human stretch reflex. J Physiol 14, 4575. Héroux ME, Brown HJ, Inglis JT, Siegmund GP & Blouin J-S (2015). Motor units in the human medial gastrocnemius muscle are not spatially localized or functionally grouped. J Physiol 593, 3711–3726. Herrington L & Al-Sherhi A (2007). A controlled trial of weight-bearing versus non-weight-bearing exercises for patellofemoral pain. J Orthop Sports Phys Ther 37, 155–160. Herrmann U & Flanders M (1998). Directional tuning of single motor units. J Neurosci 18, 8402–8416. Hinman RS, Bennell KL, Metcalf BR & Crossley KM (2002). Temporal activity of vastus medialis obliquus and vastus lateralis in symptomatic knee osteoarthritis. Am J Phys Med Rehabil 81, 684–690. Hodges PW (2011). Pain and motor control: From the laboratory to rehabilitation. J Electromyogr Kinesiol 21, 220–228. Hodges PW, Coppieters MW, MacDonald D & Cholewicki J (2013). New insight into motor adaptation to pain revealed by a combination of modelling and empirical approaches. Eur J Pain 17, 1138–1146. Hodges PW, Ervilha UF & Graven-Nielsen T (2008). Changes in motor unit firing rate in synergist muscles cannot explain the maintenance of force during constant force painful contractions. J Pain 9, 1169–1174. Hodges PW, Mellor R, Crossley K & Bennell K (2009). Pain induced by injection of hypertonic saline into the infrapatellar fat pad and effect on coordination of the quadriceps muscles. 142  Arthritis Rheum 61, 70–77. Hodges PW & Richardson CA (1999). Altered trunk muscle recruitment in people with low back pain with upper limb movement at different speeds. Arch Phys Med Rehabil 80, 1005–1012. Hodges PW & Smeets RJ (2014). Interaction Between Pain, Movement and Physical Activity: Short-term Benefits, Long-term Consequences, and Targets for Treatment. Clin J Pain 31, 97-107. Hodges PW & Tucker K (2011). Moving differently in pain: A new theory to explain the adaptation to pain. Pain 152, S90–S98. Holt G, Nunn T, Allen RA, Forrester AW & Gregori A (2008). Variation of the Vastus Medialis Obliquus Insertion and its Relevance to Minimally Invasive Total Knee Arthroplasty. J Arthroplasty 23, 600–604. Hu X, Suresh NL, Xue C & Rymer WZ (2015). Extracting extensor digitorum communis activation patterns using high-density surface electromyography. Front Physiol 6, 1–9. Huang C, Chen X, Cao S & Zhang X (2015). Muscle-tendon units localization and activation level analysis based on high-density surface EMG array and NMF algorithm. J Neural Eng 13, 1–9. Hudson AL, Taylor JL, Gandevia SC & Butler JE (2009). Coupling between mechanical and neural behaviour in the human first dorsal interosseous muscle. J Physiol 587, 917–925. Hug F, Hodges PW, Hoorn WVD & Tucker K (2014a). Between-muscle differences in the adaptation to experimental pain. J Appl Physiol 117, 1132–1140. Hug F, Hodges PW & Tucker K (2014b). Task dependency of motor adaptations to an acute noxious stimulation. J Neurophysiol 111, 2298–2306. Hug F, Hodges PW & Tucker KJ (2013). Effect of pain location on spatial reorganisation of 143  muscle activity. J Electromyogr Kinesiol 23, 1413–1420. Ireland ML, Willson JD, Ballantyne BT & Davis IM (2003). Hip Strength in Females With and Without Patellofemoral Pain. J Orthop Sport Phys Ther 33, 671–676. Jolliffe IT (1986). Principal Component Analysis and Factor Analysis, in Principal Component Analysis. Springer Series in Statistics. Springer, New York, NY. Kannus P, Natri A, Paakkala T & Jarvinen M (1999). An Outcome Study of Chronic Patellofemoral Pain Syndrome. Seven-Year Follow-up of Patients in a Randomized, Controlled Trial. J Bone Jt Surg 81, 355–363. Karst GM & Willett GM (1995). Onset timing of electromyographic activity in the vastus medialis oblique and vastus lateralis muscles in subjects with and without patellofemoral pain syndrome. Phys Ther 75, 813–823. Keenan KG, Farina D, Maluf KS, Merletti R & Enoka RM (2004). Influence of amplitude cancellation on the simulated surface electromyogram. J Appl Physiol 98, 120–131. Khayambashi K, Mohammadkhani Z, Ghaznavi K, Lyle MA & Powers CM (2012). The effects of isolated hip abductor and external rotator muscle strengthening on pain, health status, and hip strength in females with patellofemoral pain: a randomized controlled trial. J Orthop Sports Phys Ther 42, 22–29. Kouzaki M, Shinohara M (2006). The frequency of alternate muscle activity is associated with the attenuation in muscle fatigue. J Appl Physiol 101, 715–720. Kujala UM, Jaakkola LH, Koskinen SK, Taimela S, Hurme M & Nelimarkka O (1993). Scoring of patellofemoral disorders. J Arthrosc Relat Surg 9, 159–163. Laija W (1997). Conducting ANOVA Trend Analyses Using Polynomial Contrasts. Annu Meet Southwest Educ Res Assoc Austin, January 23-25. 144  Laine CM, Martinez-valdes E, Falla D, Mayer F & Farina D (2015). Motor Neuron Pools of Synergistic Thigh Muscles Share Most of Their Synaptic Input. J Neurosci 35, 12207–12216. Lankhorst NE, Bierma-Zeinstra SM & van Middelkoop M (2013). Factors associated with patellofemoral pain syndrome: a systematic review. Br J Sport Med 47, 193–206. Lankhorst NE, Bierma-Zeinstra SM & van Middelkoop M (2012). Risk factors for patellofemoral pain syndrome: a systematic review. J Orthop Sports Phys Ther 42, 81–94. Lankhorst NE, van Middelkoop M, Crossley KM, Bierma-Zeinstra SMA, Oei EHG, Vicenzino B & Collins NJ (2015). Factors that predict a poor outcome 5–8 years after the diagnosis of patellofemoral pain: a multicentre observational analysis. Br J Sports 50, 881-886. Lapatki BG, Oostenveld R, Van Dijk JP, Jonas IE, Zwarts MJ & Stegeman DF (2006). Topographical characteristics of motor units of the lower facial musculature revealed by means of high-density surface EMG. J Neurophysiol 95, 342–354. Le Pera D, Graven-Nielsen T, Valeriani M, Oliviero A, Di Lazzaro V, Tonali PA & Arendt-Nielsen L (2001). Inhibition of motor system excitability at cortical and spinal level by tonic muscle pain. Clin Neurophysiol 112, 1633–1641. Lee DD & Seung HS (1999). Learning the parts of objects by non-negative matrix factorization. Nature 401, 788–791. Lefebvre R, Leroux A, Poumarat G, Galtier B, Guillot M, Vanneuville G & Boucher JP (2006). Vastus Medialis: Anatomical and Functional Considerations and Implications Based Upon Human and Cadaveric Studies. J Manipulative Physiol Ther 29, 139-144. Lin F, Wang G, Koh JL, Hendrix RW & Zhang L (2004). In vivo and Noninvasive Three-Dimensional Patellar Tracking Induced by Individual Heads of Quadriceps. Med Sci Sport 145  Exerc 36, 93–101. Lorenz A, Müller O, Kohler P, Wünschel M, Wülker N & Leichtle UG (2012). The influence of asymmetric quadriceps loading on patellar tracking - An in vitro study. Knee 19, 818–822. Lund JP, Donga R, Widmer CG & Stohler CS (1991). The pain-adaptation model: a discussion of the relationship between chronic musculoskeletal pain and motor activity. Can J Physiol Pharmacol 69, 683–694. Madeleine P, Leclerc F, Arendt-Nielsen L, Ravier P & Farina D (2006). Experimental muscle pain changes the spatial distribution of upper trapezius muscle activity during sustained contraction. Clin Neurophysiol 117, 2436–2445. Magistris MR & Kohler A (1996). Contraction response to muscle percussion is increased in peripheral nerve conduction block. Neurology 47, 1243–1246. McCarthy Persson U, Fleming HF & Caulfield B (2009). The effect of a vastus lateralis tape on muscle activity during stair climbing. Man Ther 14, 330–337. McConnell J (1986). The management of chondromalacia patellae: a long term solution. Aust J Physiother 32, 215–223. McKeon B, Gandevia S & Burke D (1984). Absence of somatotopic projection of muscle afferents onto motoneurons of same muscle. J Neurophysiol 51, 185–194. McNeil CJ, Butler JE, Taylor JL & Gandevia SC (2013). Testing the excitability of human motoneurons. Front Hum Neurosci 7, 152. McPoil T, Warren M, Vicenzino B & Cornwall MW (2011). Variations in foot posture and mobility between individuals with patellofemoral pain and those in a control group. J Am Podiatr Med Assoc 101, 289–296. McPoil TG, Vicenzino B, Cornwall MW, Collins N & Warren M (2009). Reliability and 146  normative values for the foot mobility magnitude: a composite measure of vertical and medial-lateral mobility of the midfoot. J Foot Ankle Res 2, 6. Mellor R & Hodges P (2005). Motor unit synchronization is reduced in anterior knee pain. J Pain 6, 550–558. Merletti R, Botter A, Cescon C, Minetto MA & Vieira TMM (2010). Advances in surface EMG: recent progress in clinical research applications. Crit Rev Biom Eng 38, 347–379. Merletti R, Farina D & Gazzoni M (2003). The linear electrode array: a useful tool with many applications. J Electromyogr Kinesiol 13, 37–47. Merletti R, Rainoldi A & Farina D (2001). Surface electromyography for noninvasive characterization of muscle. Exerc Sport Sci Rev 29, 20–25. Muceli S, Falla D & Farina D (2014). Reorganization of muscle synergies during multidirectional reaching in the horizontal plane with experimental muscle pain. J Neurophysiol 111, 1615–1630. Muceli S, Jiang N & Farina D (2013). Extracting Signals Robust to Electrode Number and Shift for Online Simultaneous and Proportional Myoelectric Control by Factorization Algorithms. IEEE Trans Neural Syst Rehabil Eng 22, 623-633. Murray GM & Peck CC (2007). Orofacial pain and jaw muscle activity: a new model. J Orofac Pain 21, 263-278-288. Noehren B, Hamill J & Davis I (2013). Prospective evidence for a hip etiology in patellofemoral pain. Med Sci Sports Exerc 45, 1120–1124. Nunes GS, Stapait EL, Kirsten MH, de Noronha M & Santos GM (2013). Clinical test for diagnosis of patellofemoral pain syndrome: Systematic review with meta-analysis. Phys Ther Sport 14, 54–59. 147  O’Brien TD, Reeves ND, Baltzopoulos V, Jones DA & Maganaris CN (2010). Muscle-tendon structure and dimensions in adults and children. J Anat 216, 631–642. Owings TM & Grabiner MD (2002). Motor control of the vastus medialis oblique and vastus lateralis muscles is disrupted during eccentric contractions in subjects with patellofemoral pain. Am J Sports Med 30, 483–487. Pal S, Besier TF, Draper CE, Fredericson M, Gold GE, Beaupre GS & Delp SL (2012). Patellar tilt correlates with vastus lateralis: vastus medialis activation ratio in maltracking patellofemoral pain patients. J Orthop Res 30, 927–933. Pal S, Draper C & Fredericson M (2011). Patellar maltracking correlates with vastus medialis activation delay in patellofemoral pain patients. Am J Sport Med 39, 590-598. Pattyn E, Verdonk P, Steyaert A, Vanden Bossche L, Van den Broecke W, Thijs Y & Witvrouw E (2011). Vastus medialis obliquus atrophy: does it exist in patellofemoral pain syndrome? Am J Sports Med 39, 1450–1455. Pattyn E, Verdonk P, Steyaert A, Van Tiggelen D & Witvrouw E (2013). Muscle functional MRI to evaluate quadriceps dysfunction in patellofemoral pain. Med Sci Sports Exerc 45, 1023–1029. Peeler J, Cooper J, Porter MM, Thliveris JA & Anderson JE (2005). Structural parameters of the vastus medialis muscle. Clin Anat 18, 281–289. Piitulainen H, Rantalainen T, Linnamo V, Komi P & Avela J (2009). Innervation zone shift at different levels of isometric contraction in the biceps brachii muscle. J Electromyogr Kinesiol 19, 667–675. Poortvliet PC, Tucker KJ & Hodges PW (2014). Experimental pain has a greater effect on single motor unit discharge during force-control than position-control tasks. Clin Neurophysiol 148  126, 1378-1386. Popovic D, Gordon T, Rafuse VF & Prochazka A (1991). Properties of implanted electrodes for functional electrical stimulation. Ann Biomed Eng 19, 303–316. Rainoldi A, Melchiorri G & Caruso I (2004). A method for positioning electrodes during surface EMG recordings in lower limb muscles. J Neurosci Methods 134, 37–43. Rainoldi A, Nazzaro M, Merletti R, Farina D, Caruso I & Gaudenti S (2000). Geometrical factors in surface EMG of the vastus medialis and lateralis muscles. J Electromyogr Kinesiol 10, 327–336. Rasool G, Afsharipour B, Suresh NL, Hu X & Rymer WZ (2015). Spatial analysis of muscular activations in stroke survivors. Proc Annu Int Conf IEEE Eng Med Biol Soc EMBS, pp. 6058–6061. Rathleff MS, Rathleff CR, Crossley KM & Barton CJ (2014). Is hip strength a risk factor for patellofemoral pain? A systematic review and meta-analysis. Br J Sports Med 48, 1088-1100. Riek S & Bawa P (1992). Recruitment of motor units in human forearm extensors. J Neurophysiol 68, 100–108. Rodriguez-Falces J, Negro F, Gonzalez-Izal M & Farina D (2013). Spatial distribution of surface action potentials generated by individual motor units in the human biceps brachii muscle. J Electromyogr Kinesiol 23, 766–777. Roeleveld K, Stegeman DF, Vingerhoets HM & Van Oosterom A (1997a). The motor unit potential distribution over the skin surface and its use in estimating the motor unit location. Acta Physiol Scand 161, 465–472. Roeleveld K, Stegeman DF, Vingerhoets HM & Van Oosterom A (1997b). Motor unit potential 149  contribution to surface electromyography. Acta Physiol Scand 160, 175–183. Rojas-Martínez M, Mañanas MA & Alonso JF (2012). High-density surface EMG maps from upper-arm and forearm muscles. J Neuroeng Rehabil 9, 85. Roland MO (1986). A critical review of the evidence for a pain-spasm-pain cycle in spinal disorders. Clin Biomech 1, 102–109. Salomoni S, Tucker K, Hug F, McPhee M & Hodges P (2016). Reduced Maximal Force during Acute Anterior Knee Pain Is Associated with Deficits in Voluntary Muscle Activation. PLoS One 11, e0161487. Salomoni SE, Ejaz A, Laursen AC & Graven-Nielsen T (2013). Variability of three-dimensional forces increase during experimental knee pain. Eur J Appl Physiol 113, 567–575. Schabrun SM, Burns E & Hodges PW (2015a). New insight into the time-course of motor and sensory system changes in pain. PLoS One 10, 1–14. Schabrun SM, Christensen SW, Mrachacz-Kersting N & Graven-Nielsen T (2015b). Motor Cortex Reorganization and Impaired Function in the Transition to Sustained Muscle Pain. Cereb Cortex 26, 1878–1890. Schabrun SM, Hodges PW, Vicenzino B, Jones E & Chipchase LS (2014). Novel Adaptations in Motor Cortical Maps. Med Sci Sports Exerc 47, 681-690. Schulte E, Dimitrova NA, Dimitrov GV, Rau G & Disselhorst-Klug C (2005). Estimation of the muscle fibre semi-length under varying joint positions on the basis of non-invasively extracted motor unit action potentials. J Electromyogr Kinesiol 15, 290–299. Selfe J, Callaghan M, Witvrouw E, Richards J, Dey MP, Sutton C, Dixon J, Martin D, Stokes M, Janssen J, Ritchie E & Turner D (2013). Targeted interventions for patellofemoral pain syndrome (TIPPS): classification of clinical subgroups. BMJ Open 3, e003795. 150  Selfe J, Janssen J, Callaghan M, Witvrouw E, Sutton C, Richards J, Stokes M, Martin D, Dixon J, Hogarth R, Baltzopoulos V, Ritchie E, Arden N & Dey P (2016). Are there three main subgroups within the patellofemoral pain population? A detailed characterisation study of 127 patients to help develop targeted intervention (TIPPs). Br J Sports Med 50, 873–880. Sheehan FT, Borotikar BS, Behnam AJ & Alter KE (2012). Alterations in in vivo knee joint kinematics following a femoral nerve branch block of the vastus medialis: Implications for patellofemoral pain syndrome. Clin Biomech 27, 525–531. Smith TO, Bowyer D, Dixon J, Stephenson R, Chester R & Donell ST (2009a). Can vastus medialis oblique be preferentially activated? A systematic review of electromyographic studies. Physiother Theory Pract 25, 69–98. Smith TO, Nichols R & Harle D (2009b). Do the Vastus Medialis Obliquus and Vastus Medialis Longus Really Exist? A Systematic Review. Clin Anat 199, 183–199. Souza RB & Powers CM (2009). Differences in hip kinematics, muscle strength, and muscle activation between subjects with and without patellofemoral pain. J Orthop Sports Phys Ther 39, 12–19. Spencer JD, Hayes KC & Alexander IJ (1984). Knee joint effusion and quadriceps reflex inhibition in man. Arch Phys Med Rehabil 65, 171–177. Staudenmann D, Kingma I, Daffertshofer A, Stegeman DF & van Dieen JH (2009). Heterogeneity of muscle activation in relation to force direction: A multi-channel surface electromyography study on the triceps surae muscle. J Electromyogr Kinesiol 19, 882–895. Staudenmann D, Stegeman DF & van Dieën JH (2013). Redundancy or heterogeneity in the electric activity of the biceps brachii muscle? Added value of PCA-processed multi-channel EMG muscle activation estimates in a parallel-fibered muscle. J Electromyogr Kinesiol 23, 151  892–898. Stuart DG, Ha mm TM & Noven V (1988). Partitioning of monosynaptic Ia EPSP connections with motoneurons according to neuromuscular topography: generality and functional implications. Prog Neurobiol 30, 437-447. Suzuki H, Conwit RA, Stashuk D, Santarsiero L & Metter EJ (2002). Relationships between surface-detected EMG signals and motor unit activation. Med Sci Sports Exerc 34, 1509–1517. Svensson P, Miles TS, McKay D & Ridding MC (2003). Suppression of motor evoked potentials in a hand muscle following prolonged painful stimulation. Eur J Pain 7, 55–62. Taunton J, Ryan M, Clement D, McKenzie D, Lloyd-Smith D & Zumbo B (2002). A retrospective case-control analysis of 2002 running injuries. Br J Sports Med 36, 95–102. Taunton J, Ryan M, Clement D, McKenzie D, Lloyd-Smith D & Zumbo B (2003). A prospective study of running injuries: the Vancouver Sun Run “In Training” clinics. Br J Sports Med 37, 239–245. Te M, Baptista AF, Chipchase LS & Schabrun SM (2017). Primary Motor Cortex Organization Is Altered in Persistent Patellofemoral Pain. Pain Med 18, 2224-2234. Tenan MS, Hackney AC & Griffin L (2016). Entrainment of vastus medialis complex activity differs between genders. Muscle Nerve 53, 633–640. Tenan MS, Peng Y-L, Hackney AC & Griffin L (2013). Menstrual cycle mediates vastus medialis and vastus medialis oblique muscle activity. Med Sci Sports Exerc 45, 2151–2157. Testa M, Geri T, Gizzi L & Falla D (2017). High-density EMG Reveals Novel Evidence of Altered Masseter Muscle Activity During Symmetrical and Asymmetrical Bilateral Jaw Clenching Tasks in People With Chronic Nonspecific Neck Pain. Clin J Pain 33, 148–159. 152  Thuillier DU, Souza RB, Wu S, Luke A, Li X & Feeley BT (2013). T1r Imaging Demonstrates Early Changes in the Lateral Patella in Patients With Patellofemoral Pain and Maltracking. Am J Sports Med 41, 1813. Travell J, Rinzler S & Herman M (1942). Pain and disability of the shoulder and arm. JAMA 120, 417-422. Tsao H, Danneels LA & Hodges PW (2011a). Smudging the motor brain in young adults with recurrent low back pain. Spine 36, 1721–1727. Tsao H, Tucker KJ & Hodges PW (2011b). Changes in excitability of corticomotor inputs to the trunk muscles during experimentally-induced acute low back pain. Neuroscience 181, 127–133. Tucker K, Butler J, Graven-Nielsen T, Riek S & Hodges P (2009). Motor unit recruitment strategies are altered during deep-tissue pain. J Neurosci 29, 10820–10826. Tucker K, Hodges PW, Van Den Hoorn W, Nordez A & Hug F (2014a). Does stress within a muscle change in response to an acute noxious stimulus? PLoS One 9, e91899. Tucker K, Larsson AK, Oknelid S & Hodges P (2012). Similar alteration of motor unit recruitment strategies during the anticipation and experience of pain. Pain 153, 636–643. Tucker KJ, Fels M, Walker SR & Hodges PW (2014b). Comparison of location, depth, quality, and intensity of experimentally induced pain in 6 low back muscles. Clin J Pain 30, 800–808. Tucker KJ & Hodges PW (2009). Motoneurone recruitment is altered with pain induced in non-muscular tissue. Pain 141, 151–155. Tucker KJ & Hodges PW (2010). Changes in motor unit recruitment strategy during pain alters force direction. Eur J Pain 14, 932–938. 153  van den Hoorn W, Hodges PW, van Dieen JH, Hug F, van Dieën JH & Hug F (2015). Effect of acute noxious stimulation to the leg or back on muscle synergies during walking. J Neurophysiol 113, 224–254. Van Tiggelen D, Cowan S, Coorevits P, Duvigneaud N & Witvrouw E (2009). Delayed vastus medialis obliquus to vastus lateralis onset timing contributes to the development of patellofemoral pain in previously healthy men: a prospective study. Am J Sports Med 37, 1099–1105. Vieira T, Botter A, Minetto MA & Hodson-Tole EF (2015). Spatial variation of compound muscle action potentials across human gastrocnemius medialis. J Neurophysiol 114, 1617–1627. Vieira TM, Wakeling JM & Hodson-Tole EF (2016). Is there sufficient evidence to claim muscle units are not localised and functionally grouped within the human gastrocnemius? J Physiol 594, 1953–1954. Vieira TMM, Loram ID, Muceli S, Merletti R & Farina D (2011). Postural activation of the human medial gastrocnemius muscle: are the muscle units spatially localised? J Physiol 589, 431–443. Vieira TMM, Merletti R & Mesin L (2010a). Automatic segmentation of surface EMG images: Improving the estimation of neuromuscular activity. J Biomech 43, 2149–2158. Vieira TMM, Windhorst U & Merletti R (2010b). Is the stabilization of quiet upright stance in humans driven by synchronized modulations of the activity of medial and lateral gastrocnemius muscles? J Appl Physiol 108, 85–97. Watanabe K, Kouzaki M, Merletti R, Fujibayashi M & Moritani T (2012a). Spatial EMG potential distribution pattern of vastus lateralis muscle during isometric knee extension in 154  young and elderly men. J Electromyogr Kinesiol 22, 74–79. Watanabe K, Kouzaki M & Moritani T (2012b). Task-dependent spatial distribution of neural activation pattern in human rectus femoris muscle. J Electromyogr Kinesiol 22, 251–258. Watanabe K, Kouzaki M & Moritani T (2014a). Non-uniform surface EMG responses to change in joint angle within rectus femoris muscle. Muscle Nerve 50, 794–802. Watanabe K, Kouzaki M & Moritani T (2014b). Regional neuromuscular regulation within human rectus femoris muscle during gait. J Biomech 47, 3502-3508. Watanabe K, Kouzaki M & Moritani T (2015). Heterogeneous neuromuscular activation within human rectus femoris muscle during pedaling. Muscle Nerve 52, 404–411. Watanabe K, Miyamoto T, Tanaka Y, Fukuda K & Moritani T (2012c). Type 2 diabetes mellitus patients manifest characteristic spatial EMG potential distribution pattern during sustained isometric contraction. Diabetes Res Clin Pract 97, 468–473. Weinstabl R, Scharf W & Firbas W (1989). The extensor apparatus of the knee joint and its peripheral vasti: anatomic investigation and clinical relevance. Surg Radiol Anat 11, 17–22. Wilson NA, Press JM, Koh JL, Hendrix RW & Zhang L-Q (2009). In vivo noninvasive evaluation of abnormal patellar tracking during squatting in patients with patellofemoral pain. J Bone Joint Surg Am 91, 558–566. Wilson NA & Sheehan FT (2010). Dynamic in vivo quadriceps lines-of-action. J Biomech 43, 2106–2113. Windhorst U, Ha mm TM & Stuart DG (1989). On the function of muscle and reflex partitioning. Behav Brain Sci 12, 629–645. Witvrouw E, Callaghan MJ, Stefanik JJ, Noehren B, Bazett-jones DM, Willson JD, Earl-boehm JE, Davis IS, Powers CM, Mcconnell J & Crossley KM (2014). Patellofemoral pain: 155  consensus statement from the 3rd International Patellofemoral Pain Research Retreat held in Vancouver , September 2013. Br J Sports Med 48, 411-414. Witvrouw E, Danneels L, Van Tiggelen D, Willems TM & Cambier D (2004). Open versus closed kinetic chain exercises in patellofemoral pain: a 5-year prospective randomized study. Am J Sports Med 32, 1122–1130. Witvrouw E, Lysens R, Bellemans J, Cambier D & Guy V (2000). Intrinsic Risk Factors For the Development of Anterior Knee Pain in an Athletic Population. Am J Sports Med 28, 480–489. Wünschel M, Leichtle U & Obloh C (2011). The effect of different quadriceps loading patterns on tibiofemoral joint kinematics and patellofemoral contact pressure during simulated partial weight-bearing knee flexion. Knee Surgery, Sport, Traum, Arthrosc 19, 1099–1106. Zhou P, Suresh NL & Rymer WZ (2011). Surface electromyogram analysis of the direction of isometric torque generation by the first dorsal interosseous muscle. 8, 036028. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0364687/manifest

Comment

Related Items