UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Instrumentation and ultrasound imaging for epidural anesthesia Hor, King Wei 2007

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
831-ubc_2007-0118a.pdf [ 14.61MB ]
Metadata
JSON: 831-1.0101226.json
JSON-LD: 831-1.0101226-ld.json
RDF/XML (Pretty): 831-1.0101226-rdf.xml
RDF/JSON: 831-1.0101226-rdf.json
Turtle: 831-1.0101226-turtle.txt
N-Triples: 831-1.0101226-rdf-ntriples.txt
Original Record: 831-1.0101226-source.json
Full Text
831-1.0101226-fulltext.txt
Citation
831-1.0101226.ris

Full Text

INSTRUMENTATION AND ULTRASOUND IMAGING FOR EPIDURAL ANESTHESIA by K i n g Wei Hor B . A . S c , The University of British Columbia, 2002 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F M A S T E R OF A P P L I E D S C I E N C E in T H E F A C U L T Y OF G R A D U A T E S T U D I E S (Electrical & Computer Engineering) T H E U N I V E R S I T Y OF B R I T I S H C O L U M B I A March 2007 © King Wei Hor, 2007 ABSTRACT The loss-of-resistance technique in epidural anesthesia is the accepted standard for indicating the entry of the needle into the epidural space. In conventional epidurals, it is also the only feedback mechanism to confirm needle entry. Unsuccessful epidurals due to the technical difficulties can result in mild to severe complications. These difficulties include correctly choosing the puncture site and needle trajectory, which are determined solely by palpation and the experience of the anesthesiologist. Instrumentation of the thumb's force on the plunger of the syringe, displacement of the plunger and fluid pressure is developed for laboratory and clinical trials to study the dynamics of the loss-of-resistance technique. Instrumentation of the loss-of-resistance technique was performed on culled domestic pigs using standard epidural procedures. A static and decay model, based on physical properties and empirical data, are used for estimating the pressure from the force and displacement values. The decay model is shown to be reasonably accurate and allows the omission of the pressure sensor in clinical trials. Furthermore, the accuracy of decay model is further improved for the "smooth" protocol performed by the anesthesiologist, over the "bouncing" protocol. The loss-of-resistance, indicated orally by the anesthesiologist, is consistent with the rapid fall in all three measurements. The oral indication of the loss-of-resistance slightly lags that of the measured values and is consistent with the lag in oral communication. The instrumentation of the loss-of-resistance is further confirmed by direct and indirect measurements from ultrasound images of the epidural space and needle. However, obtaining good image quality is difficult due to the steep needle angle and the surrounding bone structures. A n adaptive spatial compounding algorithm is developed to improve important features such as the bone and epidural space. A specially constructed phantom with speed-of-sound distortion is used to compare several variations of the algorithm. The adaptive spatial compounding using median-based averaging produced image quality with the best balance for point resolution, edge resolution and noise reduction in homogeneous regions. In porcine studies, the technique shows visible improvements of the epidural space and surrounding features. 11 T A B L E OF CONTENTS Abstract • i i Table of Contents , i i i List of Figures v i List of Tables..... ix Acknowledgements x 1 Introduction ; 1 Epidural Anesthesia Overview 1 1.1 Motivation 3 1.2 Statement of Problem 7 1.3 Research Objective 8 2 Instrumentation and Modeling of the Loss-Of-Resistance Technique 9 2.1 Introduction : -9 2.2 Instrumentation 11 2.2.1 Sensors 11 2.2.2 Hardware and Software 13 2.2.3 Testing and Calibration 14 2.3 Modeling '. 14 2.3.1 Static Model 14 2.3.2 . Dynamic Model 16 2.3.3 Decay Model .' 18 2.4 Constant Force Experiment 19 2.4.1 Materials and Method ....19 2.4.2 Results and Discussion 20 in 2.5 Epidural Experiments 24 2.5.1 Materials and Method 24 2.5.2 Results and Discussion 28 2.6 Conclusion 37 3 Ultrasound Verification of the Epidural Needle Insertion 38 3.1 Introduction 38 3.2 Materials and Method 38 3.3 Results and Discussion 39 3.4 Conclusion 43 4 Adaptive Spatial Compounding 44 4.1 Introduction 44 4.2 Materials and Method 45 4.2.1 Hardware and Development Tools 45 4.2.2 Image Acquisition and Post-scan Conversion 45 4.2.3 Adaptive Spatial Compounding Method 46 4.2.3.1 Block-based Shift Vector Estimation for Warping 46 4.2.3.2 Interpolation of Shift Vectors 47 4.2.3.3 Image Remapping (Warping) 49 4.2.3.4 Compounding 49 4.2.3.5 Software Application 51 4.2.4 Experimental Verification 52 4.3 Results and Discussion 53 4.4 Conclusion 65 5 Conclusions and Future Work 66 iv 5.1 Conclusions 66 5.2 Future Work 67 References 69 Appendices 73 Appendix A : Instrumentation Circuits 73 Appendix B : Sensor Calibration 77 v LIST OF FIGURES Figure 1: A typical epidural needle 2 Figure 2: The loss-of-resistance technique 2 Figure 3: Physical interactions of the epidural needle 10 Figure 4: The SLB-25 force sensor 11 Figure 5: The PX302 pressure sensor : 12 Figure 6: The C S P R 1P65 displacement sensor 12 Figure 7: Force diagram of an angled plunger 15 Figure 8: Pressure profile of fluid in syringe using constant applied force 21 Figure 9: Displacement profile of plunger using constant applied force 21 Figure 10: Plot of resulting force on the fluid against the applied force 22 Figure 11: Pressure decay profiles of a glass epidural syringe...: 23 Figure 12: The instrumentation device 25 Figure 13: A domestic pig (Sus scrofa domestica) used for the epidural experiments 25 Figure 14: The use of the instrumentation device on the pig 26 Figure 15: Force profile of epidural procedure using "smooth" technique 28 Figure 16: Pressure profile of epidural procedure using "smooth" technique 29 Figure 17: Displacement profile of epidural procedure using "smooth" technique 29 Figure 18: Force profile of epidural procedure using "bouncing" technique 30 Figure 19: Pressure profile of epidural procedure using "bouncing" technique 30 Figure 20: Displacement profile of epidural procedure using "bouncing" technique 31 Figure 21: Comparison of estimated loss-of-resistance times 33 Figure 22: Static and decay modeling of pressure for "smooth" technique 34 Figure 23: Static and decay modeling of pressure for "bouncing" technique 34 vi Figure 24: Mean error comparison between the static and decay models 36 Figure 25: R M S error comparison between the static and decay models 36 Figure 26: Ultrasound imaging of the epidural needle . : 39 Figure 27: Measurement of the puncture path length between L 3 - L 4 vertebrae 40 Figure 28: Comparison of the puncture path lengths . 40 Figure 29: Direct measurement of the epidural space depth 41 Figure 30: Indirect measurement of the epidural space depth , 42 Figure 31: The concept of spatial compounding . 44 Figure 32: Adaptive spatial compounding application 51 Figure 33: Reference image of the phantom 54 Figure 34: Conventional spatial compounding of the phantom.. 54 Figure 35: Warped spatial compounding with averaging of the phantom 55 Figure 36: Adaptive compounding with gradient-based averaging of the phantom 55 Figure 37: Adaptive compounding with median-based averaging of the phantom...; 56 Figure 38: Selection of spots for evaluation of point features 57 Figure 39: Normalized diameter of spots '. .• 57 Figure 40: Selection of edges for evaluation of edge strength 58 Figure 41: Maximum intensity of edges 59 Figure 42: Selection of dark and light homogeneous regions for evaluation of noise 60 Figure 43: S N R and C N R of dark and light regions 60 Figure 44: Images of porcine spinal tissue using different compounding techniques 63 Figure 45: Images of porcine epidural space using different compounding techniques... 64 Figure 46: Instrumentation amplifier circuit diagram 73 Figure 47: ± 15V voltage regulator circuit diagram „ 74 vi i Figure 48: 5 V voltage regulator circuit diagram . 74 Figure 49: 10V voltage regulator circuit diagram 75 Figure 50: Circuit layout 76 Figure 51: Instrumentation circuit boards... 76 Figure 52: Wheatstone bridge circuit for amplifier calibration 77 Figure 53: Calibration plot of the amplifier for the force sensor 78 Figure 54: Calibration plot of the amplifier for the pressure sensor.. 78 Figure 55: Calibration plot of the displacement sensor including built-in amplifier 79 v m LIST OF TABLES Table 1: The resulting force on the fluid caused by the applied force 22 Table 2: Time of loss-of-resistance from the two trials shown in Figures 15-20 31 Table 3: Summary of the statistics of the decay and static models 37 Table 4: Summary of the mean epidural depths 42 Table 5: Summary of the average normalized diameter 58 Table 6: Average maximum intensity of edges 59 Table 7: Average S N R and C N R 61 ix ACKNOWLEDGEMENTS I owe particular thanks to Dr. Rob Rohling for his guidance and patience throughout the years. I thank Dr. Allaudin Kamani for giving insight into the world of epidurals and for participating as the anesthesiologist in the epidural experiments. I also thank Vick ie Lessoway for offering her expertise and service as an ultrasonographer. 1 INTRODUCTION Epidural Anesthesia Overview Epidural anesthesia, a form of regional anesthesia, is an important and widely accepted analgesia technique in obstetrics to effectively alleviate labor pain[l]-[5]. To facilitate the delivery of the local anesthetic, a catheter is inserted through a needle into the epidural space, a narrow space surrounding the dura mater within the spinal column. Although the use of epidural anesthesia has increased over a few decades, conventional epidural techniques continue to have a failure rate in the range of"6-25%[6][7]. Epidural anesthesia is considered more difficult than other regional anesthetic techniques[8][9]. The epidural anesthesia procedure (or epidural) begins by having the anesthesiologist perform palpation to identify key anatomical landmarks, generally in the mid-lumbar region of the patient's back. Palpation identifies the correct vertebral elements such as the spinous processes so the midpoint can be determined for the initial point of entry for the needle. Usually, the midpoint between the L 3 - L 4 vertebrae or between the L 2 - L 3 vertebrae is chosen for needle entry. Before the epidural needle is inserted, a local anesthetic is injected into the skin using strict aseptic technique. A Tuohy needle with a Huber tip, illustrated in Figure 1, is then inserted and advanced until the tip reaches the epidural space using the loss-of-resistance techriique[10]. The loss-of-resistance technique is a widely accepted method for indicating when the tip of the needle enters the epidural space[l 1][12]. A syringe, typically filled with air or saline, is attached to the needle and is slowly advanced with relatively little resistance through the skin, fat, and subcutaneous tissue[13]. The needle then encounters the relatively stiff and dense ligament layers known as the supraspinous ligament, interspinous ligament and ligamentum flavum, as illustrated in Figure 2a. Once the needle has encountered the very dense ligamentum flavum, the anesthesiologist feels a high resistance to injection until the needle breaches the ligamentum flavum and enters the epidural space, as shown in Figure 2b and 2c. The loss-of-resistance is felt when the needle enters the epidural space, as the air or saline is easily injected into the epidural space[14]. The syringe is then detached to allow the threading of the catheter through the needle and into the epidural 1 space. The needle is also withdrawn leaving the catheter in place to allow continuous or repeated administration of anesthetic without any further needle insertions. Figure 1: A typical epidural needle. The Tuohy needle is a hollow needle with a Huber tip, a very slightly curved tip which is designed to reduce or prevent coring. Images courtesy of CSE (http://www.csen.com/anesthesia). Figure 2: The loss-of-resistance technique, (a) The epidural needle is advanced into the interspinal ligament, (b) High resistance is felt and the saline cannot be easily injected when the needle passes through the dense ligaments, (c) The loss-of-resistance is felt when the needle enters the epidural space and the saline is easily injected. Image courtesy of World Anesthesia Online (http://www.nda.ox.ac.uk/wfsa). 2 Like all other obstetric interventions, epidural anesthesia carries risks. Complications can include backache, headache, shivering, hypotension, bladder dysfunction and inadequate pain relief[15]. More rare are the inadvertent dural puncture, fetal distress, neurologic injury (from peripheral nerve injury to paralysis), cardiac arrest, allergic shock and maternal death[16][17]. Complications arise from both the dosage of the local anesthetic and from technical difficulty. The former includes the amount and type of drug used as the local anesthetic, and the latter encompasses problems with the needle and catheter insertion: The technical difficulty depends entirely on the patient, the equipment and the skill of the anesthesiologist. Patients of different ethnicity and body mass index have varying anatomical characteristics such as the skin-to-epidural space distance[18]-[21]. Patients having different bony characteristics may require a change in needle puncture site and/or a different needle insertion trajectory. The equipment, such as glass or plastic syringe, also affects the performance of locating the epidural space due to different physical characteristics[22]. Most importantly, the success of the procedure is determined by the skill and experience of the anesthesiologist[23][24]. In routine settings, the anesthesiologist does not have any available detailed knowledge about the patient's internal spinal anatomy prior to the procedure. Without such knowledge, the anesthesiologist can only make educated estimates based on experience and external palpation. These skills, including the loss-of-resistance technique, can only be learned through observation and practice on human subjects. The presence of obesity, scoliosis or edema makes the epidural procedure even more challenging[25]. 1.1 Motivation There are several problems associated with traditional epidural anesthesia: • Obtaining competency requires a relatively high number of attempts and has significant associated risks when compared to other regional anesthetic techniques. • The loss-of-resistance technique is the only feedback mechanism for indicating entry of the needle into the epidural space. 3 • Anesthesiologists determine the location and trajectory, of the epidural needle solely by palpation and experience. Although the loss-resistance technique has been used for many years, only 61% of the punctures are successful at the first attempt[26]. Competency in epidural procedures is obtained by observation and practice. In one study, a success rate of 60% is achieved after 20 procedures and reaches 80% after 90 attempts[8]. Another study shows 60% success after 10 attempts and 84% after 60 attempts[27]. 90% is considered consistent with competency and is also claimed to require at least 60 attempts by another study[2 8]. As concluded by several groups, obtaining competency using traditional methods requires a relatively high number of attempts compared to other regional anesthetic techniques. Although residents can practice on cadavers or simulated tissue and ligaments, none provide accurate haptic feedback. Much of the experience is gained by performing the epidural anesthesia on actual human patients. However, training on human patients clearly carries a high risk of complications. Furthermore, patients varying in weight, height and ethnicity may have different epidural experiences that are unfamiliar to the anesthesiologist resulting in additional risk of complications. Since improving the learning curve while avoiding patient risks would be beneficial, there have been a number of attempts to construct epidural anesthesia simulators[29]-[31]. These simulators provide feedback of expected forces but have not been widely accepted. Some even provide training for choosing the correct puncture site and needle angle in difficult patients. However, the force feedback is based on mathematical models simulating the various anatomical structures with sub-optimal realism[31]. This may be due to subtleties and dynamic interactions that exist only while performing the actual epidural procedure on human subjects in vivo. In the past, studies have focused on physical properties of the epidural space and surrounding tissues[32][33]. Even the development of the loss-of-resistance technique was based on the observations of the different densities of the tissues, particularly the ligamentum flavum and the epidural space[34]. Anesthesiologists have relied on this technique which has remained virtually unchanged for the past seventy years. Even with 4 the increase in the physiological knowledge involving epidurals, there has been little change to the procedure. Thus, modern epidural anesthesia continues to rely on the loss-of-resistance technique as the only feedback mechanism to indicate entry into the epidural space. Although it is the accepted standard, the technique is not completely reliable[35][36]. There are no external physical characteristics of the patient that can provide information of the exact location of the epidural space. It also possible the loss-of-resistance may also occur outside the epidural space[35]. Having to solely rely on this technique means the patient is exposed to any risks associated with the technique including complications from the actual injection of saline or air into the epidural space[37]. Additional feedback that can be used in conjunction with the loss-of-resistance technique could decrease the risk of patient complications. Over the years, several research groups have come up with devices to assist in locating the epidural space[38]-[40]. However, these devices require custom-built equipment and are not widely used today. In 1980, ultrasonography was first proposed for guidance of epidural anesthesia[41]. Ultrasound has several unique advantages over other medical imaging technologies: it is portable, relatively inexpensive, real-time, immediate, and poses no known risk to the patient. Moreover, ultrasound has the potential to depict the spinal anatomy including the epidural space before or during the epidural procedure. Nevertheless, ultrasound guided epidural anesthesia has not found widespread clinical acceptance. A possible explanation is the relatively low image quality of early generation ultrasound machines. With the latest digital ultrasound systems, several groups have performed research with the use of ultrasound in regional anesthesia[25][27][42]-[55]. Due to recent successes, there is renewed interest in the use of ultrasonography[56][57]. The use of ultrasound in almost all types of regional anesthetic techniques has been shown to significantly reduce the number of puncture attempts and puncture planes, and improve quality of analgesia and patient satisfaction. Other expected benefits include a reduction in the frequency and severity of associated complications, and enhanced education and training. Furthermore, the benefits may provide an overall reduction in the .5 cost of patient care, even when the cost of the equipment and maintenance is included. Thus, continued research of ultrasound in epidural anesthesia would be very beneficial. Another aspect that anesthesiologists find challenging is the determination of the needle puncture site and trajectory solely by palpation and experience. If the insertion of the needle is unsuccessful, the anesthesiologist has to reinsert the needle causing further discomfort to the already stressed patient. Having an additional aid to assist in the guidance would be beneficial and may reduce the number of unsuccessful attempts. Again, ultrasonography is a very good candidate since rapid anatomical imaging can be used to help the anesthesiologists choose a suitable puncture site. However, one of the difficulties with ultrasonography is that it cannot image bone and features beyond most superficial bone interfaces[58]. Ultrasonic guidance is further complicated because the epidural space is surrounded by the vertebral bones[58]. Recently, there have been advances in user programmability of ultrasound systems allowing research groups access and control of hardware attributes including beam properties[59]. One such property is the probe's beam angle which may be utilized by ultrasonographers to obtain better image quality of a target feature by selecting the correct reflection angle. One study has shown substantial improvement in the visibility of a needle by adaptively steering the beam towards the strongest echoes of the needle[60]. A n additional benefit of being able to perform beam steering is to use spatial compounding, a technique that combines several images of the same scene from different view points to form a single scene. Studies have shown that spatial compounding is beneficial in many clinical applications[61]-[66]. Another study has developed an improved spatial compounding technique by introducing an intermediate warping step to produce significant improvement in point resolution, edge sharpness and speckle noise[67]. These recent advances in ultrasound research allow for further investigations in many possible applications, including epidural anesthesia. 6 1.2 Statement of Problem N o w that the issues of epidural anesthesia have been discussed, the problem statements are presented as follows: 1. There has been research and development of simulators used to improve the learning curve of the loss-of-resistance technique while reducing patient risk, but the models are based only on properties of the various anatomical structures. Very little research has been carried out on the quantification of physical characteristics of actually performing the loss-of-resistance technique. This may reveal subtleties and fundamental characteristics that may be used to improve accuracy and realism in current simulators. 2. The loss-of-resistance technique is currently the only feedback mechanism for indicating entry of the needle into the epidural space. Recently, studies have used ultrasound to identify the epidural space and the surrounding structure, but more research is needed to validate the skin-to-epidural space distance and the length of the puncture path with the simultaneous measurement of the loss-of-resistance technique. 3. Ultrasound-assisted guidance has not been widely adopted for epidurals because imaging the epidural space in the vertebral interspace has been very challenging due to the surrounding bony structures. Thus, the ultrasound probe location and beam direction are essential for obtaining good image quality. There has been no research pertaining to improving ultrasound image quality specific for epidural anesthesia. Good results may help future acceptance of ultrasound-assisted guidance in epidural anesthesia. A n additional benefit is that it may be used in other applications other than epidural anesthesia. 7 1.3 Research Objective . In conjunction with the problem statements, the dissertation presents the following research objectives: 1. One of the research objectives is to gain insight and understanding of the physical characteristics involved in epidural anesthesia using the loss-of-resistance technique. More specifically, the goal is to measure and analyze physical quantities: the force applied at the plunger by the anesthesiologist, the pressure of the fluid, and the position of plunger relative to the syringe. An instrumentation device is developed to acquire physical data and physical models are used to describe their relationships. The device is designed such that an anesthesiologist may use it in a practical manner • during clinical trials. 2. Another objective is to validate the use of ultrasound images by comparing the depth of the depicted epidural space to the length of the inserted portion of the needle when it has entered the space using the loss-of-resistance technique. Consistency of the epidural space in the ultrasound image and by the loss-of-resistance technique is essential to further fundamental research of real-time ultrasound-assisted guidance. 3. The final objective is to develop an adaptive ultrasound imaging technique that can help enhance the overall image quality for epidural anesthesia. Two-dimensional spatial compounding with warping is used as the basis for these investigations. Objectives 1, 2 and 3 are individually described in Chapters 2, 3 and 4, respectively. 8 t 2 INSTRUMENTATION AND MODELING OF T H E LOSS-OF-RESISTANCE TECHNIQUE 2.1 Introduction In order to gain an understanding of the physical characteristics involved in epidural anesthesia, initial research begins with the investigation and identification of the characteristics involved in the loss-of-resistance technique. An instrumentation device is constructed to measure and capture the identified characteristics, and designed for the use by an anesthesiologist in clinical settings. Finally, analysis of the loss-of-resistance and several physical models relating the characteristics are performed using the instrumentation device in epidural experiments. Physical characteristics are determined by examining the interactions of the anesthesiologist and patient with respect to the syringe, as shown in Figure 3. The anesthesiologist typically supports the barrel and needle with one or both hands while using his thumb to push the plunger. The anesthesiologist applies a force Fa onto the plunger relative to the barrel overcoming the frictional force F f to cause a change in displacement D of the plunger with respect to the barrel. As a consequence of Fa, the force of the plunger F P acts on the fluid inside the barrel resulting in positive pressure of the fluid, P. If the needle tip is inserted into the dense ligament flavum, the flow of the fluid, v is significantly impeded causing an increase in the fluid pressure. The anesthesiologist "feels" the resistance because the force applied on the plunger is no longer able to move the plunger due to the high pressure P. Once the needle tip enters the epidural space, the flow of the fluid is no longer impeded and the pressure decreases rapidly causing a substantial displacement of the plunger. The condition is known as the loss-of-resistance. 9 I) V 7 Plunger Barrel Needle Tissue Skin Figure 3 : Physical interactions of the epidural needle. T h e three c h a r a c t e r i s t i c s m e a s u r e d b y the i n s t r u m e n t a t i o n d e v i c e are the a p p l i e d f o r c e Fa, f l u i d p r e s s u r e P, a n d p l u n g e r d i s p l a c e m e n t D. T h e c h a l l e n g e o f d e v e l o p i n g i n s t r u m e n t a t i o n f o r e p i d u r a l p r o c e d u r e s is that m e a s u r e m e n t s m u s t b e p e r f o r m e d i n a s ter i l e e n v i r o n m e n t w h i l e m i n i m i z i n g pa t i en t d i s c o m f o r t . T h e p r e p a r a t i o n t i m e m u s t b e s h o r t , n o part o f the d e v i c e c a n be a t t a c h e d to the pat i ent , a n d the a n e s t h e s i o l o g i s t s h o u l d b e a b l e to p e r f o r m the e p i d u r a l p r o c e d u r e u s i n g the d e v i c e w i t h m i n i m a l t r a i n i n g a n d o b s t r u c t i o n . C o n s i d e r i n g the d i f f e r e n t p h y s i c a l c h a r a c t e r i s t i c s a n d c o n s t r a i n t s f o r c l i n i c a l a p p l i c a t i o n , it is r e a s o n a b l e to m e a s u r e the f o l l o w i n g p h y s i c a l c h a r a c t e r i s t i c s : • T h e a p p l i e d f o r c e Fa. • T h e f l u i d p r e s s u r e P. • T h e p l u n g e r d i s p l a c e m e n t D. In c l i n i c a l se t t ings , d i r e c t l y m e a s u r i n g the f l u i d p r e s s u r e is d i f f i c u l t b e c a u s e the p r e s s u r e s e n s i n g d e v i c e i n c l u d i n g the t r a n s d u c e r a n d c a b l e s m u s t b e s t e r i l i z e d . F o r this r e s e a r c h , the p r e s s u r e m e a s u r e m e n t s are u s e d as r e f e r e n c e d a t a f o r a n a l y s i s a n d it w i l l b e s h o w n 10 that the pressure can be determined indirectly from the applied force and plunger displacement in certain conditions. Although there are many useful characteristics such as the force of the needle against the tissue and the position of the needle tip relative to the patient, measuring these quantities is difficult and impractical because accurate sensing devices require attachment to the patient, cannot be sterilized, or are too large and cumbersome. Only the loss-of-resistance with the plunger is measured in this research. 2.2 Instrumentation Building the appropriate device to measure each characteristic requires appropriate sensors, hardware, software and calibration which are described in the following subsections. 2.2.1 Sensors Measuring the three physical quantities, force, pressure, and displacement requires three individual sensors. Each sensor is chosen based on several constraints such as accuracy, range and practicality. The SLB-25 force sensor manufactured by Transducer Techniques (Temecula, C A ) , shown in Figure 4, is used to measure Fa. The sensor has a mini-spherical load button used to measure compression force. This force sensor is very small and has a footprint diameter of 9.5mm. The SLB-25 has a maximum compression load of 251bs which is sufficient for measuring the thumb force used in the epidural procedure. A custom-built stainless steel harness that can be fitted on an anesthesiologist's thumb is used to mount the sensor. Sterility is maintained by wearing the sensor under a sterile glove. Figure 4: The SLB-25 force sensor. The sensor is used to measure the force of the thumb acting on the plunger. Image courtesy of Transducer Techniques. I I The PX302 pressure sensor manufactured by Omega Engineering, Inc. (Stamford, CT) , shown in Figure 5, is used to measure P. The sensor uses a corrugated stainless steel diaphragm that is fluid filled and has a pressure range between 0 and 300psi. Typically, the expected pressure range of the fluid in the epidural syringe is between 0 and lOpsi. A custom-built stainless steel adaptor is used to attach the rubber tube to the epidural syringe. Figure 5: The PX302 pressure sensor. The sensor is used to directly measure the fluid pressure of the syringe. Image courtesy of Omega Engineering, Inc. The C S P R IP65 displacement sensor manufactured by M T S System Corp. (Cary, N C ) , shown in Figure 6, is used to measure D. The sensor, based on magnetostriction principles, uses a small ring magnet situated along the transducer, a thin cylindrical rod. The maximum stroke length is 72.3mm and can sufficiently measure the full extension of the plunger. The ring magnet is attached to the plunger and the transducer rod is attached to the barrel by a custom-built stainless steel harnesses. The ring need not touch the rod, so friction should be negligible and allows the anesthesiologist to retain the full feeling of loss-of-resistance. Sterility is maintained by covering the sensor with a sterile drape before mounting it to the sterilized harness attached to the syringe barrel. Figure 6: The CSPR IP65 displacement sensor. The sensor is used to measure the displacement of the plunger with respect to the barrel. Image courtesy o f M T S System Corp. 12 2.2.2 Hardware and Software Additional circuitry is required to power the sensors and interface with the digital capture board. The force sensor requires a 5 V voltage regulator and a differential voltage amplifier. The displacement sensor also requires a 5V voltage regulator, but it already has a built-in voltage amplifier. The pressure sensor requires a 10V voltage regulator and a differential voltage amplifier. Additionally, +15V and -15V voltage regulators are needed to power the two differential amplifiers. The selection of the circuit components are based on the requirements for optimal operation in epidural procedures. CircuitMaker 2000 (Altium Ltd. , Carlsbad, C A ) is used for design of the circuit layout. Two identical printed circuit boards are fabricated by A P Circuits (Calgary, A B ) . The circuit components and designs are described in Appendix A . The two assembled circuit boards are housed in a black plastic box with openings to allow entry of the power and signal cables from the sensors and computer. A single ±25V (max 1A) power supply unit is required to power the entire instrumentation device. The amplified signal cables, from the box, are connected to the Q8 terminal board (Quansar Inc., Markham, ON) . The Q8 hardware-in-the-loop control board and its associated terminal board, both manufactured by Quansar Inc. (Markham, ON) , are used to capture the voltage signals from the sensor to a P C . The control board contains 14-bit analog-to-digital converters that measure the voltage range between -10V and +10V, and can sample data at a rate up to 192kHz. A PCI interface card to the Q8 board is installed into a P C (2.8GHz Xeon H T , 512KB L 2 cache, 400MHz F S B , 2 G B R A M ) for capture of the three measurements. The R T X driver (Ardence, Inc., Waltham, M A ) is installed into the PC to provide real-time operation in the Windows X P (Microsoft Corp., Redmond, W A ) operating environment. WinCon 4.0 (Quansar, Inc., Markham, O N ) is the real-time control application for Windows that integrates with Simulink (The MathWorks, Inc., Natick, M A ) . The software allows the user to create Simulink models that are automatically built into the C code using the Visual C++ 2003 (Microsoft Corp., Redmond, W A ) compiler. 13 The sensor data is captured in real-time by WinCon and stored directly onto the local hard drive in MAT-f i l e s . Model simulation and analysis are performed using M A T L A B 7.0.1 (The Math Works, Inc., Natick, M A ) . 2.2.3 Testing and Calibration The complete treatment of testing and calibration of the sensors and amplifiers are discussed in Appendix B . In summary, the relationships between the physical quantities and the sensor voltage are the following: F a = 1 . 2 8 3 F o u l > r c e +0.487 Eq. 1 P = 3 -483F o u t p r e s s u r e +0.2043 Eq. 2 D = 2 6 . 0 8 F o u t d l s p ] a c e m e n l - 3 1 . 8 6 Eq. 3 Fa is compression force with lbs units, P is fluid pressure with psi units, D is the displacement with mm units, and Voui is the sensor voltage with V units. 2.3 Modeling A research goal is to determine any relationship between the physical characteristics. In particular, is it possible to use force and displacement measurements to estimate the pressure? If so, it would allow the pressure sensor to be omitted in clinical trials and make the instrumentation even less obtrusive. The next subsections discuss three possible models: the static, dynamic and decay models. Each model incorporates different physical and empirical properties. A n analysis is performed on the models using experimental data to determine their suitability. 2.3.1 Static Model The static model describes a non-dynamic system which assumes the system has no motion or fluid flow. The static model establishes the basic relationship between the physical characteristics and its principles are used for more complex models. The fundamental relationship describing the pressure P(i) varying over time t o f an incompressible static fluid is 14 P(t)-^p- Eq.4 where Fp(t) is the force exerted on the fluid over an area A. For the case of the epidural needle, shown in Figure 3 , P(i) is the pressure of the fluid inside the barrel, FP(t) is the force exerted by the plunger onto the fluid inside the barrel, and A is the inner cross-sectional area of the barrel. To relate the applied force Fa(t), a close examination of the syringe reveals a small non-zero angle 0 exists between the central axis of the plunger and barrel due to design and manufacturing tolerances. Therefore, the plunger can be pushed into the barrel with an applied force FJj) that has a non-zero perpendicular component, as shown in Figure 7. Although it may be possible to push the plunger perfectly parallel to the barrel, it is very unlikely for a human in practical settings, including epidural procedures. Fa cosf? Plunger Figure 7: Force diagram of an angled plunger. Although exaggerated in the diagram, a very small non-zero angle 9 exists between the central axis of the plunger and barrel. The perpendicular component of F a ( /) is counteracted by the normal force F^(t). Since the friction force Ff(t) is proportional to F^(t), the equation relation friction force to the applied force is F{(t) = vFa(t)sm0 where p is the coefficient of friction of the glass-glass interface. Although 0 depends on the position of the plunger relative to the barrel, the range is assumed to be small so 0 is 15 assumed to be constant. The force acting on the fluid Fp(/) is the net force of the parallel components and is given by F P ( / ) = F a ( O c o s 0 - / / F , ( O s i n 0 Eq. 5 Let ^ ( c o s ^ s i n ^ ) , then Eq . 5 can be written as Fv{t) = KFz(t) Eq.6 where ka describes the complex interactions of the glass-glass interface of the plunger and barrel. This derivation is an oversimplification of the real interaction, and 9, ju are unknown but allows a correction factor to be incorporated into one parameter, ka. The value of ka is determined empirically and assumed constant. These complex interactions include friction due to the imperfections of the plunger and barrel, and the "wetting" of the fluid of the glass-glass interface. A value of ka close to unity implies little loss of force being transferred from Fa(i) to FP(t). The analysis and estimation of ka are discussed in Section 2.4. Combining Eq . 4 and Eq . 6 yields the following: i P(t) = k * F ^ Eq. 7 The relationship shown in Eq . 7 is used to compare the force and pressure data measured by the instrumentation device. 2.3.2 Dynamic Model The nature of the epidural procedure is such that fluid is continuously injected into the tissue to detect the loss-of-resistance, and therefore, the dynamics of the fluid must be considered. For simplicity, the fluid is. assumed to be incompressible, non-viscous and has steady flow. From conservation principles, the continuity equation states that the volume flow rate at any point in the fluid is constant: v A = constant where v is the speed of the fluid and A is the cross-sectional area. Another conservation principle, Bernoulli 's equation is given by P + p g x + j p v 2 = constant 16 where P is the pressure of the fluid, p is the density of the fluid, g is the acceleration due to gravity exerted on the fluid, and x is the height of the fluid. Since the syringe for the epidural is typically inserted horizontally, the change in pressure due to different heights is negligible. If Bernoulli 's equation is applied to the fluid flowing through two tubes (i.e., the barrel and the needle) with different cross-sectional areas, their relationship is given by Px+\pv2 =P2+\pv22 Eq.8 Similarly, the continuity equation between two tubes is ' v, Ax - v2 A2 Eq. 9 Combining Eq . 8 and Eq . 9 yields the following: A2 P2 -P< = T P v , 2 ( l - - V ) Eq.10 Eq. 10 can be used to estimate the pressure difference between the barrel and needle. Given the approximate ratio of the cross-sectional areas of the barrel and needle is 19, and the upper-limit speed of the plunger is lOmm/s (verified in later clinical tests), the difference in pressure is calculated to be approximately 0.003psi. The resulting pressure difference is relatively small and does not significantly effect the instrumentation since the pressure sensor does not measure at that level of accuracy. For the epidural procedure, the measured pressure difference is approximately in the range of 0 and lOpsi. Therefore, it can be assumed that the pressures from which the fluid flows from the barrel to needle and other cylindrical connections are approximately constant. Hence, this dynamic model is not relevant and is not used in modeling the physical characteristics. A corollary is that measurements from the pressure sensor attached at the end of the connector are sufficiently accurate and require no additional compensation due to fluid motion to the sensor. 17 2.3.3 Decay Model Some of the fluid is known to leak past the plunger, s o E q . 7 needs to be corrected i f the leakage is significant. In general, the leakage is dependant on the plunger force, displacement of the plunger, and time, so the general pressure decay Pdccay is introduced as follows: " ^ ( 0 = ^ ( ^ ( 0 ^ ( 0 , 0 A simple expression for Pdccay is needed so that Fa(t) and D{t) can be easily used to estimate P{i). Two cases are investigated: a stationary and a moving plunger. When the plunger is stationary, the pressure is expected to decay exponentially from the initial pressure (at the time when the plunger stops moving) caused by FP. Since the plunger is motionless, small changes in the applied force Fa does not affect the initial pressure because it is countered by static friction. If the change in Fa is large, it w i l l cause the plunger to move. When the plunger is in motion, pressure does not decay (although some leakage still likely occurs) because it is continually and directly affected by Fp . Thus, the pressure decay simplifies to the following: P(t) = dz>(0 * 0 A _ _ _ _ _ _ " ~ 7 ~ «tP(Q _ A 1 At Eq. 11 Therefore, substituting Eq . 6 into Eq . 11 yields the following: * 0 P(t) = ka Fa ( 0 dD(i) A • ' * _________ ""P _____ Eq. 12 > a, - 0 where z is the exponential time constant and /, is the time at which the plunger stops moving, T and k3 are estimated experimentally by applying a range of constant Fa (see Section 2.4). The relationship shown in Eq . 12 is used to compare the force and distance to the pressure data measured by the instrumentation device. 18 2.4 Constant Force Experiment To estimate ka and x, a set of experiments are performed by applying different constant forces to the plunger while measuring the average fluid pressure. The needle end of the plunger is sealed such that the fluid only leaks through the plunger-barrel interface. A s the fluid leaks continually, the plunger moves until it is fully depressed and stops. The decay time constant is estimated by analyzing the decaying pressure of the fully depressed plunger. The constant, force experiments mimic to some degree the constant-force protocol ("smooth" technique) of the anesthesiologist. 2.4.1 Materials and Method The epidural syringe, obtained from the JH-0550 epidural catheterization kit (Arrow International, Inc., Reading, P A ) is a 5ml glass syringe. The barrel of the syringe has an inner diameter of 12.46±0.02mm. The barrel is supported vertically by a clamp with the plunger on the topside. Water is drawn into the syringe while ensuring there is little or no air bubbles. To reduce errors, the barrel and syringe interface is always wetted before each trial. The pressure sensor is attached to the needle-seat using a three-way stopcock. The third opening in the stopcock is closed off to prevent leakage from this end. The displacement sensor is attached to the syringe using the custom-built harness. Using pre-determined weights, a constant downward force is applied to the plunger while the sensor data is acquired and stored to the P C for post-analysis. Three weights 1.40T±0.0021bs, 3.441±0.0021bs and 4.842±0.0021bs are used in the experiment. For each weight, five trials are performed and data acquired at a sampling rate of 0.01s. To estimate ka, the average pressure of each trial is calculated by averaging the measured pressures between the time interval when the readings are stable (not oscillating) and the plunger becomes fully depressed and stopped (just before the onset of decay). Manual selection of the time intervals for the averaging is performed. Additionally, the standard deviation of the average pressure is also calculated in the same time intervals. The average pressures for each weight are averaged and multiplied with the cross-section area to determine the average force exerted on fluid (see Eq . 4). The 19 best-fit line (based on least-squares) is determined for the plot of the applied force and the force exerted on the fluid. The estimated ka is the slope of the best-fit line. To estimate x, the decay portion of each trial is fitted to ah exponential curve to determine its individual exponential time constants, x is the average of the time constants for all the trials. 2.4.2 Results and Discussion The pressure and displacement data for a typical experimental trial are shown in Figure 8 and Figure 9, respectively. A t time ^Os , the weight has not been added to the system. Once the weight is added, there is a significantly large change in pressure and displacement. For approximately 200 seconds, there are significant pressure oscillations because the plunger is almost fully extended causing slight swaying. Since the anesthesiologist normally inserts the plunger to at least half length (e.g., at ?=300s), the earlier oscillating part of the data is ignored. When the displacement level is at approximately /=580s, the plunger is completely depressed and the pressure decay is observed. The fluid is observed to be leaking through the plunger-barrel interface. The leak continues until the fluid pressure equalizes with atmosphere pressure (and fluid friction). The decay is only approximately exponential due to the complex interactions of the fluid leak along the plunger-barrel surface. In clinical settings, the plunger only remains motionless in the order of seconds such that only a small amount of decay is observed and can be approximated using the exponential decay model. The key observations of Figures 8 and 9 is that there is an interval with constant pressure, followed by an interval with a substantial decay. This supports the use of Eq . 12, which divides the equation into a pseudo-static part and a static part. For the first pseudo-static interval, the constant drop of D(t) suggests a constant leak rate through the plunger-barrel interface. For the second static-decay interval, the decay suggests the leakage is non-constant and decreases as the pressure decreases. Since water is close to incompressible, this decay suggests that other compressibility and viscosity factors are at work, likely from small amounts of air introduced in the system. • s 20 100 200 300 400 500 600 700 800 Time (s) Figure 8: Pressure profile of fluid in syringe using constant applied force. The applied force acting on the plunger is 3.441bs. 40 i 0' 1 1 1 1 1 — 1 0 100 200 300 400 500 600 700 800 Time (s) Figure 9 : Displacement profile of plunger using constant applied force. The applied force acting on the plunger is 3.441bs. 21 The mean and standard deviation of the pressure for each trial are calculated by manually selecting the interval between the non-oscillating pressure and prior to the decay (/=200s to 7=580s). The cross-sectional area of the inner barrel is 0.1890±0.0004in 2 and is used to compute F P . The mean values of FP and Fa along with their standard deviations are given in Table 1, and the graph is shown in Figure 10. A linear fit is applied to the plot and its R2 value is 0.9998 implying a reliable fit and suitable use of constant ka. The slope is &a=0.8998, which is approximately 10% of unity. The pressure of the fluid is caused by approximately 90% of the applied force and the rest is lost by frictional forces. It should be noted that for very low values of Fa, the relationship is, to a slight degree, not directly proportional to Fp because static friction prevents the plunger from moving. In the case of the epidural procedure, the linear relationship is valid since the anesthesiologist constantly exerts a relatively large amount of applied force to move the plunger. Table 1: The resulting force on the fluid caused by the applied force. Fa (lbs) FP (lbs) 1.401±0.002 1.281±0.005 3.441±0.002 3.08±0.05 4.842±0.002 4.33±0.02 5 Figure 10: Plot of resulting force on the fluid against the applied force. The slope is of the best-fit line is the estimated ka value. 22 Figure 11 shows the decay portion of the pressure data for all fifteen trials. The decay time constant was determined for each trial by fitting an exponential curve to the interval when the decay occurs. The plot is intended to illustrate the large variability of the decay curves which is possibly due to the fluid leaking along different paths in the plunger-barrel interface. Different plunger orientations may cause slight gaps or closures along the imperfect glass-glass interface. It also possible that the syringe was worn down through repeated use as it was only designed for one-time-use. A s mentioned, compressibility and viscosity factors of the various materials in the syringe may also play a role. 300 T i m e (s) Figure 11: Pressure decay profiles of a glass epidural syringe. The decay portions from the profiles (fifteen trials) are plotted for comparison. The average decay constant r is estimated to be 23±8s. The large deviation is not as significant in the actual epidural procedure because the plunger only remains motionless for up to one or two seconds. If the plunger remains motionless for two seconds', the largest error is approximately 7%. Although the decay constant is computed for a completely depressed plunger, the small change in time constant w i l l have little •impact on the calculations in the actual epidural procedure. 23 For real epidural pressures, additional leakage wi l l occur directly into the tissue. For the pseudo-static part of Eq. 12, the pressure calculations should be unaffected. For the static-decay portion, it is assumed that there is a very small leakage into tissue, especially ligaments. Adding the tissue leakage term would decrease the effective r, but, as mentioned, it should not have a large affect on the pressure calculations for short time intervals. 2.5 Epidural Experiments The objective of the epidural experiments is to analyze the loss-of-resistance technique and physical models on data from actual epidurals performed by an anesthesiologist: The goal is to show the loss-of-resistance from the measured data is consistent with the loss-of-resistance felt by the anesthesiologist. Another goal is to validate the static and decay models with epidural data. 2.5.1 Materials and Method Before performing the epidural procedure, the instrumentation device must be prepared. The instrumentation system consists of the sensor devices, circuit and capture box, power supply and P C (see Sections 2.2.1 and 2.2.2). Strict aseptic techniques are not necessary, but for the majority of the tests, they are still followed so that the procedure remains as similar as possible to the standard procedure. The harness for the force sensor is fitted and attached to the anesthesiologist's thumb. Medical gloves are worn over the harness and sensor to provide non-slip support.to the force sensor and to prevent contamination. The syringe's needle seat is fitted with a three-way stopcock to allow the attachment of the pressure sensor to its third opening using a 1cm stiff intravenous tube. The displacement sensor harness is attached to both the syringe's barrel and plunger. The epidural syringe is a 5ml glass syringe that obtained from the JH-0550 epidural catheterization kit (Arrow International, Inc., Reading, P A ) . The instrumentation sensors and the epidural syringe are shown in Figure 12. 24 Force Sensor Displacement Sensor Pressure Sensor Figure 12: The instrumentation device. The force, pressure and displacement sensor are used to instrument the loss-of-resistance technique. The epidural procedure is performed by an anesthesiologist on two pigs (Sus scrofa domestica), as shown in Figure 13. The pigs, obtained from a local meat store, were culled and prepared for human consumption the same day as the experiments. Organs such as the eyes, stomach and intestines had already been removed, and the pigs had been refrigerated for several hours prior to the experiments. Figure 13: A domestic pig (Sus scrofa domestica) used for the epidural experiments. 25 The epidural procedure is performed by the anesthesiologist in a manner consistent with clinical practice. His hold on the syringe is slightly different to compensate for the instrumentation device, but the loss-of-resistance technique remains unchanged. Two such motion patterns are used in clinical practice: "smooth" and "bouncing". For the "smooth" technique, the anesthesiologist continuously applies an approximately constant force, and for the "bouncing" technique, the anesthesiologist repeatedly applies and releases the plunger in quick succession. Ten trials are performed using each technique and the punctures take place either between the L 3 - L 4 and L 2 - L 3 interspaces to prevent overuse of one interspace. Figure 14 shows the epidural needle with the instrumentation device injected into the L 3 - L 4 interspace in a typical trial. Water is used as the injection fluid and is carefully filled into the syringe and all other parts including the tubes and pressure sensor. During the procedure, a computer operator monitors the device and acquires the data at a sampling rate of 0.01s. When the loss-of-resistance is felt, the anesthesiologist orally indicates the event to the operator so that the time is recorded for analysis. _ta "NS. ^> Figure 14: The use of the instrumentation device on the pig. The epidural needle is inserted into the L 3 - L 4 interspace. Analysis is performed by examining the results for consistency in the loss-of-resistance indicated by the anesthesiologist and the estimations from the profiles. A l l three measurements should show a simultaneous rapid fall when a loss-of-resistance is 26 encountered. To estimate the loss-resistance for each physical measurement, a moving average filter with an interval size of 0.1s is used to remove the majority of the noise. Since the pressure and displacement profiles continually decrease when the fluid is injected into the tissue or ligaments, the time when the minimum slope occurs is chosen to be time when loss-of-resistance also occurs. The force profiles nearby the loss-of-resistance tend to vary depending on the anesthesiologist's actions so the time of loss-of-resistance is found by averaging the time at 10% of the local maximum and minimum values. The paired t-test is used to compare all three times obtained from the force, displacement, and pressure profiles. The mean time of the three estimated times is compared against the time orally indicated by the anesthesiologist. The next step is to examine the physical models discussed in Section 2.3. The estimated pressure is calculated using the force, and displacement profiles, and Eq . 7 and Eq. 12. The estimated pressure values are compared with the actual values using error statistics. Suppose the error Eh at time index /, is defined to be E = P. - P. • i /.measured /, estimated where P,,measurcd is the measured pressure value of the pressure sensor, and P/,cstimated is the pressure estimated by either the static (Eq. 7) or decay model (Eq. 12). Then the mean error £mcan measured over the number of samples N in the a time interval is given by i-\ i The standard deviation OE of the mean error is defined as , — f(E. - E f Eq.14 Another useful measure, the root mean square (RMS) error Erms is defined as ^s=J^5U2 Eq.15 Both the mean error and its standard deviation are used to measure the average and the spread of the error. The additional R M S error is also used since the pressure differences 27 can either be positive or negative. The unpaired t-test is used to compare the errors and standard deviations between the "smooth" arid "bouncing" techniques. 2.5.2 Results and Discussion After acquisition of the raw voltage data, the force, pressure and displacement data are converted to the appropriate physical values using Eq . 1, Eq . 2 and Eq . 3, respectively. Figures 15-17 show the three profiles for the "smooth" technique of a typical trial. Similarly, Figures 18-20 show the three profiles for the "bouncing" technique of a typical trial. The solid vertical line in each plot indicates the time of loss-of-resistance determined by the respective profile. The dashed vertical line in each plot indicates the time of loss-of-resistance orally communicated by the anesthesiologist. The estimated times, shown in Figures 15-20, are summarized in Table 2. 0 5 10 15 Time (s) Figure 15: Force profile of epidural procedure using "smooth" technique. The solid vertical line indicates the time of loss-of-resistance determined by averaging the time at 10% of the local maximum and minimum values. The dashed vertical line indicates the time of loss-of-resistance orally communicated by the anesthesiologist. 28 0 30 15 Time (s) Figure 16: Pressure profile of epidural procedure using "smooth" technique. The solid vertical line indicates the time of loss-of-resistance determined by minimum slope. The dashed vertical line indicates the time of loss-of-resistance orally communicated by the anesthesiologist. 15 20 25 30 Time (s) Figure 17:' Displacement profile of epidural procedure using "smooth" technique. The solid vertical line indicates the time of loss-of-resistance determined by minimum slope. The dashed' vertical line indicates the time of loss-of-resistance orally communicated by the anesthesiologist. 29 0> o O UH 0 2 4 6 8 10 Time (s) Figure 18: Force profile of epidural procedure using "bouncing" technique. The solid vertical line indicates the time of loss-of-resistance determined by averaging the time at 10% of the local maximum and minimum values. The dashed vertical line indicates the time of loss-of-resistance orally communicated by the anesthesiologist. 0 2 4 6 8 10 Time (s) Figure 19: Pressure profile of epidural procedure using "bouncing" technique. The solid vertical line indicates the time of loss-of-resistance determined by minimum slope. The dashed vertical line indicates the time of loss-of-resistance orally communicated by the anesthesiologist. 30 35 30 25 I 20 <u o 1 r .22 15 Q 10 I --• i \ • i \ A i 1 1 t i i t i \ \ 0 12 14 16 1J 2 4 6 8 10 Time (s) Figure 20: Displacement profile of epidural procedure using "bouncing" technique. The solid vertical line indicates the time of loss-of-resistance determined, by minimum slope. The dashed vertical line indicates the time of loss-of-resistance orally communicated by the anesthesiologist. Table 2: Time of loss-of-resistance from the two trials shown in Figures 15-20. Force Profile Pressure Profile Displacement Profile Anesthesiologist Smooth 25.0s 24.6s . 25.2s 26.0s Bouncing 15.2s 15.3s 14.2s . 15.4s When the anesthesiologist uses the "smooth" technique, the applied force actually shows significant variability indicating the applied force is not truly constant. Rather, the anesthesiologist is trying to apply the appropriate amount of force to move the plunger in a slow continuously decreasing manner. The slow changes are used for additional feedback by the anesthesiologist to ensure the plunger is not stuck and to ensure the fluid has consistently high pressure such that the resistance is felt when entering the dense ligaments. The "bouncing" technique differs from the "smooth" technique, but it still provides the same feedback. The plunger's displacement "bounces" because the fluid injected into the ligament is forced back into the syringe because of ligament's high 31 density and elasticity[68]. Even when the needle has entered the dense ligaments, the plunger's displacement continually decreases, indicating the fluid is still slowly being injected. ' ' , To compare the estimated times when loss-of-resistance occurs over the different profiles, paired t-tests (a=0.05) are performed on each paired set over all trials. There are no significant differences between any of the three time of loss-of-resistance obtained from the force, pressure or displacement profile. Therefore, the times of loss-of-resistance for any of the profiles are consistent with each other. There is also no advantage of either the "smooth" or "bouncing" technique in loss-of-resistance analysis. Furthermore, the time of loss-of-resistance can be sufficiently determined from any of the three profiles: Since there is no significant difference for the times of loss-of-resistance between the three profiles, the mean times are calculated and compared with the times orally indicated by the anesthesiologist. The paired t-tests (oc=0.05) results conclude that the times indicated by the anesthesiologist are significantly larger than the estimates from any of the sensors or the calculated mean times for the sensors. To compare the different values, the mean times are subtracted from each estimated value in the same trial. The relative differences are shown in Figure 21, where trials 1 to 10 consist of values from the "smooth" technique and trials 11 to 20 are from the "bouncing" technique. The first three bars of each trial are the relative time difference of the force, pressure and displacement profiles from the mean. The last bar in the set indicates the relative time difference of the anesthesiologist's oral time, and it lagged, in all trials, the mean time. The times indicated by the anesthesiologist are, on average, 0.7±0.3s later than the mean. This discrepancy is consistent with the time it takes for the anesthesiologist to conclude the needle has reached the epidural space and to orally relay the information to the operator. 32 1.5 _l Force • Pressure BDispl. • Anesth. -0.5 A -1.0 -I 1 2 3 4 5 6 7 I I 9 10 II 12 13 14 IS'. 16 17 18 19 20 Smooth Bouncing Trial Figure 21: Comparison of estimated loss-of-resistance times. The mean time of loss-of-resistance is the average of the three time estimates of the force, pressure and displacement profiles. The relative difference of each of the four times (including the anesthesiologist's oral time) of loss-of-resistance with the mean time are shown over all the trials. The static and decay models described by Eq . 7 and Eq . 12 are then used to estimate the pressure from the force and displacement data. The estimated constants, ka and T , from the two models, are discussed in Section 2.4. Because the data from pressure sensor measures the fluid pressure directly, it is the most accurate pressure data and it is used as the reference for comparing the estimated pressures. The estimated and measured pressure profiles for both the "smooth" and "bouncing" technique are shown Figure 22 and Figure 23, respectively. These plots are from the same trials shown in Figures 15-20. The time of loss-of-resistance from the pressure profile and the orally communication are also shown. 33 0 5 10 15 20 25 30 Time (s) Figure 22: Static and decay modeling of pressure for "smooth" technique. The solid vertical line indicates the time of loss-of-resistance determined by pressure profile. The dashed vertical line indicates the time of loss-of-resistance orally communicated by the anesthesiologist. The arrows point to examples of when fluid leakage occurs while the plunger is motionless. 10 Time (s) Figure 23: Static and decay modeling of pressure for "bouncing" technique. The solid vertical line indicates the time of loss-of-resistance determined by pressure profile. The dashed vertical line indicates the time of loss-of-resistance orally communicated by the anesthesiologist. The arrows point to examples of when fluid leakage occurs while the plunger is motionless. 34 For the "smooth" technique, the decay model is clearly a better estimate than the static model. The "smooth" technique has relatively longer constant'displacement and therefore, the decay model is suited for accounting the loss of pressure from leakage. Although the decay model accounts for leakage, the model only estimates decay with a single time constant. The pressure data shows small variations in the decay rate. Possible explanations are that there are small amounts of leakage into the tissue and ligaments, and the time constant for the leakage through the plunger-barrel interface is position dependant (how much the plunger is inserted). A t the time when loss-of-resistance occurs, all estimates rapidly fall, but the pressure estimates slightly lag the actual pressure. A likely reason is that the plunger moved too quickly and consequently, the anesthesiologist was unable to properly apply an axial force upon the plunger so the force reading simply show the greater levels of friction. For the "bouncing" technique, the estimated pressures also follow the rapid oscillations of the force profile, but shows small oscillations with decaying peaks. In some instances, the decay model shows the appropriate pressure decay, but it significantly differs from the measured pressure. The decay model is unable to estimate the decay pressure due to the rapidly changing force and displacement profile. For the majority, the estimates are nearly identical since the plunger is always in motion. A t the time when loss-of-resistance occurs, all estimates fall to zero, but at a slower rate than the oscillations. A possible explanation is that the anesthesiologist is rapidly "bouncing" the plunger until he believes he feels the loss-of-resistance. A t that point in time, he slowly pushes the plunger to confirm his "hunch". The mean errors and standard deviations are shown in Figure 24 and the R M S errors are shown in Figure 25. Using the unpaired t-test (ct=0.05), there are significant reductions in both the standard deviations (by an average of 31 %) and R M S errors (by an average of 40%), but not for the mean error values for the "smooth" technique. The results conclude the decay model is more accurate than the static model for the "smooth" technique. However for the "bouncing" technique, there are no significant differences in the standard deviations, mean and R M S errors indicating neither model performs better: On average, the "smooth" technique has smaller errors and standard deviations for both models compared to the "bouncing" technique, as summarized in Table 3. Overall, the 35 decay model is the better pressure estimator and suggests it should be used in clinical settings to replace the pressure sensor. Furthermore, the "smooth" technique should be used for greater accuracy. 2.5 2.0 1.5 1.0 'JL 0.5 P c oo C S -0.5 -1.0 -1.5 -2.0 -2.5 • Static _ Decay II. m I1 iii in in m in in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1 8 19 20 Smooth Bouncing Trial Figure 24: Mean error comparison between the static and decay models. The vertical error bars represent the standard deviation of the mean error values. 1.2 E 0.8 U _ 0.6 0 0 • Static • Decay 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Smooth Bouncing Trial Figure 25: RMS error comparison between the static and decay models. 36 Table 3: Summary of the statistics of the decay and static models. The mean values are computed from the values of the mean error, standard deviation and RMS error. Mean Error • Static R M S Error Static Mean Error Decay R M S Error Decay Smooth 0.1±0.8psi l.Opsi -0.1±0.6psi 0.7psi Bouncing 0.2±1.0psi 1.2psi 0.2± l.Opsi 1.2psi 2.6 Conclusion A n instrumentation device measuring the applied force of the plunger, pressure of the fluid and the displacement of the plunger relative to the barrel was successfully constructed and calibrated. The instrumentation device is simple, unobtrusive and can be used in a sterile clinical setting. Depending on personal preference, the anesthesiologist can apply either a "smooth" or "bouncing" force upon the plunger when performing the loss-of-resistance technique. The times of loss-of-resistance are estimated for the force, pressure and displacement measurements and shown to be equivalent. Neither the "smooth" nor "bouncing" technique has any distinct advantage for determining loss-of-resistance. The anesthesiologist indicated that he had felt the loss-of-resistance shortly after the sudden drop in the measured values by approximately 0.7±0.3s, which is consistent with the expected oral response time. The static and decay models are used to relate the estimate the pressure from the force and displacement measurements. The static model accounts for friction, and the decay model deals with both friction and fluid leakage. The decay model produced better estimates for the "smooth" technique, but similar estimates for the "bouncing" technique. Overall, the decay model is more accurate of the two and is should be considered as the replacement for direct pressure measurements in clinical trials. 37 3 ULTRASOUND VERIFICATION OF THE EPIDURAL NEEDLE INSERTION 3.1 Introduction In the previous chapter, an instrumentation device successfully measured the time of the loss-of-resistance for indicating entry of the needle tip into the epidural space. This part of the research is to confirm that the depth of the epidural space, as depicted in ultrasound, is similar to the actual depth, as measured along the needle. This is similar to a study by Grau et al.[44], except for the use of the additional loss-of-resistance measurements to confirm the needle has actually entered the epidural space instead of simply relying on the anesthesiologist's belief. Furthermore, the purpose is to gain an understanding of the issues involved in ultrasonography of the epidural space. 3.2 Materials and Method The epidural procedure is performed by an anesthesiologist on two pigs, as described in Section 2.5.1. A Voluson 730 (GE Healthcare, Chalfont St. Giles, Buckinghamshire, U K ) with a real-time 4D convex 1.5-5MHz probe was used to capture ultrasound images of the puncture site during the epidural procedure. The 17Ga x 8.57cm epidural needle is obtained from the JH-0550 epidural catheterization kit (Arrow International, Inc., Reading, P A ) . Once the needle breaches the epidural space by the loss-of-resistance technique, the needle is marked at the base of the puncture and ultrasound is used to image the needle and the epidural space, as shown in Figure 26. The built-in software caliper is used to measure the puncture path length, the distance between the base of the puncture and the tip of the needle, in the ultrasound image. To physically measure the puncture path length, the needle is withdrawn and measured with calipers from the mark to the needle tip. A paired t-test analysis is performed on the ten paired measurements (ultrasound and actual needle), and the mean, standard deviation and R M S errors are calculated. 38 In a second, more limited study, the epidural depth is directly estimated by manually identifying the epidural space in the ultrasound image without the needle. The depth is measured vertically to the skin surface (shortest distance) since the needle path is unknown. That depth can then be compared to a depth indirectly measured by estimating the length between the tip of the needle and the skin directly above it (shortest length). Three trials were performed for each of the L 3 - L 4 and the L4-L5 interspace. Figure 26: Ultrasound imaging of the epidural needle. The ultrasound probe is place sagittally between the L 3 - L 4 vertebrae to image the needle and the epidural space. 3 .3 Results and Discussion A n ultrasound image of a typical epidural puncture between the L 3 - L 4 vertebrae is shown in Figure 27. The software calipers measure the distance between the two x ' s marking the surface of the puncture and the tip of the needle. Figure 28 compares the results o f all the ten paired measurements. The puncture path length varies significantly by up to 16mm because the anesthesiologist is able to choose different puncture sites and needle trajectory. The mean error and standard deviation are 0.0±0.5mm, and the R M S error is 0.5mm. The paired t-test concludes there is no significant difference between the two measurements (a=0.0001) confirming that the ultrasound measurements are 39 consistent with the actual measurements when the needle itself is visible in the ultrasound. Figure 27: Measurement of the puncture path length between L3-L4 vertebrae. The ellipse shows the needle location and trajectory. The two *'s at the ends of the ellipse mark the puncture path length of the needle. [_ • Actual • Ultrasound 1 1 1 1 1 1 1 1 2 3 4 5 6 7 8 9 10 Trial Figure 28: Comparison of the puncture path lengths. The lengths were measured from physical markings of the needle and the ultrasound image. A l l ten trials are measured from the needle in the L 3 - L 4 interspace. 40 Direct and indirect measurements of the depth of the epidural space are shown in Figure 29 and Figure 30, respectively. The direct method is performed by manually identifying the epidural space, seen typically as the doublet echoes, and measuring its depth. The indirect method estimates the epidural depth geometrically using the puncture path length of the needle. The results of for the L 3 - L 4 and L 4 - L 5 epidural space depth are given in Table 4. The small variation in depth is explained by how hard the probe is depressed into the skin causing deformation. Nevertheless, the mean depth of the L 3 - L 4 is significantly less than the depth of L 4 - L 5 , which is also consistent with human anatomy[21]. For each interspace, the direct method is similar to the indirect method and within the error tolerance. Therefore, the depth of the epidural space is consistent with the depth of the needle tip. Furthermore, the use of the loss-of-resistance technique to reach the epidural depth is consistent with the measurements from the ultrasound image. Figure 29: Direct measurement of the epidural space depth. The two x 's mark the skin-to-epidural space distance. The ellipse contains the doublet echoes that are typical of the epidural space. 41 Figure 30: Indirect measurement of the epidural space depth. The ellipse shows the needle location and trajectory. The two x ' s mark the skin-to-epidural space distance. The epidural depth is measured along line from the needle tip to the skin directly above it. Table 4: Summary of the mean epidural depths. The direct method measures the manually identified epidural depth in the ultrasound image. The indirect method measures the depth by geometrically computing the tip of the needle. Direct Indirect L 3 - L 4 29.5±0.8mm 29±2mm L4-L5 35.7±1.7mm 37±4mm In this study, we found that the main difficulty is obtaining good images of both the needle and epidural space. On many occasions, several attempts are made to optimally position the probe in order to obtain the strongest echo of the needle. Because the needle and epidural space are surrounded by bone, there is a very small range o f positions which yield satisfactory images. Furthermore, the probe has to be placed alongside the needle making it very difficult to visualize the needle due to the steep angle. 42 3.4 Conclusion Ultrasound measurement of the puncture path length of the needle is accurate and reliable. The location of the needle in the image is consistent with the actual location but the puncture path length varies significantly depending on the choice of puncture site and trajectory. The depth of the epidural space is consistent with the measurements of depth of the needle using the loss-of-resistance technique. However, obtaining optimal ultrasound images of the epidural space and needle is very difficult due to the steep angle of the needle and surrounding bone structures. If ultrasound-assisted guidance is to be widely used in epidural procedures, then good image quality depicting the needle and epidural space is needed. 43 4 ADAPTIVE SPATIAL COMPOUNDING 4.1 Introduction The use of ultrasound imaging to examine the epidural space has been particularly challenging. To image the epidural space, the patient must arch her back causing the spinal elements to move apart. Then the probe must be placed in a specific position and angle to obtain the optimal image of the epidural space. Other important features such as bony surfaces may not be visualized because the probe is at a sub-optimal angle for good reflection of the structures. The motivation of this research is to improve feature visualization by adaptively enhancing features at different beam angles and combining these enhancements to show an overall improved image quality. This is done with a technique called adaptive spatial compounding. The concept of spatial compounding is illustrated in Figure 31. Figure 31: The concept of spatial compounding. It is the process of averaging (compounding) multiples views of the same region taken from different points of view. For two-dimensional ultrasound, multiple images are obtained by electronically steering the sound beam through a range of angles while the probe is stationary. At each angle, the image is constructed. The compounded image is the average of the set of beam-steered images. 44 Spatial compounding is a technique that signal-averages the images over varying beam angles. However, this technique often results in blurry compounded images when there is misalignment among the individual images from refraction and speed-of-sound errors. A n improved spatial compounding technique was proposed by introducing an intermediate warping step that non-rigidly realigns the features in each of the individual image frames before averaging[67]. This warped spatial compounding technique is used as the starting point of this research to further improve image quality. 4.2 Materials and Method 4.2.1 Hardware and Development Tools The Ultrasonix 500RP (Ultrasonix Medical Corp., Burnaby, B C ) equipped with a 38mm linear 4 - 9 M H z probe and a 40mm curvilinear l - 9 M H z probe is used to acquire the set of beam-steered images. The pre-scan-cOnverted frames are stored locally onto the system's hard disk drive. The frames are transferred to a standard P C (2.8GHz Xeon H T , 512KB L 2 cache, 4 0 0 M H z F S B , 2 G B R A M ) for post-scan conversion (to Cartesian coordinates), compounding and analysis. A l l applications including image acquisition, post-scan conversion, compounding and display are developed using Visual C++ 2003 (Microsoft Corp., Redmond, W A ) . The Microsoft Foundation Class ( M F C ) library is used for G U I development, and the Integrated Performance Primitives (IPP) 2.0 library (Intel Corp, Santa Clara, C A ) is used for high-speed image processing and numerical computation. Evaluation and analysis of the results are performed using M A T L A B 7.0.1 (The MathWorks, Inc., Natick, M A ) . 4.2.2 Image Acquisition and Post-scan Conversion A n application is developed to acquire pre-scan-converted images from the ultrasound system using the application programming interface (API) provided by Ultrasonix. The A P I is used to access various system functions and probe properties. More specifically, the application allows the user to select the beam angle sweep range and the size of the angle increment. Once the user begins the capture process, the application automatically selects the lowest beam angle, and captures the pre-scan-45 converted frames (individual beam-steered image). Then the next beam angle of the probe is set to capture the next frame. The process is repeated until it captures the frame of the final beam angle of the specified range. In this research, the captured frames consist of the reference frame of angle 0°, and eight beam-steered frames of angles ±2° , ±4° , ±6° and ±8° . Frames with beam-steered angles beyond ±10° degraded significantly. The pre-scan-converted data frames are received as a series of rectangular gray-scale data for a given beam angle. Typically, it can take up to one second for each frame capture because the Windows Messaging System is used as the communication protocol between the software application and A P I . . The pre-scan converted frames are copied from the ultrasound system to the P C for scan-conversion and processing. Scan-conversion involves transforming the pre-scan-converted frames, by using the beam-angle and probe type, into either into a trapezoidal or curved image (depending on the type of probe) in Cartesian coordinates for further processing or display. 4.2.3 Adaptive Spatial Compounding Method The spatial compounding technique uses non-rigid (warped) image registration prior to compounding. The warped images are then compounded using one of several possible methods, each with its own merits and demerits. First, warping is described, then several compounding techniques. 4.2.3.1 Block-based Shift Vector Estimation for Warping The warping process corrects misalignment of features prior to compounding the frames. The basic principle is to divide each frame into small blocks and determine the optimal alignment for each block to the reference image. Warping is described by the set of registration vectors from the blocks. This means after scan-conversion, all beam-steered frames (±2°, ±4°, ±6° and ±8°) are warped with the registration vectors to match the reference frame (0°). Registration is based on correlating blocks of pixels between the beam-steered frames and pixels in a predefined search area of the reference frame. The blocks are defined such that the centers are evenly spaced horizontally and vertically, and may overlap each other depending on the size and number. The size of each block is 96 46 by 96 pixels to ensure that each block contains significant features. A 5 by 5 set of blocks are thus defined on each target frame. Each block in the beam-steered frame is compared to a corresponding block of the same location in the reference frame. It is also compared around neighboring locations based on a horizontal and vertical search range of ±10 and ±4 pixels, respectively. This search range encompasses reasonable translational misregistration (relative to refraction and speed of sound maximum errors) and assumes that rotation and shear is small relative to the block size. Matching is based on the statistical correlation between the target and reference frames. Given two blocks, the reference block A and the beam-steered block B, and a shift vector (Ax, Ay), then the pixel A(x,y) corresponds to pixel B(x+Ax, 7+Ay). Let I denote the set of all indices where B contains a valid ultrasound pixel that is within the borders of the ultrasound image. These corresponding pixel pairs can be arranged into vectors A-, and Bh where /el. Let fxA and pB be the respective averages of Aj and Bt over all /el. Then the correlation coefficient C is defined by The value of C ranges between -1.0 to +1.0, where a value of +1.0 implies a perfect determined by finding the highest correlation value within the search range. The process is repeated over all blocks for each beam-steered image. 4.2.3.2 Interpolation of Shift Vectors Once all the shift vectors for each block have been determined for each beam-steered image, a shift vector for each pixel is computed using radial basis function (RBF) interpolation technique. The R B F technique offers smooth interpolation and is computationally practical for the relatively small number of points in this application. The radial basis function approximation is described here for the two dimensional case. Let x=(x,y) be a two-dimensional point (in a Cartesian coordinate system) and c = Eq. 16 match of the two image blocks. The best match and its associated shift vector are 47 (fi,...fy) be the set of N two-variable real-valued functions with corresponding distinct points {xi,.. .,xN] such thatfrflxi). Then the radial basis function s(x) is Eq. 17 where p is a low-degree polynomial of degree M, is a real-valued constant, || • || denotes the Euclidean norm, and (f> is single-variable real-valued function. The space of all polynomials of at most M degrees in 2 variables is denoted by n2M. The solution can be found by satisfying the interpolation condition s(xi) = fi / = ! , . . . , TV Eq. 18 and applying the side condition £ A,. _,(%,) = 0 V ? e . Eq. 19 There are several choices for <f> and p. For two-dimensional registration, a common choice is the thin-plate spline, where $>)=r2log(>) is the biharmonic equation and p(x)=ao+ci]X+a2y is the affine trend function[69][70]. Thus, Eq . 18 and Eq . 19 imply the coefficients of the radial basis functions and the affine trend function can be solved using following linear system ' A Q W f f ^ ) T 0 V where A = (a , , ) = Wx, . 1 Q = Eq.20 i,j = \,...,N \ - (XX,...,A.N)J a = ( a 0 , a , , a 2 ) T f = ( / „ . . . , fN)T 48 To obtain interpolation of the shift vectors for each pixel, the R B F technique is applied separately to the horizontal and vertical components. If the set of vertical components from the shift vectors for a single beam-steered image is represented as f with the corresponding block location (at the center), then X and a can be solved using Eq. 20. A fast matrix inversion routine provided by the IPP library is used to solve the linear system. The technique is repeated for the horizontal component for the same image. Once X and a have been determined for both components, the shift vectors for all pixels are interpolated using Eq . 17. 4.2.3.3 Image Remapping (Warping) Warping is performed by remapping the pixels using their corresponding shift vectors and resampling to a regular grid is performed using nearest-neighbor interpolation. This step is performed using the image-processing routines of the IPP library. 4.2.3.4 Compounding Both conventional and warped spatial compounding perform a basic signal-averaging step where the valid pixel intensity values for all beam-steered and reference images are averaged to form the final compounded image. More formally, let Pi(x,y) be the valid pixel intensity value at coordinates (x,y) where i is the image index in I(x,y), which is the set valid pixels (within the image border) of all the images. Then the intensity of the pixel of the compounded image is given by P*vm&(x>y) = i T , 1 , i Z J O Eq.21 where | • | denotes the cardinality of the set. One of the problems with this method is that features present in some frames are reduced in overall intensity after averaging. Edge responses may appear stronger in some beam-steered images than others because an edge perpendicular to the echo plane reflects back a higher energy echo. The idea is to develop methods that adaptively enhance edges perpendicular to the echo plane when the images are compounded. Two new adaptive 49 methods are developed in this research: a gradient-based and a median-based compounding technique. The gradient-based compounding method is used to sharpen edges including point features (spots). To detect the edges of features, the Laplacian of Gaussian filter is initially convolved with each warped image to obtain an edge response while filtering speckle noise. The suitable kernel size for speckle reduction is determined empirically to be a 9 by 9 mask with a standard deviation of 1.5 pixels in order to obtain the strongest gradient or edge response of the significant features while ignoring the smaller size speckle noise. The values of the computed gradient at each pixel location are used as weights for computing the weighted-average of each compounded pixel. The values of the weighted-average are used as the final intensity value of the compounded image. The gradient value produced by the Laplacian of Gaussian at each pixel of an image is G,(x,y). Then the weighted-average for the pixels of the compounded image using the .gradient-based technique is given by ]^G,(x,y) /? , . (* , 7 ) ^gradient (*, X> = ' 1 ( j c y ) Eq. 22 The goal of this approach is to retain strong reflections/edges in the compounded image by weighting their contributions more than weak reflections/edges at the same anatomical locations. The median-based compounding method is an extension of the gradient-based method. Once the gradient responses G,(x,v) have been computed, an additional decision making step is introduced for each pixel. Given a set of overlapping pixels at a particular location, i f more than half of the frames that have a pixel value above the threshold value ^threshold, then the median intensity for that pixel Over all the images is used as the pixel intensity of the compounded image. Otherwise, the gradient weighted-average of the intensities is used as the pixel intensity of the compounded image. Gthrcshoid is determined empirically by choosing a suitable value that gives the best edge response without introducing speckle noise. Again , the goal is to further improve the edge response by 50 performing median computation on images with stronger reflections/edges rather than weighted-averaging which tends to blur the edge responses. More formally, define the set of pixels P(x,y) to be P(„, y) = {Gt (x,y) | G,(x,y)>G threshold ' Then the weighted-average for the pixels in the compounded image using the median-based technique is given by median (x,y) = JMedian(P(x,v)) , |P(x,v) | > ^ \Pp*m*(*»y) .otherwise Eq. 23 4.2.3.5 Software Application A multi-threaded, object-oriented M F C application, shown in Figure 32, is developed to perform scan-conversion, adaptive spatial compounding algorithms and image display. The application is able to load one or more beam-steered images acquired by a linear or curvilinear probe, and view the pre-scan converted or scan-converted image(s) and the resulting compounded image. O tonv. cane Qw_fpcw-f l i n M - M I * I » | [ J i-**JI*fkfJ\- *S4 M.QQmt Mod-;hijiCunfcUrd fcflt* cwtktt:. 1**1 t «n*ptf. >*iMM«) a-rjt 's«c!. CVr^ Hnq ruw WP*-: Uudhg i..«««Me. Tot* 1 ttwjn. Figure 32: Adaptive spatial compounding application. The application can compute and display beam-steered images, conventional and adaptive spatial compounded images for the linear and curvilinear probes. 51 4.2.4 Experimental Verification A number of experiments are conducted to evaluate image quality of five available methods: no compounding, spatial compounding, warped spatial compounding with averaging, adaptive spatial compounding with gradient-based averaging and adaptive spatial compounding with median-based averaging. Experimental verification is performed on the data of a specially-constructed agar phantom obtained from the experiments in [67]. The phantom consists of 39% glycerol (by mass) mixed with water to form the solvent. Square columns of "white agar", consisting of 30 g/L cellulose fibre, were embedded in the clear agar substrate. Small bright points were created by stirring air bubbles into the agar mixture just before pouring. This phantom was specifically designed to investigate image quality for point features, edge features and homogeneous regions. To evaluate point features, bounding boxes over manually selected air bubbles from the test phantom are established. Pixels that have intensity values that are at least half as bright as the brightest pixel in the bounding box are counted. The normalized diameter is computed and expressed as the diameter of a circle with the same area as the counted pixels. The normalized diameters are compared over all the compounding methods using the paired t-test. Edge sharpness is quantified by measuring the intensity transition of a manually chosen boundary. The average intensity profile is the average of 30 intensity profiles evenly spaced, 2.2mm wide that perpendicularly crosses the boundary. The maximum value of the average intensity profile convolved with the derivative of a Gaussian with (7=2 is compared over all the compounding methods using the paired t-test. The amount of speckle within a homogeneous region is measured using the signal-to-noise ratio (SNR). The signal-to-noise ratio for a region is defined as S N R = Eq. 24 where x a v g is the average intensity of the region and crx is its variance. S N R is computed for both the clear and white homogeneous regions of the image. 52 The contrast-to-noise ratio (CNR) is used to compare a light against a dark region. The contrast-to-noise ratio is defined as where x a v g and y a v g are the average intensity of the light and dark regions, respectively, and o x and o y are their respective variances. C N R is computed on manually selected dark regions consisting of a clear homogeneous regions, and light regions consisting of white homogeneous regions. A qualitative analysis is also performed on porcine tissue. The ultrasound images are acquired during the epidural experiments described in Section 2.5 and 3.2. A n ultrasonographer steadily positions the probe to minimize motion error during frame acquisition. Acquisition requires several seconds due to the slow rate of change in the beam angle. Features such as muscle tissue, bone structures and the epidural space are qualitatively compared. 4.3 Results and Discussion The phantom used for experimental verification is depicted in Figure 33 using standard ultrasound. The different spatial compounding techniques, applied to the set of phantom images, are shown in Figures 34-37. Qualitatively, the resulting images show remarkable differences relative to each other. Conventional spatial compounding appears to be the most blurred, and warped spatial compounding appears to recover some image quality, while the adaptive spatial compounding techniques recover most of the sharpness. The amount of speckle noise present in the compounded images also varies substantially. x. avg -y, avg C N R = Eq.25 53 Figure 3 3 : Reference image of the phantom. Figure 35: Warped spatial compounding with averaging of the phantom. Figure 37: Adaptive compounding with median-based averaging of the phantom. Manual selections of spots are shown in Figure 38, their normalized diameters are plotted in Figure 39, and the summary of their averages are given in Table 5. Conventional compounding has the most blurring and hence a larger average spot diameter. Applying the paired t-test (a=0.05), shows that both the conventional and warped spatial compounding with averaging have significantly larger normalized diameter relative to the reference image. Both gradient-based and median-based techniques have no significant difference relative to the reference image implying that both techniques are able to recover the resolution of the point features. Furthermore, the normalized diameters from both techniques are significantly smaller than those of the warped spatial compounding with averaging. 56 Figure 38: Selection of spots for evaluation of point features. 1.6 1.4 1.2 1.0 0.8 —I u "c3 0.6 o 0.4 0.2 0.0 • Reference B Warped Comp w/ Avg • Adaptive Comp w/ Med • Conventional Comp • Adaptive Comp w/ Grad J l I L J I J ] 1 2 3 4 5 6 7 8 9 10 11 12 13 Feature Number Figure 39: Normalized diameter of spots. 57 Table 5: Summary of the average normalized diameter. Image Type Average Normalized Diameter (mm) Reference 0.86±0.17 Conventional Comp. 1.1 ±0.2 Warped Comp. w/ A v g 0.93±0.18 Adaptive Comp. w/ Grad 0.87±0.17 Adaptive Comp. w/ M e d 0.87±0.18 Manual selections of edges are shown in Figure 40, their maximum slopes are plotted in Figure 41, and the summary of their averages are given in Table 6. Again , conventional compounding produces edges with the least edge strength. The paired t-test (et=0.05) shows that the maximum slopes from conventional compounding are significantly less than those of the reference image. A l l three warped compounding methods recover most of the edge sharpness of the reference image. Figure 40: Selection of edges for evaluation of edge strength. 58 3 0 0 • Reference • Conventional Comp B Warped Comp w/ Avg • Adaptive Comp w/ Grad • Adaptive Comp w/ Med 9 10 II 12 Figure 41: Maximum intensity of edges. Table 6: Average maximum intensity of edges. Image Type Average Max Slope (intensity/mm) Reference 170±40 Conventional Comp. 140±50 Warped Comp. w/ Avg 160±50 Adaptive Comp. w/ Grad 160±50 Adaptive Comp. w/ Med 160±50 Manual selections of dark and light regions are shown in Figure 42, the results of the SNR and CNR calculations are plotted in Figure 43, and the summary of their averages are given in Table 7. The paired t-test (a=0.05) shows that only the gradient-based technique is not significantly improved over the reference. All other techniques have significantly better noise reduction. 59 Figure 42: Selection of dark and light homogeneous regions for evaluation of noise. The lines connecting the rectangles indicate the pairings used in the C N R evaluation. • Reference •Conventional Comp 5 Warped Comp w/ Avg • Adaptive Comp w/ Grad • Adaptive Comp w/ Med S N R 1 (light) S N R 2 (light) S N R 3 (light) S N R I (dark) S N R 2 (dark) S N R 3 (dark) C N R I C N R 2 C N R 3 Figure 43: SNR and CNR of dark and light regions. The C N R values are calculated using the paired regions shown in Figure 42. 60 Table 7: Average SNR and CNR. Image Type Average SNR-Light Average SNR-Dark Average C N R Reference 10±2 6.5±0.7 6.3±1.0 Conventional Comp. 16±4 11±3 9±2 Adaptive Comp. w7 A v g 15±4 11±3 9±2 Adaptive Comp. w/ Grad 12±2 8.9±1.8 6.8±1.7 Adaptive Comp. w/ M e d 13±3 9.2±2 7.2±1.9 Test data sets are acquired from a porcine specimen and the resulting images are shown in Figure 44. Conventional compounding produces an image consisting of slightly blurred bone structure, reduced speckle, and slightly better defined muscle striations. Warped compounding with averaging improves the image with some sharper muscle definition, but still contains blurred bone structure. Adaptive compounding with gradient averaging further sharpens edges of the bone and muscle. Adaptive compounding with median averaging produces similar results to the image from the gradient averaging. Another set of the images consisting of the epidural space and surrounding bone are shown in Figure 45. Conventional compounding produces blurry images with poor visualization of the epidural space. Warped compounding with averaging slightly recovers the epidural space feature, but some significant blurring occurs in the featureless regions. Adaptive compounding with gradient averaging produces slightly sharper edges on the bone and epidural space, but the blurring in the featureless regions remain. Adaptive compounding with median averaging produces sharper edges along the bone and epidural space, and the contrast around those features is significantly higher which brings out some more detail around the features. Furthermore, the blurring in the featureless regions is noticeable reduced. Overall for the porcine test data, the adaptive spatial compounding with median-based averaging produced the best compounded image with similar or better image 61 quality than the reference image depending on the region. However, the improvements are modest, but a small improvement can have a large impact for this application. In Figure 45, the epidural space can be seen in the reference image because the ultrasonographer took great care in the positioning of the probe. For an anesthesiologist with limited experience in ultrasonography, the epidural space may not appear in the reference image so the adaptive spatial compounding technique may help the anesthesiologist image the epidural space more quickly and without the need for optimal probe placement. However, one of the issues with warping is that significant misregistration may .occur in featureless regions because the block matching algorithm is unable to determine the best match. Since there are little or no features in such regions, the correlation values are very low resulting in possible misregistration, which may actually blur the image more than conventional compounding. Although this can occur, it is not as important since the region is presumed featureless and of less interest. Due to the lengthy acquisition time, the images may possibly suffer from slight motion error. However, the process of non-rigid registration and warping allows for correction of a small degree of motion error. This additional benefit is well suited for clinical settings where tissue motion may be even more significant. 62 (d) (e) Figure 4 4 : Images of porcine spinal tissue using different compounding techniques. Two vertebral elements are shown with overlying muscle, fat and skin, (a) Reference, (b) conventional, (c) warped with averaging, (d) adaptive with gradient averaging and (e) adaptive with median averaging. 63 Figure 45: Images of porcine epidural space using different compounding techniques, (a) Reference, (b) conventional, (c) warped with averaging, (d) adaptive with gradient averaging and (e) adaptive with median averaging. 64 4.4 Conclusion Conventional spatial compounding performed the worst in point and edge features, but it performed the best in noise reduction of homogeneous regions. Warped spatial compounding with averaging can satisfactorily recover point and edge features with excellent noise reduction in homogenous regions. Adaptive spatial compounding with gradient-based averaging can excellently recover point features but at the cost of some increased noise in homogenous regions. Adaptive spatial compounding with median-based averaging performed the best in point feature recovery while maintaining excellent edge features with good noise reduction in homogenous regions. For imaging specific features, such as the epidural space, point and edge resolution is likely more important than noise reduction in homogenous regions. The adaptive spatial compounding with median-based averaging also performed the best with the porcine data by sharpening bone, muscle and epidural structures, and increasing the contrast in particularly noisy areas. Therefore, adaptive spatial compounding with median-based averaging is the overall best compromise but it comes at the cost of increased computational power. 65 5 CONCLUSIONS AND FUTURE WORK 5.1 Conclusions The loss-of-resistance, together with the applied force, fluid pressure and plunger displacement are successfully instrumented in porcine studies. The times of the loss-of-resistance characterized by the rapid fall o f the force, pressure and displacement profiles are equivalent and very close to the oral indication of the loss-of-resistance felt by the anesthesiologist. Static and decay models are developed to obtain reasonable estimates of the pressure from the force and displacement measurements. The decay model is more accurate and offers better estimates compared to the static model. Furthermore, accuracy is further increased when the "smooth" technique is used in epidural procedures compared to the "bouncing" technique. Therefore, the pressure can be reasonably estimated by using only the force and displacement. Because the instrumentation device is simple, unobtrusive and sterilizable, and the pressure sensor can be omitted, the instrumentation device can be used in clinical trials. It is also recommended that anesthesiologists use the "smooth" technique during clinical trials to improve the accuracy of the pressure estimates. Ultrasound can reliably and accurately measure the puncture path length of the needle. Furthermore, the depth of the epidural space is consistent with ultrasound measurements and the loss-of-resistance technique (in porcine studies). However, it is very difficult to use ultrasound to visualize the epidural space and needle due to the steep needle angle and surrounding bony structure. The angle and position dependencies result in sub-optimal image quality. Therefore, there is a need to improve ultrasound image quality for epidurals. New algorithms for adaptive spatial compounding are developed to improve several aspects of image quality. The adaptive spatial compounding using a median-based averaging gives the best balance for point resolution, edge resolution and noise reduction in homogeneous regions. This algorithm also provides the best image quality of the . epidural space for the porcine studies because point and edge resolution are especially important. 66 5.2 Future Work One of the future goals is to use the instrumentation device in clinical trials to study the loss-of-resistance technique on pregnant women. Further epidural studies are needed to gain more insight in the dynamics of the loss-of-resistance technique. Additionally, it would be beneficial to consider measuring other physical characteristics such as the elasticity of the tissues or the needle position with respect to the patient. Model improvement is another aspect of future study so that simulations can be made more realistic and accurate. The use of ultrasound can be beneficial in epidural procedures for image-assisted guidance. It has been shown that the needle and epidural space can be accurately measured. So the next step is to track the needle using an automated technique. A needle enhancement using beam steering has already been developed by others, but it must be adapted for use in epidural anesthesia to account for anatomy such as bones and the steep needle angles. Detection of the epidural space with ultrasound is mainly a noise-limited and a human-perception problem, which is especially obvious in the human study. Acceptance o f ultrasound-assisted guidance is possible when it can be used to reliably detect the epidural space, especially by an anesthesiologist with limited experience in ultrasound interpretation. The adaptive spatial compounding algorithm is a step in the right direction, but there are several areas for future work in improving the algorithm including real-time implementation and registration. With the recent introduction of commercial multi-core C P U s , implementation of special-purpose threading tasks would greatly speed up computation. For example, cross-correlation can be performed on one C P U while R B F interpolation can be performed in another C P U . Frames can be sent through a pipeline architecture that performs compounding "on-the-fly" as new frames are acquired. Another area for future work is to improve registration by developing an adaptive technique that improves matching in near or completely featureless regions. Furthermore, an adaptive technique can be developed to automatically choose the optimal parameters (i.e., filter size, Gthrcshoia) for the gradient and median-based averaging algorithms. Although the ultrasonographer who participated in the studies prefers the images 67 p r o d u c e d b y the a d a p t i v e s p a t i a l c o m p o u n d i n g t e c h n i q u e o v e r the o thers , f u r t h e r i n v e s t i g a t i o n i n the i m a g e a s s e s s m e n t b y m a n y u l t r a s o n o g r a p h e r s w o u l d a l s o b e b e n e f i c i a l . 68 REFERENCES [I] Anim-Somuah M , Smyth R, Howell C. Epidural versus non-epidural or no analgesia in labour. Conchrane Database Syst Rev, 2005;(4):CD000331. [2] Declercq E R , Sakala C, Corry M P . Listening to mothers: report of the First National U . S . Survey of Women's Childbearing Experiences. New York: Maternity Care Association/Harris Interactive Inc, 2002. [3] Seebacher J, Malassine P. Anesthesia and delivery. Rev Prat, 1999;49:167-71. [4] Eberte R L , Norris M C . Labour analgesia: A risk-benefit analysis. Drub-Saf, 1996;14(4):239-51. [5] Vincent R D Jr, Chestnut D H . Epidural analgesia during labor. A m Fam Physician, 1998;58(8):1785-92. [6] Le Coq G , Ducot B , Benhamou D . Risk factors of inadequate pain relief during epidural analgesia for labour and delivery. Can J Anaesth, 1998;45:719-23. [7] Watts R W . A five-year prospective analysis of the efficacy, safety, and morbidity of epidural anaesthesia performed by a general practitioner anaesthetist in an isolated rural hospital. Anaesth Intensive Care, 1992;20:348-53. [8] Konrad C, Schupfer G , Wietlisbach M , Gerber H . Learning manual skills in anesthesiology: Is there a recommended number of cases for anesthetic procedures? AnesthAnalg , 1998;86:635-9. [9] Kopacz D J , Neal J M , Pollock JE . The regional anaesthesia "learning curve". What is the minimum number of epidural and spinal blocks to reach consistency? Reg Anesth, 1996;21:182-90. [10] Baumgarten R K . Importance of the needle bevel during spinal and epidural anesthesia. Reg Anesth, 1995;20:234-8. [II] Bei l in Y , Arnold I, Telfeyan C, Bernstein H H , Hossain S. Quality of analgesia when air versus saline is used for identification of the epidural space in the parturient. Reg Anesth Pain Med , 2000;25:596-9. [12] Howel l T K , Prosser D P , Harmer M . A change in resistance? A survey of epidural practice amongst obstetric anaesthetists. Anaesthesia, 1998;53;238-43. [13] Norman D . Epidural analgesia using loss of resistance with air versus saline: does it make a difference? Should we reevaluate our practice? A A N A J, 2003;71:449-53. [14] Bernards C M . Epidural.and spinal anesthesia. In: Barash P G , Cullen B F , Stoelting R K , editors. Cl inical Anethesia. 4th ed. Philadelphia, Lippincott Will iams & Wilkins , 2001:689-713. [15] Paech M J , Godkin R, Webster S. Complications of obstetric analgesia and anaesthesia: a prospective analysis of 10,995 cases. Int J Obstet Anesth, 1998;7:5-11. [16] Kantor G . Obstetrical epidural anesthesia in a rural Canadian hospital. Can J Anaesth, 1992;39:390-3. [17] Haire D B . Obstetric drugs: their effects on mother and child. In: Rothman, B K . ... Encyclopedia o f Childbearing, New York, Henry Holt and Company, 1993. [18] Emmanouil S, Eleni M , Ioanna S, Er i f i l l i A , Athina V . Prediction of the distance from the skin to lumbar epidural space in the Greek population, using mathematical models. Pain Practice, 2005;5:125-34. 69 Watts R W . The influence of obesity on the relationship between body mass index and the distance to the epidural space from the skin. Anaesth Intensive Care, 1993;21:309-10. Harrison G R , Clowes N W . The depth of the lumbar epidural space from the skin. Anaesthesia, 1985;40:685-7. Hoffman V L , Vercauteren M P , Buczkowski P W , Vanspringel G L . A new combined spinal-epidural apparatus: measurements of the distance to epidural and subarachnoid space. Anaesth, 1997;52:350-5. Butler B D , Warters R D , E lk JR, Davies I, Abouliesh E . Loss of resistance technique for locating the epidural space: evaluation of glass and plastic syringes. Can J Anaesth, 1990;32:438-9. Stone P A , Kilpatrick A W , Thorburn J. Posture and epidural catheter insertion. The relationship between ski l l , experience and maternal posture on the outcome of epidural catheter insertion. Anaesthesia, 1990;45:920-3. Downing JW. Technical hazards of epidural analgesia. S Af r M e d J, 1980;58:356-8. Grau T, Leipold R W , Conradi R, Martin E . Ultrasound control for presumed difficult epidural puncture. Acta Anaesthesiol Scand, 2001;45:766-71. De Filho G R , Gomes H P , Da Fonseca M H , Hoffman JC, Pederneiras S G , Garcia J H . Predictors of successful neuraxial block: a prospective study. Eur J Anaesthesiol, 2002;19:447-51. Grau T, Bartusseck E , Conradi R, Martin E , Motsch J. Ultrasound imaging improves learning curves in obstetric epidural anesthesia: a preliminary study. Can J Anaesth, 2003;50:1047-50. Kopacz D J , Neal J M , Pollock JE. The regional anesthesia "learning curve". What is the minimum number of epidural and spinal blocks to reach consistency? Reg Anesth, 1996;21:182-90. Hiemenez L , Stredney D , Schmalbrock P. Development of the force-feedback model for an epidural needle insertion simulator. Irt: Westwood JD, Hoffman H M , Stredney D , Weghorst S, Morgan K . Medicine Meets Virtual Reality, IOS Press and Ohmsha, 1998:272-7. Dang T, Annaswamy T M , Srinivasan M A . Development and evaluation of an epidural injection simulator with force feedback for medical training. Stud Health Technol Inform, 2001;81:97-102. . M a g i l l J, Anderson B , Anderson G , Hess P, Pratt S. Multi-axis mechanical simulator for epidural needle insertion. International Symposium on Medical Simulation, L N C S 2004;3078:267-76. Rodiera J, Calabuig R, Aliaga L , Espinosa W , Hobeich F, Oferil F , Gual A . Mathematical analysis of epidural space location. Int J C l i n Monit Comput, 1995;21:213-7. Usubiaga JE, Wikinski J A , Usubiaga L E . Epidural pressure and its relation to spread of anesthetic solutions in epidural space. Aaesth Analg, 1967;46:440-6. Dogliotti A M . Anaesthesia/Chicago, SB Debour, 1939. Carden E , Ori A . The BIP test: A modified loss of resistance technique for confirming epidural needle placement, Pain Physician, 2006;9:323-5. 70 [36] Hoffmann V L , Vercauteren M P , Vreugde JP, Hans G H , Coppejans H C , Adriaensen . H A . Posterior epidural space depth: safety of the loss of resistance and hanging drop techniques. B r J Anaesth, 1999;83(5):807-9. [37] Saberski L R , Kondamuri S, Osinubi O Y . Identification of the epidural space: is loss of resistance to air a safe technique? A review of the complications related to the use of air. Reg Anesth, 1997;22:3-15. [38] Lechner TJ , van Wil jk M G , Maas A J , van Dorsten F R , Drost R A , Lagnenberg C J , Teunissen L J , Cofnelissen P H , van Niekerk J. Clinical results with the acoustic puncture assist device, a new acoustic device to identify the epidural space. Anesth Analg, 2003;96:1183-7. [39] Jacob S, Tierney E . A dual technique for identification of the epidural space. Anaesthesia, 1997;52:141-3. [40] Shah JL . Identification of the epidural space. B r J Hosp Med , 1991;46:60-3. [41] Cork R, Krye J, Vaughan R. Ultrasonic localization of the lumbar epidural space. Anesthesiology, 1980:52;513-6. [42] Rapp H J , Folger A , Grau T. Ultrasound-guided epidural catheter insertion in children. Anesth Analg, 2005;101;333-9. [43] Rapp H J , Grau T. Ultrasound imaging in pediatric regional anesthesia. Can J Anesth, 2004;51:277-8. [44] Grau T, Leipold R W , Fatehi S, Martin E , Motsch J. Real-time ultrasonic observation of combined spinal-epidural anaesthesia. Eur J Anaesthesiol. 2004;21:25-31. [45] Grau T. The evaluation of ultrasound for neuraxial anesthesia. Can J Anesth, 2003 50:6,Rl-8. [46] Grau T, Leipold R W , Conradi R, Martin E , Motsch J. Efficacy o f ultrasound imaging in obstetric epidural anesthesia. J C l i n Anesth, 2002;14:169-75. [47] Grau T, Leipold R W , Horter J, Conradi R, Martin E , Motsch J. Paramedian access to the epidural space: the optimum window for ultrasound imaging. J C l i n Anesth, 2001;13:213-7. [48] Grau T, Leipold R W , Conradi R, Martin E , Motsch J. Ultrasound imaging facilitates localization of the epidural space during combined spinal and epidural anesthesia. Reg Anesth Pain Med , 2001;26:64-7. [49] Grau T, Leipold R W , Horter J, Conradi R, Martin E , Motsch J. The lumbar epidural space in pregnancy: visualization by ultrasonography. B r J Anaesth, 2001;86:798-804. [50] Perlas A , Chan V W , Simons M . Brachial plexus examination and localization using ultrasound and electrical stimulation: a volunteer study. Anesthesiology, 2003;99:429-35. [51] Chan V W . Applying ultrasound imaging to interscalene brachial plexus block. Reg Anesth Pain Med, 2003;28:340-3. [52] Chan V W , Perlas A , Rawson R, Odukoya O. Ultrasound-guided supraclavicular brachial plexus block. Anesth Analg, 2003;97:1514-7. [53] Schafhalter-Zoppoth I, McCul loch C E , Gray A T . Ultrasound visibility of needles used for regional nerve block: an in vitro study. Reg Anesth Pain M e d , 2004;29:480-8. 71 Gray A T , Collins A B , Schafhalter-Zoppoth I. Sciatic nerve block in a child: a sonographic approach. Anesth Analg, 2003;97:1300-2. Wallace D H , Currie J M , Gilstrap L C , Santos R. Indirect sonographic guidance for epidural anaesthesia in obese pregnant patients. Reg Anesth, 1992;17:233-6. Marhofer P, Willschke H , Greher M , Kapral S. New perspectives in regional anesthesia: the use of ultrasound - past, present, and future. Can J Anaesth, 2005;52:R1-R5. Marhofer P, Greher M , Kapral S. Ultrasound guidance in regional anaesthesia. B r J Anaesth, 2005;94:7-17. Cittadini G , Martinoli C. Ultrasound and the bone: a difficult relationship. Radiol Med , 1995;89:12-7. Rohling R, Fung W , Lajevardi P. P U P I L : programmable ultrasound platform and interface library. Proc. of Medical Image Computing and Computer Aided Interventions, 2003, L N C S 2879. Cheung S, Rohling R. Enhancement of needle visibility in ultrasound guided percutaneous procedures. Ultrasound Med B i o l , 2004;30:617-24. L i n D C , Nazarian L N , O'Kane P L , McShane J M , Parker L , Merritt C R B . Advantages of real-time spatial compounding sonography of the musculoskeletal system versus conventional sonography. A R J 2002;179:1629-31. Huber S, Wagner M , M e d l M , Czembirek H . Real-time spatial compounding imaging in breast ultrasound. Ultrasound M e d B i o l , 2002;28:155-63. Entrekin R R , Porter B A , Sillesen H H , Wong A D , Cooperberg P L , F ix C H . Real-time spatial compounding: application to breast, vascular, and musculoskeletal ultrasound, Semin Ultrasound C D M R , 2001;22:50-64. Entrekin R, Jackson P, Jago JR, Porter B . Real time spatial compounding imaging in breast ultrasound: technology and early clinical experience. Medicamundi, 1999;43:35-43 Kofoed SC, Gronholdt M L , Wilhjelm JE, Bismuth J, Sillesen H . Real-time spatial compound imaging improves reproducibility in the evaluation of atherosclerotic carotid plaques. Ultrasound M e d B i o l , 2001;27:1311-7. Jespersen S K , Wilhjelm JE, Sillesen H . In vitro spatial compounding scanning for improved visualization of atheroscleroris. Ultrasound M e d B i o l , 2000;26:1357-62. Groves A , Rohling R. Two dimensional spatial compounding with warping. Ultrasound M e d B i o l , 2004;30:929-42. Behrsin JF, Briggs C A . Ligaments of the lumbar spine: a review. Surg Radiol Anat, 1998;10:211-219. Bookstein F L . Principal warps: thin-plate splines and the decomposition of deformations. I E E E Trans. Pattern Anal . Mach. Intell., 1989;11:567-85. Carr JC, Fright W R . Surface interpolation with radial basis functions for medical imaging. I E E E Trans. Med. Imag., 1997;16:96-107. 72 APPENDICES Appendix A: Instrumentation Circuits The differential amplifier design for the sensors is illustrated in Figure 46. The amplifier exhibits high SNR, linear frequency response, and high input impedance. The voltage gain is adjusted by varying the resistor R. + 15 V -15 V Figure 46: Instrumentation amplifier circuit diagram. The circuit is designed to amply the small, differential output voltages of both the force and pressure sensors. 73 Figure 47, Figure 48 and Figure 49 illustrate the circuits for regulating specific voltages. The regulators exhibits low drift, high SNR and can be powered by a ±25V power supply unit. Adjusting the resistor R for the 10V voltage regulator fine tunes the output voltage. Input -(~^ ~VSUpp|y) Input -("VSUpp|y) LM7815 G N D 0.33 nF 2.3 nF Tant. 0.01 HF _k - Output (+15V) 1N4005 G N D v i n . v„ LM7915 1.5 nF Tant. 1N4005 • Output (-15V) Figure 47: ±15V voltage regulator circuit diagram. The circuit is designed to power the.differential amplifiers of both the force and pressure sensors. Input ("*~V S Upply) Output (+5V) 0.01 uF Figure 48: 5V voltage regulator circuit diagram. The circuit is designed to power both the force and displacement sensor. 74 1N4005 Input— (~*~VSUpp|y) LM317 ADJ 1.5 uF Tant. '2kO. 1N4005 5.9V Zener .3kn Output (+10V) 1 5 M F Elect. 5 C K 1.5 uF Tant. Figure 49: 10V voltage regulator circuit diagram. The circuit is designed to power pressure sensor. The P C B layout and fully assembled circuit board are shown in Figure 50 and Figure 51, respectively. The two P C B s are identical, but the circuit configuration differs for the different sensors. The power, signal, ground and shield cables are also connected to the circuit boards using banana plugs. 75 Figure 50: Circuit layout. The layout is designed for two different circuit configurations. The red paths are for the top layer and the blue paths are for the bottom layer. Figure 51: Instrumentation circuit boards. The two identical circuit boards each have different circuit configurations allowing for complete power and amplification of the instrumentation device. 7 6 Appendix B: Sensor Calibration Two calibration stages are required for each sensor subsystem: the sensor device and the amplifier circuit. A l l the sensors are pre-calibrated by the manufacturers and are reported to have insignificant non-linearity errors. However, the amplifiers require gain adjustment and calibration in order to amplify signal voltages of the sensors to the appropriate levels. The gain is adjusted such that the maximum output voltage is 10V since it is the maximum allowable voltage for the Q8 control board. The amplifier is calibrated using a Wheatstone bridge configuration, as shown in Figure 52. The resistor, Rl is adjusted around the value of Rl such that the differential voltage AV0Xlt is desired. A range of differential voltages are applied to the inputs of each amplifier and the resulting output voltages are measured and plotted. The plots of the output voltage against the differential input voltage and their calibration curves are shown in Figure 53, Figure 54 and Figure 55. Because the amplifier for the displacement sensor has a built-in amplifier, its calibration differs from the calibration of the amplifiers for the force and pressure sensors. Calibration is performed by measuring (with calipers) the positions of the magnet with respect to the transducer, and also measuring the resulting output voltages. Figure 52: Wheatstone bridge circuit for amplifier calibration. The circuit is used to calibrate the amplifiers by creating small differential voltages in the order of millivolts. 77 12 l AV i n, f o r c e (raV) Figure 53: Calibration plot of the amplifier for the force sensor. A reference differential voltage source is applied to the input of the amplifier, and the resulting output voltages are measured. 12 1 Figure 54: Calibration plot of the amplifier for the pressure sensor. A reference differential voltage source is applied to the input of the amplifier, and the resulting output voltages are measured. 78 1 . 0 -0.5 -0.0 -I , 1 : 1 i 1 1 : 1 1 0 10 20 30 40 . 50 60 70 80 D (mm) Figure 55: Calibration plot of the displacement sensor including built-in amplifier. The position of the ring magnet is varied, and the resulting output voltages are measured. The R2 values of the linear calibration plot of Figure 53, Figure 54 and Figure 55 are 0.9996, 0.999997 and 0.99996, respectively. A s expected, the linear calibration curves are reliable and can be used for establishing a linear relationship between the input and output values. The amplifier for the force sensor is slightly less reliable than the amplifier for the pressure sensor since it needs to amplify smaller differential voltages which are more susceptible to noise. The linear relationship for the amplifier of the force, pressure and displacement sensors are respectively o^ut, force 1 .V1 12 A \f0Ke - 0.3794 Eq. 26 J / o U l , p r e S s u r e = 0 . 1 4 3 6 A ^ p r e s s u r e - 0 . 0 5 8 6 . E q . 27 ^dbpu-_=0.0383D + 1.2212 Eq. 28 A Fin is the input differential voltage with m V units, D is the displacement with mm units, and F0ut is the output with V units. The calibrated values reported by the manufacturers for the force and pressure sensor are respectively F a = 2.195 A F i n > r c e Eq.29 79 P = 0.5000 A ^ p r e s s u r e Eq.30 Fa is compression force with lbs units and P is fluid pressure with psi units. The physical characteristics Fa, P and D must be a function of Voul, since the output voltages are the only values captured by the computer for analysis: After substituting Eq . 26 into Eq . 29, and Eq . 27 into Eq : 30, and performing algebraic rearrangement of Eq. 28, the equations for the three sensors are as follows: F a = 1 . 2 8 3 F o u l i f o r c e + 0.487 P = 3-483 F o u t > p r e s s u r e + 0.2043 D = 26 .08F o u l > d i s p l a c e n i e n t -31 .86 The three equations are the basis for converting the raw voltage values captured by the computer into meaningful physical quantities for simulation and analysis of the physical models. 80 

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.831.1-0101226/manifest

Comment

Related Items