THE USE OF EMG FOR LOAD PREDICTION DURING MANUAL LIFTING

Transcription

1 THE USE OF EMG FOR LOAD PREDICTION DURING MANUAL LIFTING by Sonya Chan A thesis submitted to the Department of Electrical and Computer Engineering In conformity with the requirements for the degree of Master of Science (Engineering) Queen s University Kingston, Ontario, Canada (September, 27) Sonya Chan, 27

2 Abstract The Ergonomics Research Group at Queen s University, supported by the Workplace and Safety Insurance Board, has been developing an on-line system to estimate peak and cumulative joint loading in the workplace. This study will aid the project by examining the muscle activation levels (MALs) in upper extremity and trunk muscles during a manual lifting task using both hands. It was hypothesized that MAL s are correlated with the magnitude of the load in the hands and thus could be used to predict the load which in turn will be used to predict the lower back moments. Alterations in the muscle activation patterns due to lifting different loads were examined. Electromyographic signals (EMG) and kinematic data were recorded from different sites on the trunk and upper limb as subjects lifted a load from the floor to a shelf using squat, stoop and freestyle lift techniques. All raw EMG data were processed to obtain the linear envelopes (LE) which provides estimates of the MAL s. The peak, mean and area of the linear envelopes were calculated. Using regression analysis, a relationship between the parameters and load lifted was found to exist. A non-linear parallel cascade type architecture was used to develop a model to predict the load in the hands. The model uses the EMG parameters as inputs and fits the data via linear and non-linear cascades to the output, i.e. the load in the hands. A model was successfully developed for the squat lift posture using the area, peak and mean of the zero-normalized EMG LE recorded from the erector spinae (L4 level), with a prediction error of ± 1.3kg and for the stoop posture, a prediction error of ± 2.34kg. Given the predicted loads, moments in the lower back were computed using the method of Hof (1992). ii

3 Acknowledgements I would like to thank my supervisor, Dr. Evelyn Morin for all of her generous support and patience in this project and in me. I would like to thank Susan Reid, Alison Godwin and Dr. Mohammad Abdoli for taking the time to answer my many questions. Lastly, I would like to thank my friends and family for their ongoing support and encouragement. iii

4 Table of Contents Abstract...ii Acknowledgements...iii Table of Contents...iv List of Figures...vi List of Tables...vii Chapter 1 Introduction Introduction and Motivation Literature Review Lifting and Lower Back Moments EMG and Lower Back Moments EMG-Force Identification...5 Chapter 2 Methods Data Collection Data Processing and Analysis...11 Chapter 3 Results EMG Signal Processing Lift Symmetry Correlation Analysis Load Prediction Squat Lift Stoop Lift Moments...21 Chapter 4 Discussion and Conclusion Discussion Conclusion Future Work / Recommendations...28 References...3 Appendices...4 Appendix A - Letter of Information and Consent form...32 Appendix B - EMG Linear Envelopes...37 Appendix C - Lift Characteristics...45 iv

5 Appendix D Parallel Cascade Coefficients for Squat and Stoop Lift...47 Appendix E - Moment in the Lower Back...69 v

6 List of Figures Figure 2-1 EMG and Fastrak Sensor Positioning...8 Figure 2-2 Lift styles...11 Figure 2-3 Parallel cascade structure with n cascades...13 Figure 3-1 EMG LE's...14 Figure 3-2 Prediction Model Structure...18 Figure 3-3 Samples of stoop kinematics and EMG LE s...21 Figure 3-4 Moment at the lower back...23 vi

7 List of Tables Table 2-1 Mean and standard deviation of subject physical characteristics...7 Table 2-2 Mean and standard deviation of subject physical characteristics for subject Table 3-1 L4 EMG LE non-normalized Parameter Lift Characteristics...15 Table 3-2 Mean and standard deviation of the orientation of the sternum (in degrees) about the transverse plane during lifting...16 Table 3-3 Correlation (R2) values between the EMG parameters and load lifted for all squat and stoop lifts for subjects Table 3-4 Results in leave one subject out testing for squat lift...19 Table 3-5 Predicted load Statistics for Squat Lift...19 Table 3-6 Results in leave one subject out testing for Stoop Lift...2 Table 3-7 Predicted load Statistics for Stoop Lift...2 Table A-1 Subject Anthropometrics...36 Table C-1 L4 EMG LE Parameter Lift Characteristics...45 Table D-1 Parallel cascade coefficients for squat lift...47 Table D-2 Parallel cascade polynomial coefficients for squat lift...5 Table D-3 Parallel cascade coefficients for stoop lift...59 Table D 4 Parallel cascade polynomial coefficients for stoop lift..62 Table E-1 Body segment parameters taken from De Leva (1996 )...69 vii

8 1.1 Introduction and Motivation Chapter 1 Introduction Low back disorders are commonly developed with repetitive manual lifting (NIOSH, 1992). According to the Bureau of Labor Statistics (BLS), back injuries account for approximately 2% of all injuries and illnesses in the workplace and 75% of back injuries occurred while the employee was performing a manual lifting task. McGill (1997) states that low back injury occurs when the applied load exceeds the strength of the tissue and discusses the theory behind low back pain to aid in understanding the development and the prevention of lower back injuries in the workplace. A link between moments in the back and lower back pain has been established, where moment is the magnitude of force applied to a rotational system at a distance from the axis of rotation (Winter, 24). In this work, a link segment model that includes the weight of the load in the hands, the weight of the different body segments, acceleration of the body segments and the inertial properties of the segments (Hof, 1992) was used to find the total moment in the lower back during manual lifting. As the load in the hands increases, the moment in the low back increases and the muscle forces are increased to counteract the moment. The electromyogram (EMG) is a recording of a small measurable signal that is emitted by a muscle as it contracts (Basmajian and de Luca 1985). As the strength of the muscle contraction increases, the amplitude of the EMG increases. EMG has been used in the modeling of spinal loading (Sparto et al. 1998; Mientjes et al. 1999; Granata and Marras, 1993). These studies provide a link between EMG in the muscles in the back during certain well-defined tasks and the forces acting on the lower back. 1

9 In manual lifting tasks, knowing the value of the weight in the hands is crucial for determining the forces and moments acting on the lower back. Thus the objective of this study is to examine and define a relationship between EMG in specific muscles and the load being lifted in a manual lifting task. This is part of a larger project to develop a body-worn system to examine body motion (kinematics) and EMG as workers perform manual lifting tasks in the workplace. The data will then used to predict lower back moments and the risk of developing lower back pain. 1.2 Literature Review Lifting and Lower Back Moments Many studies have been and are currently being conducted to gain a better understanding of lower back pain, its development and how it can be prevented in the workplace. Much investigation has focused on the analysis of lower back moments during different types of lifts and for different loads. Bazgari et al. (26) provides an in-depth look into lift posture and the effects it has on the moments in the lower back during squat and stoop lifts. They stated that although there is a rationale that a squat lift style is safer than a stoop lift, there was still controversy over which lift style is actually better. Their subjects were asked to perform squat and stoop lifts with and without a load of 18N in the hands. From the different moment trends, the results suggest that the squat lift has reduced net moments and muscle forces. This however depends on the rotation at the thorax, pelvis and lumbar spine and the position of the external load, speed of movement and the curvature of the spine. De Looze et al. (1993) investigated the joint moments and muscle activity in the lower back during lifting and lowering tasks. Subjects lifted and lowered a load of 15.3 kg using squat and 2

10 stoop lift postures. Their results showed that the lowering tasks consistently had peak moments that were slightly lower than the lifting tasks. From observations of the acceleration curves, it was concluded that the slight increase in the peak moment during lifting was due to the larger inertial effects of the actual lifts. Along with kinematics, EMG from the erector spinae muscle, at the L5/S1 level in the back, was examined and it was found that there is more muscle activity (higher EMG amplitudes) in the lifting tasks than in the lowering tasks, similar to the moment results EMG and Lower Back Moments Although moments play an important role in understanding the effects of lifting on lower back pain, EMG can also provide insight into the effects of lifting. EMG can accurately represent co-contractions, the simultaneous activation of agonist and antagonist muscles around a joint. Lavender et al. (1992) examined the EMG response of the trunk muscles to external moments with different magnitudes and directions. Their analysis showed that there is a high correlation between the magnitude of the moments and the level of EMG in the trunk muscles. Abdoli et al. (25) examined the EMG in the erector spinae (ES) at the T9 and L4 levels, external obliques and rectus abdominis during manual lifting, with and without an on-body personal lift augmentation device called the PLAD. The subjects performed stoop, squat and freestyle lifts with 5kg, 15kg and 25kg loads in the hands. EMG amplitudes in both the erector spinae sites were used as indicators as to whether or not the PLAD provided assistance in lifting. The PLAD was found to have significantly lowered EMG amplitudes in the ES by 14.4% to 27.6% and thus reduced the required muscular effort in the back. The results of this study also showed that as the load in the hands increased, back muscle activity increased indicating a possible relationship between EMG and load lifted. 3

11 Arjmand and Shirazi-Adl (26) researched the effects of forward lean on muscle force with and without a load in the hands. Their results indicate that as the external load and forward lean increases, muscle force increases. These results also indicate a possible relationship is present between EMG in the back muscles and load in the hands. Granata and Marras (1995) presented a model which estimates spinal loads and trunk moments based on measured EMG signals and external force data. EMG data from ten trunk muscles and kinematic data from a lumbar motion monitor were used as model inputs. The EMG data were assumed to be linearly related to net muscle force. The tensile force in each muscle was modeled as a function of the EMG signal, muscle cross-sectional area, gain (defined as the muscle force per unit area), muscle length and velocity. The EMG signal was normalized by using the EMG signal value of the muscle when the subject elicits a maximum voluntary contraction (MVC). The cross-sectional area of the muscle, as well as the origin and insertion of each muscle was calculated based on subject anthropometry. Although the muscle crosssectional area at the transverse plane through the lumbo-sacral junction is usually employed, some muscles were modeled using the maximum cross-sectional area since the authors believed some muscles provided more force contribution than could be represented by the area at the lumbo-sacral junction. The calculated muscle force was modeled as acting along a straight line between the muscle origin and insertion. The muscle gain was calculated by comparing musclegenerated trunk moments with measured applied moments. High correlation coefficients (R 2 values) were found between the actual and predicted moments in the sagittal and frontal planes. The error in predictions was found to be less than 15% of the peak sagittal moment and less than 24% of the peak frontal moment. The authors conclude that it is possible to apply an EMGassisted model which reasonably estimates trunk kinetics during free-dynamic lifting exertions. 4

12 Marras and Davis (21) developed a normalization technique for the trunk muscles by finding an EMG-force relationship that does not require a subject to produce an MVC. While MVC normalization is commonly used to normalize EMG levels across subjects, experimental error in the form of uncertainty in whether or not true maxima are actually exerted exists. MVC normalization is not an ideal exercise, and subjects with or without low back pain may not necessarily exert a true MVC. The proposed normalization technique used sub-maximal exertions to determine an EMG-force (moment) relationship, which was found to be linear and could be extrapolated to obtain an estimate of the expected maximum contraction (EMC). The EMC was used as the reference point for normalization. The EMC normalization technique showed a difference of less than 8% in the estimated maximum value compared to the traditional MVC recording technique. Based on the results, the proposed EMC normalization technique was considered a feasible alternative to the MVC normalization and showed that different techniques can be used in the normalization of EMG. Dolan and Adams (1993) used EMG activity from the erector spinae at the T1 and L3 levels to estimate the compressive forces acting on the spine and also to find an EMG-extensor moment relationship. They found that there was a linear relationship between the extensor moment and EMG. To find the spinal compression during lifting, subjects were asked to lift kg, 1kg, 2kg and 26kg weights while EMG was recorded from the back muscles. It was found that a higher velocity lift had a higher peak extensor moment and higher compressive force EMG-Force Identification There are many different ways to model a relationship between EMG and force or moments. Korenberg (1991) developed parallel cascade identification for nonlinear systems. The structure uses parallel cascades of alternating dynamic linear and static nonlinear elements, where the 5

13 linear and nonlinear elements are determined by the user. To train the model, the first cascade is given an input and models it to give a known output and with minimal error. The subsequent cascade then models the errors at the output of each preceding cascade until a solution is reached. Another non-linear identification procedure is fast orthogonal search (FOS), also developed by Korenberg (1988). Mobasser et al. (27) used FOS to estimate elbow-induced wrist force. The FOS estimate uses a summation of linear or nonlinear functions to create a model. The linear and nonlinear functions are defined by the user and chosen based on minimizing the mean square error. EMG was recorded from the biceps brachii, triceps brachii and brachioradialis with arm motion restricted to elbow extension and flexion in the horizontal plane. Data were recorded under isometric, isotonic and light load conditions. Using EMG from the three muscles and joint angle data, the FOS model was able to predict the force generated with an average mean square error of less than 28%, depending on the training data. In summary, lower back moments during manual lifting are proven indicators of the risk of developing low back pain during manual lifting. To obtain an accurate estimate of the lower back moments, the load being lifted needs to be known. It has been established that EMG activity in the back muscles is higher for higher moments (De Looze et al. 1993; Lavender et al. 1992) and the EMG amplitude in the back muscles increases with increasing load in the hands (Abdoli et al. 25, Arjmand and Shirazi-Adl, 26). Thus the objective of this study is to develop a model to predict load lifted during manual lifting tasks using EMG data recorded from arm and trunk muscles. The parallel cascade structure was chosen as the basis model for predicting the load due to its simplicity and robustness. The predicted load, along with the recorded kinematic data, will then be used to estimate the moment at the lower back. 6

14 Chapter 2 Methods An experimental study was designed to collect EMG and kinematic data from the upper body during a manual lifting task. Twelve adult males with no evidence of back pain were recruited to participate in this study. The physical characteristics of the volunteers are given in Table 2-1. Subjects read and signed an information and consent form approved by the Queen s University Research Ethics Board (Appendix A). Table 2-1 Mean and standard deviation of subject physical characteristics 2.1 Data Collection Physical Characteristics Mean ± SD Age(yr) 22.3 ± 3.5 Height (m) 1.78 ±.7 Weight (kg) 79.4 ± 17.7 Data were collected in the Biomechanics Laboratory in the Physical Education Centre at Queen s University. EMG data were recorded using a Bortec surface EMG (semg) recording system and kinematic data were obtained using a Polhemus Fastrak motion tracking system. The EMG data were sampled at 1Hz per channel and the motion tracking data were collected at 3Hz. The EMG recordings were obtained from six muscle sites on the right side of the body: biceps brachii, triceps brachii, lateral deltoid, thoracic erector spinae (T9), lumbar erector spinae (L4) and external oblique. Prior to attaching the electrodes, the skin at the recording sites was cleaned with alcohol and a small amount of conductive gel (Sigma gel Electrode Gel, Parker Laboratories Incs., Fairfield, NJ) was applied to lower the skin impedance. Subjects were outfitted with six pairs of bipolar surface electrodes (Kendall-LTP Meditrace TM, Chicopee, MA) 7

15 that are 1cm in diameter and were placed side by side over the muscles of interest with a center to center separation of 3cm. Each bipolar pair was oriented such that the electrode axis was parallel to the underlying muscle fibres. A single reference electrode was placed on the spine at the C7 level. The motion tracking sensors were placed approximately over the center of mass (COM) of the right upper arm, the right forearm, the head, neck and trunk combined, and over the L4 vertebra. The motion sensors collected positional and orientation information for each segment. Stretchy adhesive medical grade fabric was placed over all sensors to keep them in place. Each subject s age, height, weight, arm length and trunk height were also recorded and are given in Appendix A. The placement of the sensors can be seen in Figure 2-1. Figure 2-1 Sensor Positioning: the bipolar EMG electrodes were attached as followed: 1. right biceps brachii 2. triceps brachii 3. lateral deltoid 4. thoracic erector spinae (T9) 5. lumbar erector spinae (L4) 6. external oblique 7. reference electrode attached over the spine at the C7 cervical vertebra. The positions of the Fastrak motion sensors are attached as follows: a. forearm b. upper arm c. sternum d.l4 8

16 Maximum voluntary static contractions (MVC) were recorded for each of the measured muscles. For the back muscles (erector spinae at T9 and L4), the subject laid down on a bench face down with his legs strapped down, with one strap over the ankles and one strap over the thighs, to keep the lower body in place. The subject then arched his back, to lift his upper body off the bench while an assistant pushed down on his shoulders to provide resistance to ensure maximum contraction. For the external oblique, the subject laid down on a bench face up with his legs strapped down, with one strap over the ankles and one strap over the thighs, to keep the lower body in place. He then performed an abdominal crunch with a twist trying to touch the right elbow to the left knee. While the abdominal crunch was being performed, an assistant provided resistance by pushing on the subject s right shoulder. For the biceps brachii, the subject s shoulder was flexed to a 9 angle and his upper arm was placed on a flat surface so that it was parallel to the floor. The subject s elbow was flexed to a 9 angle and he grasped a fixed bar and pulled towards himself with maximal effort. For the triceps brachii, the arm was placed in the same position as previously described and the subject pushed away from himself against the fixed bar. For the deltoid, the subject stood with his arm at his side with no bend in the elbow. The subject lifted his right arm up to about 9 to the right so that it was parallel to the floor and his shoulder was abducted. While his arm was in this position, a strap was placed tautly over the arm. With the strap taut, the subject was asked to try to abduct the shoulder to greater than 9 with maximal effort. Each of these exercises was performed a total of two times to ensure that an absolute MVC was achieved. The duration of each task was approximately three seconds with a minimum of two minutes break in between each performance of the exercises to avoid muscle fatigue. Subjects were then asked to perform a series of symmetric manual lifting tasks. Five load conditions were evaluated: subjects lifted a rectangular box, with handles on the right and left 9

17 sides, weighing 5kg, 1kg, 15kg, 2kg or 25kg. Three different lift types were performed: squat, stoop and freestyle (Figure 2-2). To perform a squat lift, the subject kept his back straight and upright during the duration of the lift so that there was almost no bend in the lower back. The subject bent his knees, keeping his feet stationary to grasp the handles of the box, lift the box, place it on the shelf in front of him and let go of the handles. To perform the stoop lift, the subject kept his legs straight and stationary so that there was almost no bend in the knees. The subject bent from the hips to lift the box and place it on the shelf. Since the stoop and squat lift were symmetric lifts, the subject s feet were placed parallel and in line with each other. The freestyle lift was performed to the comfort of the subject. The usual stance was to have the feet staggered and a combination of bending the back and knees was used for the lift. For each lift style and load, subjects repeated the lift-lower cycle three times in a row. A total of 45 lift cycles were performed in an experimental session (5 loads x 3 postures x 3 trials). For each lift posture, the loads were presented in randomized order. Subjects 7-12 also performed a lift with no load in the hands (zero-load) after the all the lifts were performed. The subject characteristics for subjects 7-12 are shown in table 2-2. Table 2-2 Mean and standard deviation of subject physical characteristics for subject 7-12 Physical Characteristics Mean ± SD Age(yr) 21.5 ±.67 Height (m) 1.76 ±.4 Weight (kg) ±

18 a b c Figure 2-2 Lift styles - a. squat b. stoop c. freestyle 2.2 Data Processing and Analysis All EMG signals were processed using Matlab v7.. Since the EMG recording system used had no pre-filtering of the data, the raw EMG data were band-pass filtered at 3 4Hz to filter out low and high frequency noise. After filtering, the data were rectified and then low pass filtered using a second order Butterworth filter at 2.7Hz to obtain the linear envelopes (LE s) (Winter, 24). Lift duration was defined as the time between when the subject grasped the box to lift it off the floor and when the subject placed the box onto the shelf and let go of the handle. This was determined with the aid of two switches, one on the handle and one on the bottom of the box. When the box was at rest, a constant (dc) voltage was recorded by the EMG recording system. When the subject grasped the handle of the box, this voltage dropped to a different dc level, and the dc level dropped again when the box was lifted off the surface on which it was resting. The EMG LE curves between the start and end of each lift were used in the subsequent analysis. The LE s from all lifting trials were normalized to the maximum EMG activity attained during the MVC trials (MVC normalization). To find the MVC value, the LE of the MVC trials was analyzed to find a window of two seconds where the average LE value was the highest. The data over the two second window were averaged to give a single MVC value. The EMG LE s of 11

19 the lifts were then divided by the MVC value found for each muscle to normalize across subjects. The mean value, peak value and area under the graph were calculated for the MVC normalized EMG LE s and non-normalized EMG LE s. To find the mean, the average value was found over the linear envelope. To find the peak, the maximum value of the linear envelope was found and integration was used to find the area under the graph. For the six subjects who performed a zero load lift, the mean, peak and area of the non-normalized EMG LE s were also normalized using the mean, peak and area of the EMG LE for the no load lift (zero-normalization condition). As well, the mean, peak and area of the no load lift were subtracted from the mean, peak and area of the non-normalized EMG LE s (zero-minus condition). A correlation comparison between the load lifted and mean value, peak value and area under the graph for the MVC normalized, zero-normalized and zero-minus EMG LE s was calculated using Microsoft Excel 22 for subjects The muscles with parameters with the highest correlation values were selected for use in the load estimation model. A parallel cascade type model (Korenberg, 1991) (Figure 2-3) was trained to identify the load lifted, where the inputs x 1, x 2 and x 3 are the LE mean, peak and area parameters for the relevant muscle for each subject and y is the predicted load. G and A are non-linear functions and n represents the number of cascades. 12

20 Figure 2-3 Parallel cascade structure with n cascades 13

21 Chapter 3 Results 3.1 EMG Signal Processing A representative set of EMG LE s from all muscles for a 25kg squat lift are shown in Figure Subject 7 25kg Squat Lift 1 25 Subject 7 25kg Squat Lift Deltoid Tricep Subject 7 25kg Squat Lift 1 6 Subject 7 25kg Squat Lift Bicep Oblique Subject 7 25kg Squat Lift 1 8 Subject 7 25kg Squat Lift L T Figure 3-1 EMG LE's - EMG LE s for one subject performing a squat lift with a 25kg load. The muscle recording site is indicated on each graph. Note: the vertical axis scales are not the same for all graphs. 14

22 From Figure 3-1, the LE for the deltoid shows low EMG for the first half of the lift which likely means that the lateral deltoid was not activated until the placement part of the lift. The triceps brachii shows very low EMG signals throughout the entire lift and shows a small amount of activity during placement. The biceps brachii show a similar pattern to the deltoid except there is relatively more activity at the beginning of the lift. The external oblique shows relatively low signal amplitude throughout the entire lift. The erector spinae at the L4 and T9 level are activated throughout the lift. The EMG pattern shown in Figure 3-1 show only subject 7 s EMG LE and does not represent all of the subjects. Table 3-1 gives the L4 non-normalized EMG LE parameter lift characteristic across all subjects 7-12, the rest of the L4 EMG LE parameter lift characteristics can be seen in Appendix C. The coefficient of variation shows that there is variability in the EMG LE patterns when performing the different lifts. Table 3-1 L4 non-normalized EMG LE Parameter Lift Characteristics Lift Type Duration L4 Area L4 Peak L4 Mean and Load Mean± Standard 5kg Squat 1.87± ± ± ± Deviation (SD) Coefficient of Variation (CV) 1kg Squat Mean± SD 2.9± ± ± ± CV kg Squat Mean± SD 2.18± ± ± ± CV kg Squat Mean± SD 2.41± ± ± ± CV kg Squat Mean± SD 2.37± ± ± ± CV kg Stoop Mean± SD 1.87± ± ± ±98.9 CV kg Stoop Mean± SD 2.1± ± ± ± CV kg Stoop Mean± SD 2.15± ± ± ± CV kg Stoop Mean± SD 2.8± ± ± ±1471. CV kg Stoop Mean± SD 2.33± ± ± ± CV

23 3.2 Lift Symmetry Since the lifts performed were considered symmetric, the kinematic data of the lifts were analyzed to verify this assumption. Table 3-2 gives the mean and standard deviation of the orientation of the sternum about the transverse plane during lifting. A large standard deviation about the mean angle indicates that rotation about the transverse plane occurred during lifting. There was relatively little twisting during the squat lifts. However, during the stoop lifts, most notably in subject 1, there was more twisting of the trunk during the lift. Table 3-2 Mean and standard deviation of the orientation of the sternum (in degrees) about the transverse plane during lifting Lift Type Mean ±SD Mean ±SD Mean ±SD Mean ±SD Mean ±SD Mean ±SD and Load Subject 7 Subject 8 Subject 9 Subject 1 Subject 11 Subject 12 5kg Squat.83± ± ± ± ± ±1.58 1kg Squat.54±.52.77± ± ± ± ± kg Squat 4.95± ± ± ± ± ±2.14 2kg Squat -.71± ± ± ± ± ± kg Squat 2.22± ± ± ± ± ±1.37 5kg Stoop.21± ± ± ± ±1.45.8±2.14 1kg Stoop 3.92± ± ± ± ± ± kg Stoop.7± ±1.93.2± ± ± ±4.57 2kg Stoop.8± ± ± ± ± ±3.8 25kg Stoop 3.57± ± ± ± ± ± Correlation Analysis Results of the correlation analysis for the squat and stoop lift styles are given in Table 3-3 only for subjects who performed a zero load lift (subjects 7-12). The zero-normalized and zero minus EMG parameters exhibited higher correlations with the load lifted. Also, the correlation values between the load lifted and the EMG parameters for the biceps brachii, anterior deltoid and erector spinae at the L4 level were higher than for the other muscles sites. Thus, these parameters were selected for the use in the parallel cascade model to predict the load being lifted. Given that the biceps brachii, anterior deltoid and erector spinae at the L4 level were selected for further analysis, EMG LE s from these muscles for each load lifted are shown in Appendix B. 16

24 Table 3-3 Correlation (R2) values between the EMG parameters and load lifted for all squat and stoop lifts for subjects 7-12 Correlation with load lifted Squat MVC Normalization Squat Zero- Normalization Squat Zero- Minus Stoop MVC Normalization Stoop Zero- Normalization Stoop Zero- Minus Bicep Area Bicep Mean Bicep Peak Deltoid Area Deltoid Mean Deltoid Peak L4 Area L4 Mean L4 Peak Oblique Area Oblique Mean Oblique Peak T9 Area T9 Mean T9 Peak Tricep Area Tricep Mean Tricep Peak Load Prediction Squat Lift Using a nonlinear parallel cascade algorithm (Korenberg, 1991), a model was developed to predict load lifted from the EMG LE parameters using the parameters that were most highly correlated with load. The parallel cascade algorithm for each muscle site was tested separately because a single muscle site would be best for the portable monitoring system. After trying different parameter sets, the model that was developed for the squat lift posture uses area (x 1 ), peak (x 2 ) and mean (x 3 ) of the zero-normalized EMG LE recorded from the erector spinae (L4 level). The system used is shown in Figure 3-2 and the coefficients g k 11, g k 12, g k 21, g k 22, g k 31, 17

25 g k 32 and a kd, where k = 1 n are given in Appendix D. There are 91 cascades (n) with polynomials to the 25 th degree (d). This particular cascade gave the most accurate load prediction with an error of ±1kg. Since only subjects 7-12 performed zero load lifts, the model was developed using the EMG data from those subjects. Figure 3-2 Prediction Model Structure - Parallel cascade structure for the squat lift data with n cascades: where x 1, x 2, x 3 represent the LE mean, peak and area parameters for each subject and y is the predicted load A leave one subject out prediction of the load lifted was done. To do this, the EMG parameters from 5 subjects were used as inputs to the cascade in Figure 3-2 and the cascade was trained to obtain the loads lifted. Using the trained model, the EMG parameters from the missing subject were used as the input into the model to obtain a prediction of the load lifted. This was repeated six times and a different subject s data was tested each time. The results are shown in 18

26 Table 3-4. Since the different loads lifted were in increments of 5kg, errors in the load predictions that were less than ± 2.5kg were deemed acceptable. Table 3-4 Results in leave one subject out testing for squat lift Error in Load Missing Subject Prediction 7 ± 1.3 kg 8 ± 1.3 kg 9 ±.55 kg 1 ± 1.3 kg 11 ± 1.3 kg 12 ± 1.3 kg To test the model, the EMG parameters from the six test subjects were used as inputs to the model and the load lifted for each individual lifting trial was predicted. Table 3-5 gives the mean and standard deviation of the loads predicted for each of the six subjects Stoop Lift Table 3-5 Predicted load Statistics for Squat Lift Actual Weight (kg) Standard Deviation Mean Using the same cascade configuration as the squat lift (Figure 3-2), a parallel cascade solution was developed to predict load lifted from the EMG signal parameters during a stoop lift. The coefficients g 1, g 2, g 3 and a nd are given in Appendix D. The model that was developed for the stoop lift posture uses area (x 1 ), peak (x 2 ) and mean (x 3 ) of the zero-normalized EMG LE recorded from the erector spinae (L4 level). There are 1 cascades (n) with polynomials to the 19

27 19 th degree (d). This particular cascade gave the most accurate load prediction with an error of ±2.34kg. The model was developed using the EMG data from the subjects Results of the leave on out testing are shown in Table 3-6. Table 3-6 Results in leave one subject out testing for Stoop Lift Error in Load Missing Subject Prediction 7 ± 2.34kg 8 ± 2.34 kg 9 ± 2.29 kg 1 ± 2.34 kg 11 ± 2.34 kg 12 ± 2.34 kg To test the stoop model, the EMG parameters from the test subject were used as inputs to the model and the load lifted for each individual lifting trial was predicted. Table 3-7 gives the mean and standard deviation of the loads predicted for each of the six subjects. Table 3-7 Predicted load Statistics for Stoop Lift Actual Load (kg) Mean Standard Deviation Since the error in the load prediction was higher than for the squat lift, an examination of the stoop lift EMG LE s and kinematics was carried out. It was discovered that the patterns of muscle activation and kinematics were not consistent across all subjects. Lift kinematics and EMG LE s from two subjects are shown in Figure 3-3. In this case, subject 7 lifts the box towards his body before straightening his trunk. This is evident from the high EMG activity in the biceps during the lifting phase. During the placement phase, the biceps and deltoids are both active and the L4 has increasing EMG as the subject straightens to a standing position. Subject 9 does a 2

28 gradual lift where the arms and trunk come up together. The deltoid shows activation early on in the lift to stabilize before lifting and high activation in the placement phase of the lift. The bicep activates when the load is being brought towards the body and stays on until the load is placed on the shelf. The L4 shows relatively low activation levels throughout the whole lift. 2 Subject 7 Stoop 25kg 16 Subject 7 25kg Stoop Lift Position (cm) Forearm-z Upper arm-z Trunk-z Time (s) Subject 9 Stoop 25kg 12 Subject 9 25kg Stoop Lift Position (cm) Forearm-z Upper arm-z Trunk-z Time (s) Figure 3-3 Samples of stoop kinematics (left) and EMG LE s (right) for subject 7(top) and subject 9(bottom) 3.5 Moments Using the predicted loads for the squat lift, the moment at the lower back (L4 level) was calculated using an intersegmental model (Hof, 1992). 21

29 From Hof (1992), the moment equation is set to be M lk k d i i i= dt ( r r ) F [( r r ) m g] + [( r r ) m a ] + ( I w ) = r lk r Term1 k i lk i= Term2 i k i lk i= Term3 i i Term4 (1) where M lk is the intersegmental force F r is the force plate reaction force m i is the mass of segment i g is gravity a i is the acceleration of the center of mass of segment i r r is the position of the center of pressure of the force plate r i is the position of the center of mass of segment i I i is the moment of inertia of segment i w i is the angular velocity of segment i d dt ( ) I i w i reduces to I i α i for a planar movement, where α i is the angular acceleration of segment i. Since there was no ground reaction force measured, F r was replaced with the gravitational force of the box (F box = m box g) and r r is replaced with r box, the position of the center of mass of the box. The first term of equation 1 calculates the moment due to the box, the second term calculates the sum of the moments of each body segment due to the gravity, the third term calculates the sum of the moments of each body segment due to their acceleration and the fourth term calculates the sum of the moments of each body segment due to inertia. The segments that are included in the 22

30 model are the all upper body segments (hand, forearm, upper arm, head and trunk), to find the center of mass of the segments, segment masses and inertial parameters determined in previous studies (De Leva, 1996) were used and are given in Appendix E. To calculate the moments in the lower back, the kinematic data was filtered with a second order Butterworth lowpass filter with cutoff frequency of 1.5Hz.To calculate the accelerations of the segments, the recorded displacements and orientation of the segments were differentiated. Since the predicted loads in the hands for the squat lift had an error of ± 1.3 kg, it was reasonable to bin the predicted loads into load categories of 5kg, 1kg, 15kg, 2kg and 25kg. An example of the calculated moments using the binned loads and the filtered kinematic data is shown in Figure 3-4. All computed moments are shown in Appendix E. The calculated moments show that as the load in the hand increases, the moment in the back also increases causing the muscles in the lower back to work harder. Figure 3-4 Moment at the lower back (L4 level) for 5kg, 15kg and 25kg manual loads for subject 12 In summary, a load prediction model was trained and successfully developed for the squat and stoop lift postures. The load prediction error for the squat lift was ± 1.3kg and the stoop lift was ± 2.34kg and the moments in the lower back were calculated. 23

31 Chapter 4 Discussion and Conclusion 4.1 Discussion MAL s in upper extremity and trunk muscles were examined during manual lifting tasks. To do this, specific parameters the mean, peak and area were extracted from the LE s of the recorded EMG. The EMG data were normalized to a baseline determined either from MVC or zero lift LE s or non-normalized EMG data were processed by subtracting the zero load EMG parameters (mean, peak and area). A high correlation between the EMG LE parameters of the biceps brachii, erector spinae at the L4 level and lateral deltoid and the magnitude of the load lifted was found. Using parallel cascade type architecture, a mathematical model was developed, using the zero-normalized EMG LE parameters of the L4, to predict the load in the hands for a squat and stoop lift. The primary significance of this work is that a solution was found to the problem of predicting the load in the hands during manual lifting. Another significant result is that data from a zero load lift is used for EMG normalization rather than MVC. As seen in Table 3-1, the EMG characteristics that were processed using zero-normalization and zero-minus showed better correlation with load than values which were MVC-normalized. Zero-normalization and zero-minus uses information from dynamic contractions to normalize the EMG LE unlike other non-mvc normalization techniques (e.g. Marras and Davis, 21, Mobasser et al. 27) which use static contractions. The dynamic contraction used in this study was the actual motion the subject was required to perform only without the added load in the hands. This type of normalization not only accounts for recording factors (e.g. skin impedance, electrode size and position) but also accounts for any postural differences that may occur when performing the lifts. Another reason for not using MVC normalized data is because producing an 24

32 MVC can be very strenuous for the subject. For the subject to obtain a true MVC, multiple exertions are needed to ensure that it is a maximum exertion. In a previously published work, a non-mvc normalization technique (Marras and Davis, 21) was developed for the ease of the subject. A zero weight lift is a simple task to perform and can be easily performed in the workplace, allowing the load prediction algorithm to be incorporated into a real-time monitoring device. The squat lift prediction model uses the mean, peak and area parameters from the erector spinae (L4 level) EMG as the inputs. Although the L4 zero normalized parameters did not give the highest correlation to weight, it was not a surprise that this muscle gave the best prediction. The erector spinae are a group of muscles that support the spinal column. Since there is minimal flexion and extension in the squat lift, one can conclude that during the squat lift, the erector spinae at the L4 level provided stabilization and support. The heavier the load being lifted, the more the erector spinae will have to work to stabilize and support the spine. Since there is very little movement in the trunk during the squat lift, fewer factors are affecting the way this muscle is working. The stoop lift proved to be a more difficult lift to model. Although a proper prediction could be made, the prediction error for the stoop lift is greater than that of the squat lift. This can be attributed to inconsistencies in how the lift was performed. The kinematic data and EMG LE s show that some subjects lifted the box towards the body before the trunk was straightened and other subjects straightened and pulled the box towards the trunk simultaneously. For example, in Figure 3-3, subject 7 s biceps brachii show high muscle activation at the beginning of the lift and at the placement of the box. Subject 9 s biceps brachii is not activated until well into the lift and stays on till the box is placed on the shelf indicating that the arm motion is one swift movement 25

33 unlike subject 7. Since the actual lift performed was not consistently similar, the data did not provide enough information to make a more accurate prediction. A model for the freestyle lift has yet to be determined. Upon investigation, the freestyle data showed higher variability than the squat and stoop data. Since a model was successfully developed for the squat and stoop data, it is expected that a model can be developed for the freestyle lift posture given that more data can be acquired. Subject anthropometrics play a key role in the effects of lifting weight in the hands. For subjects who are heavier, the weight of a manual load is a smaller percentage of their total body weight and may require less energy and thus lower muscle activation levels MAL s when lifting. However, this does not take into account a subject s ability to lift or his overall fitness level which will also impact the effects of lifting a load in the hands. In this study, data were normalized to minimize the anthropometric effects. It may be possible that load prediction can be improved by including relevant subject characteristics as input variables to the identification algorithm. The kinematic data plays a key role in determining the type of lift being executed. The L4 marker gives the best indication as to which lift posture is being used. In the stoop lift, the L4 marker will show minimal movement vertically (from head to toe) and in the squat lift, there would be significant vertical movement. Thus, the type of lift can be predicted from the kinematic information, and this can also be used as extra input information for an improved predictive model. The net sagittal moments were calculated using Hof s (1992) linked segment model under the assumption that the lifts were performed symmetrically. The motion trackers were placed only on the right arm and it was assumed that the left arm would move symmetrically. Table 3-2 shows 26

34 that this is not necessarily true for all the lifts. This can alter the true net moments as the left arm may not be moving symmetrically with the right arm. Therefore, the calculated moments should be validated. To validate the moment calculations, force plate data collection is needed to compare the existing hands down linked segment model to a feet up linked segment model. 4.2 Conclusion Examination of the forces and moments at the lower back is needed to better understand joint loading and injury. To find the forces and moments associated with manual lifting, the weight of the load being lifted must be known. The model presented here, predicts load lifted in the hands using MAL data from trunk and arm muscles. This information will be incorporated into an online system for detecting peak and cumulative joint loading at the lower back which is being developed as part of a larger study. For the squat lift posture, a nonlinear parallel cascade model was developed to predict load lifted from the EMG LE parameters which were most highly correlated with load. In searching for a general solution with an acceptable error of <±2.5kg, different muscle sites and different parallel cascade models were explored extensively. Although the L4 zero-normalized EMG LE parameters did not have the highest correlation to load lifted, these gave the best results and the model is shown in Figure 3-2. This particular cascade gave the most accurate load prediction with an error of ±1.3kg. For the stoop lift, a model that predicted the weight lifted was developed using the same structure as the squat lift model. It uses the L4 zero-normalized EMG LE parameters as the input and gives a load prediction with maximum error of ±2.34kg. Since the maximum error of the predictions were ±1.3kg and the loads lifted were 5kg, 1kg, 15kg, 2kg and 25kg, the predicted weight could be binned to these specific weights and the moments could be calculated. This can only be done because the loads lifted were known. If the loads lifted were 27

35 unknown, and error <±2.5kg is still acceptable. The loads lifted are in increments of 5kg and the peak bending moment for subject 7 varies by a maximum of 3 N m, this is shown in Appendix D. With a maximum peak bending moment of 22 N m the error in the estimated bending moments is less than 14% of the maximum peak bending moment. 4.3 Future Work / Recommendations There are many ways this study can be further developed to improve load prediction. The study only looked at using peak, mean and area of the EMG LE s as input parameters to the parallel cascade. Other inputs, such as anthropometric data or duration of the lift, can be measured and used. New models and algorithms could be developed to provide a more accurate solution for the different lifts and a solution for the freestyle lift. The lift postures performed in this study were structured and defined. The lifts were considered symmetric, no twisting of the body was done and the feet were stationary during the entire lift. Less rigid lifting could be examined to provide a more practical model. Data from more subjects performing different types of lifts and a larger variety of different weights could provide a more precise prediction and improve on the usability of prediction models. Since there can be a considerable number of different lift style and weight combinations, models can be developed according to specific workplaces. Kinematic data has been paired with EMG data in previous studies to predict force and moments in the back. This could be another way to increase the precision of the prediction model. Investigations into which type of kinematics would be best in combination with EMG could also aide in the prediction of the weight lifted. The ability to predict the weight lifted during manual lifting can be used in monitoring and data collection in the workplace. Using this information, cumulative loading on the lower 28

36 back can be reasonably estimated and limits can be established to prevent the development of lower back pain or injuries in workers. 29

37 References Abdoli-E, M., Agnew, M.J., Stevenson, J.M., An on-body personal lift augmentation device (PLAD) reduces EMG amplitude of erector spinae during lifting tasks, Clinical Biomechanics, vol. 21, pp , 25. Arjmand, N., Shirazi-Adl, A., Model and in vivo studies on human trunk load partitioning and stability in isometric forward flexions, Journal of Biomechanics, vol. 39, pp , 26. Basmajian, J.V., de Luca, C.J., Muscles Alive: Their Functions Revealed by Electromyography, 5 th Edition, Balitimore: Williams and Wilkins, Bazgari, B., Shirazi-Adl, A., Arjmand, N., Analysis of squat and stoop dynamic lifting: muscle forces and internal spinal loads, European Spine Journal, vol. 16, pp , 26. Bureau of Labor Statistics, U.S. Department of Labor, Workplace Injuries and Illnesses, Report USDL-95-58, Washington, DC, De Leva, P., Adjustments to Zatsiorsky-Seluyanov s Segment Inertia Parameters, Journal of Biomechanics, vol. 29, pp , De Looze, M.P., Toussaint, H.M., Van Dieen, J.H., Kemper, H.C.G., Joint moments and muscle activity in the lower extremities and lower back in lifting and lowering tasks, Journal of Biomechanics, vol. 26, pp , Dolan, P., and Adams, M.A., The relationship between EMG activity and extensor moment generation in the erector spinae muscles during bending and lifting activities, Journal of Biomechanics, vol. 26, pp , Granata, K.P., and Marras, W.S., An EMG-assisted model of trunk loading during free-dynamic lifting, Journal of Biomechanics, vol. 28, pp , Hof, A.L., An explicit expression for the moment in multi-body system, Journal of Biomechanics, vol. 25, pp , Korenberg, M., Identifying nonlinear difference equation and functional expansion representations: the fast orthogonal algorithm, Annals of Biomedical Engineering, vol. 16 pp , Korenberg, M., Parallel cascade identification and kernel estimation for nonlinear systems, Annals of Biomedical Engineering, vol. 19 pp , Lavender, S.A., Tsuang, Y.H., Andersson, G.B.J., Hafezi, A., Shin, C.C., Trunk muscle cocontraction: the effects of moment direction and moment magnitude, Journal of Orthopaedic Research, vol 1, pp ,

38 Marras, W.S., and Davis, K.G., A non-mvc EMG normalization technique for the trunk musculature, Journal of Electromyography and Kinesiology, vol. 11, pp. 1-18, 21. McGill, S., The biomechanics of low back injury: Implications on current practice in industry and the clinic, Journal of Biomechanics, vol. 3, pp , Mientjes, M., Norman, R., Wells, R., McGill, S., Assessment of an EMG-based method for continuous estimates of low back compression during asymmetrical occupational tasks, Ergonomics, vol.42, pp , Mobasser, F., Eklund, M., Hashtrudi-Zaad, K. Estimation of elbow-induced wrist force with EMG signals using fast orthogonal search, IEEE Transactions on Biomedical Engineering, vol. 54, pp , 27. National Institute for Occupational Safety and Health (NIOSH). Musculoskeletal Disorders and Workplace Factors: A Critical Review of Epidemiologic Evidence for Work-Related Musculoskeletal Disorders of the Neck, Upper Extremity, and Low Back, NIOSH Publication No , Cincinnati, OH, Sparto, P., Parianpour, M., Marras, W., Granata, K., Reinsel, T., Simon, S., Effect of Electromyogram-force relationships and methods of gain estimation on the predictions of an Electromyogram-driven model of spinal loading, Spine, vol. 23, pp , Winter, D., Biomechanics and Motor Control of Human Movement, 3 rd Edition, New Jersey: John Wiley & Sons,

39 Appendix A Letter of Information and Consent form MEASUREMENT OF MUSCLE ACTIVITY IN THE BACK AND ARMS DURING MANUAL LIFTING Dear You are being invited to participate in a research study being done by the Ergonomics Research Group at Queen s University. We will read through this consent form with you and describe the procedures in detail. Please feel free to ask question at any time. Purposes and Aims of the Study: The purpose of this study is to determine the load that someone is lifting by examining EMG signals and the motions of the body. EMG is a small signal that is emitted by your muscles. In this study, EMG will be recorded from your thoracic erector spinae, lumbar erector spinae (muscles in the back), oblique (stomach muscles), deltoid, bicep and tricep (muscles in the arm). These EMG signals and the motion of your trunk and arms will be measured while you are lifting a box from the ground to a platform and back. This data will be used to determine how your muscles react to lifting different weights. Study Details: Firstly, your height, weight, arm length and trunk height will be measured and recorded. Small sensors will then be attached to the skin on the right side of your back and right arm in the pattern shown on your right. Prior to attaching the sensors, your skin will be cleaned with rubbing alcohol and a non-irritating gel will be applied to your skin to improve sensor contact. Motion sensors will be placed on your hand, forearm, upper arm, neck and lower back. The position of these body segments will be detected by these sensors in the presence of a low power magnet source. To keep all the sensors in place, medical grade stretchy adhesive fabric tape will used. Sensor Attachment EMG Sensors Motion Sensor If you do not have appropriate clothing (shorts and t- shirt), it will be provided for you. 32

40 Procedure: Part 1: Maximum Voluntary Contractions Once the EMG sensors are in place, maximum voluntary contractions will be measured. This part of the experiment is to create a baseline for the EMG measurements by contracting the muscles of interest as hard as you can. For the back muscles, you will be asked to lie down on a bench face down with your legs strapped down to keep the lower body in place. Then you will be asked to arch up to lift your upper body off the bench. To provide resistance, someone will be pushing down on your shoulders to ensure maximum contraction. For the oblique (stomach) muscles, you will be asked to lie down on a bench face up and legs strapped down, and again, with someone pushing down on your shoulders while you do an abdominal crunch. A small twist is needed while you do the stomach crunch in order to target the oblique. For the arm muscles, your arms will be placed on a platform with your elbow bent to a 9 degree angle. You will be asked to pull on a fixed bar towards yourself to flex the bicep muscle. For the triceps, you will push against the bar with your elbow at a 9 degree angle. For the deltoid, while standing with your arm at your side and straight, you will be asked to lift your arm up to about 9 degrees with applied resistance. For the bicep and triceps, the amount of force exerted will be recorded from the load cell. You will then be asked to exert a certain amount of force (less than MVC) and hold it for 3 seconds while EMG is being recorded. This will be done for 3 different force levels. This is done to provide a non-mvc baseline for normalization. Each of these exercises will be performed a total of 2 times each and you will be asked to hold each MVC for 3 seconds. You will take a minimum of 2 minutes break (please ask if you need more time) in between each of the exercises. Part 2: Lifting Exercises After the MVC exercises are completed, the Fastrak motion sensors will be attached and you will be asked to do a series of lifting exercises. Each lift will consist of lifting a box from the floor onto a shelf that is at the height of your waist and then lift the same box from the shelf and put it on the floor. The lift will be done in a stoop, squat and freestyle positions as shown on the right hand side. For each lifting position, you will be asked to lift a box containing different weights. These weights will be 5kg, 1kg, 15kg, 2kg and 25 kg and will be chosen in a random order. So for each chosen position and chosen weight, you will lift the box from the floor and put it on a shelf and then lift the box from the shelf and put it back on the floor. This will be done 3 times in a row. The lifts will be done at your own pace. After each box you will be given a rest period of a minimum of 1-3 minutes depending on the weight lifted. 33

41 Risks of Participation: The risks of participating in this study include skin irritation due to sensitivity under the EMG sensors and muscle discomfort following the high force contractions. If you experience any unpleasant effects, please report this to the research assistant or the principal investigator immediately. You will be reminded to immediately report unpleasant effects at the beginning of each lifting segment. Benefits of Participation: While you will not benefit directly from this study, the results may improve our understanding of how different loads affect the back and in the future may contribute to our understanding of low back pain. You will receive a payment of $1. per hour for participating in the study. Exclusion Criteria: You will not be considered for this study if you: o o o o o o Have restricted back mobility or a history of back problems, including chronic back pain or any other back problems Have a history of dizziness, fainting, irregular heartbeat, or severe headaches Have been diagnosed with carotid or coronary artery disease Have been diagnosed with high blood pressure Have breathing problems Are overweight such that the muscle signals are of poor quality Confidentiality: All information obtained during the course of this study is strictly confidential and your anonymity will be protected at all times. You will be identified by a subject number, not your name. Recorded information will be kept on file under your subject number only. This signed consent form will be stored in a locked file and will be available only to the principle investigator. You will not be identified in any publications or reports. Voluntary Nature of the Study: You participation in this study is completely voluntary. You may withdraw from this study at any time without penalty by indicating to the investigator that you do not wish to continue. You may request that the data collected be discarded. 34

42 Withdrawal of subject by principal investigator: You may be withdrawn from this study if the investigator feels that you are becoming overly fatigued or uncomfortable because of the tasks required or if you experience back pain or discomfort during the procedure. Liability: In the unlikely event you are injured during this study, the investigators are trained in first aid. By signing the consent form, you do NOT waive your legal rights nor release the investigator(s) and sponsors from their legal and professional responsibilities. Subject Statement and Signature: I have read and understand the consent form for this study. The purposes, procedures and technical language have been explained to me. I have been given sufficient time to consider the above information and to seek advice if I choose to do so. I have had the opportunity to ask questions which have been answered to my satisfaction. I am voluntarily signing this form. I will receive a copy of this consent form for future reference. If at any time I have further questions, problems or adverse effects, I can contact the principal investigator, Evelyn Morin at (evelyn.morin@queensu.ca) or Director of the School of Kinesiology and Health Studies Dr. Jean Cote at (jc46@post.queensu.ca). If I have any questions regarding my rights as a research subject I can contact Dr. Albert Clark, Chair of Research Ethics Board at (clarkaf@post.queensu.ca). By signing this consent form, I am indicating that I agree to participate in this study. Signature of Subject Date I understand that I may be photographed during the study. By signing below, I give my consent for use of my photograph in publication and presentation of this work. I understand my identity will be masked in any such publications and presentations. I understand that if I do not sign below, I will not be photographed during the study. Signature of Subject Date By signing this consent form, I confirm that I have carefully explained to the subject the nature of the above research study. I certify that, to the best of my knowledge, the subject understands clearly the nature of the study and the demands, benefits and risks involved to participants in this study. Signature of Witness 35 Date

43 Table A-1 Subject Anthropometrics 36 Subject number Age (years) weight (lbs) height (inches) trunk height (C7 to L4/L5) (inches) elbow to shoulder (inches) elbow to wrist (inches) wrist to knuckle (inches) head and neck (C7 to Top of Head) (inches)

44 Appendix B EMG Linear Envelopes EMG linear envelopes of the Lateral Deltoid, Biceps Brachii and Erector Spinae at the L4 Level for the Second Lift Trial of the Stoop and Squat Lift for all Loads and Subjects. 7 Subject 7 5kg Squat Lift 2 7 Subject 7 5kg Stoop Lift Subject 7 1kg Squat Lift 2 16 Subject 7 1kg Stoop Lift Subject 7 15kg Squat Lift 2 16 Subject 7 15kg Stoop Lift

45 14 Subject 7 2kg Squat Lift 2 16 Subject 7 2kg Stoop Lift Subject 7 25kg Squat Lift 2 16 Subject 7 25kg Stoop Lift Subject 8 5kg Squat Lift 2 6 Subject 8 5kg Stoop Lift Subject 8 1kg Squat Lift 2 7 Subject 8 1kg Stoop Lift

46 12 Subject 8 15kg Squat Lift 2 14 Subject 8 15kg Stoop Lift Subject 8 2kg Squat Lift 2 14 Subject 8 2kg Stoop Lift Subject 8 25kg Squat Lift 2 16 Subject 8 25kg Stoop Lift Subject 9 5kg Squat Lift 2 6 Subject 9 5kg Stoop Lift

47 12 Subject 9 1kg Squat Lift 2 1 Subject 9 1kg Stoop Lift Subject 9 15kg Squat Lift 2 12 Subject 9 15kg Stoop Lift Subject 9 2kg Stoop Lift 2 14 Subject 9 2kg Squat Lift Subject 9 25kg Squat Lift 2 12 Subject 9 25kg Stoop Lift

48 7 Subject 1 5kg Squat Lift 2 6 Subject 1 5kg Stoop Lift Subject 1 1kg Squat Lift 2 12 Subject 1 1kg Stoop Lift Subject 1 15kg Squat Lift 2 14 Subject 1 15kg Stoop Lift Subject 1 2kg Squat Lift 2 14 Subject 1 2kg Stoop Lift

49 16 Subject 1 25kg Squat Lift 2 16 Subject 1 25kg Stoop Lift Subject 11 5kg Squat Lift 2 9 Subject 11 5kg Stoop Lift Subject 11 1kg Squat Lift 2 12 Subject 11 1kg Stoop Lift Subject 11 15kg Squat Lift 2 16 Subject 11 15kg Stoop Lift

50 16 Subject 11 2kg Squat Lift 2 16 Subject 11 2kg Stoop Lift Subject 11 25kg Squat Lift 2 16 Subject 11 25kg Stoop Lift Subject 12 5kg Squat Lift 2 12 Subject 12 5kg Stoop Lift Subject 12 1kg Squat Lift 2 16 Subject 12 1kg Stoop Lift

51 16 Subject 12 15kg Squat Lift 2 16 Subject 12 15kg Stoop Lift Subject 12 2kg Squat Lift 2 16 Subject 12 2kg Stoop Lift Subject 12 25kg Squat Lift 2 16 Subject 12 25kg Stoop Lift

52 Appendix C Lift Characteristics Table C-2 L4 EMG LE Parameter Lift Characteristics L4 MVC L4 MVC L4 MVC Lift Type Duration L4 Area L4 Peak L4 Mean Normalized Normalized Normalized and Load Area Peak Mean 5kg Squat Mean Standard Deviation (SD) Coefficient of Variation (CV) kg Squat Mean SD CV kg Squat Mean SD CV kg Squat Mean SD CV kg Squat Mean SD CV kg Stoop Mean SD CV kg Stoop Mean SD CV kg Stoop Mean SD CV kg Stoop Mean SD CV kg Stoop Mean SD CV

53 L4 Zero L4 Zero L4 Zero L4 Zero Lift Type L4 Zero L4 Zero Normalize Normalized Normalized Minus and Load Minus Area Minus Peak d Area Peak Mean Mean 5kg Squat Mean Standard Deviation (SD) Coefficient of Variation (CV) kg Squat Mean SD CV kg Squat Mean SD CV kg Squat Mean SD CV kg Squat Mean SD CV kg Stoop Mean SD CV kg Stoop Mean SD CV kg Stoop Mean SD CV kg Stoop Mean SD CV kg Stoop Mean SD CV

54 Appendix D Parallel Cascade Coefficients for Squat and Stoop Lift Load Prediction Solution Table D-1 Parallel cascade coefficients for squat lift Cascade Number (k) g k 11 g k 12 g k 21 g k 22 g k 31 g k

55

56

57 5 Table D-2 Parallel cascade polynomial coefficients for squat lift Cascade Number (k) a k 1 a k 2 a k 3 a k 4 a k 5 a k 6 a k 7 a k 8 a k E E E E E E E E E E E+9 2.1E E+9-7.8E E E+9 6.1E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+4 8.5E E E E E E E E E E E+45-2.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+3 2.6E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+29

58 E E E E E E E E E E E E E E E E E E E E E E E E E+42 2.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+4-2.1E E E E E E E E E E E E E E E E E E E E E E E E E E E+3 2.8E E E E E E E E E E E E E E E E E E E E E E E E E E E E+5-6.1E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+52-2.E E E+46-3.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+66

59 E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+77-2.E E E E E E E E E E E E+8 6.9E E E E E E E E E E E E E E E E E E E E E E E E E E+13-5.E E E E E E E E E E E E E E E E E+7 4.5E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+79-4.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+8-2.5E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+9-3.7E E E E E E E E E E E E E E E E E E E E+8 1.3E E E E E E E E E E E E E E E E E E E E E E+73

60 53 Cascade Number (k) a k 1 a k 11 a k 12 a k 13 a k 14 a k 15 a k 16 a k 17 a k E E E E E E E E E E+9 1.3E E E+6-7.1E E E+8 2.3E+7-1.3E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+2-5.7E E E E E E E E E E E E E E E E E E E E E E E+4 5.7E E E E E E E E E E E E E E E E E E E E E E E E E+5-1.9E E E E E E E E E E E E E E E E E E E E E E+2-1.3E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+27 1.E E E+2 5.5E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+3 2.8E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+21-2.E E E E+15

61 E E E E E E E E E E E E E E E E E E E E+33-1.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+47-2.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+3-1.2E E E E+46 3.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+5 3.7E E E E E E E E E E E+4 3.8E E E E E E E E E E E E E E E E E E E E E E E E E+5 6.6E E E E E E E E E E E+4-1.7E E E E E E E E E E E E E+27 1.E E E E E E E+2-3.E E E E E E E+3 6.1E E E E E E E E E E E E E E E E E E E E E E E E E E E E+3 1.4E E E E E E E E E E E E E E E E E E E E+36

62 E E E E+45 6.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+44 3.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+5 5.5E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+5 2.5E E E E E E E E E E E E E E E E E E E+66 8.E E E E E E E E E E E E E E E E E E+6-3.3E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+49 5.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+35

63 56 Cascade Number (k) a k 19 a k 2 a k 21 a k 22 a k 23 a k 24 a k 25 a k E E E E E E E E E E E E E+8-4.1E E E E E E E E E E+13-2.E E E E E E E+9 1.8E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+9-1.8E E E E E E E E E E E E E E E E E E E E E E+8 5.9E E E E E E E E E E+1-4.9E E E E E E E E E E E E E+9-3.6E E E E E E E E E E E E E E E E E E E E E E E+5-1.2E E E E E E E E E+9 3.9E+8-1.9E

64 E E E E E+9 1.5E E E E E E+8 1.4E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+9-9.3E+7 8.4E E E E+19 8.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+14-2.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+9-2.4E E E E E E+8-1.1E E E E E E E E E E E E E E E E E E+8-1.5E E E E E E E E E E E E E E E E E+2-2.9E E E

65 E E E E+2-1.7E E E E E E E E E E E E E+2-2.1E E E E E E E E E E+3 2.8E E E E E E E E E E E E E E E E E E E E E E+15 2.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E

66 Table D-3 Parallel cascade coefficients for stoop lift Cascade Number (k) g k 11 g k 12 g k 21 g k 22 g k 31 g k E

67

68

69 62 Table D-4 Parallel cascade polynomial coefficients for stoop lift Cascade Number (k) a k 1 a k 2 a k 3 a k 4 a k 5 a k 6 a k 7 a k 8 a k 9 a k E E E E E E E E E E E E E E+1 5.2E+1-2.4E+1-2.9E E E E E E E+ 1.33E E E+3-4.2E E E E E E E+5-1.3E E E E E E E E E E E+7-3.5E+7 2.5E E E E+9 1.5E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+5-3.1E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+4 2.7E E E E E E E E E E E E E E E E E E E E E E E E E E E+2-1.1E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+3 1.4E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+39

70 E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+3 3.3E E E E E E E E E E E E E E E E+5 1.5E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+4-1.2E E E E E E E E E E E+48-6.E E E E E E E E E E E E E E E E E E E E E E E+4 6.6E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+4 1.1E E E E E E E E E E E E E E E E E E E E E E E E E+4 3.2E E+35 2.E E E E E E E E E E+46 1.E E+41

71 E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+35 5.E E E E E E E E E E E E E E E E E E E+4 9.2E E E E E E E E E E E E E E E+5 7.8E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+4 5.3E E E E E+5 7.3E E E E E+4-3.1E E+34-1.E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+5-4.8E E E E E E E E E E+5 4.8E E E E E E E E E E+49 8.E E E E E E E E E E E E E E E E E E E+5-1.3E E E E E E E E E E E E E E E E E E E E E+5 2.4E E E E E E E E E E+5-5.4E E+46-4.E E E E E E E E E E E E E E E E E+5-1.9E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+31

72 91-3.3E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+5-2.7E E E E E E E E E+5 3.E E E E E E E E E Cascade Number (k) a k 11 a k 12 a k 13 a k 14 a k 15 a k 16 a k 17 a k 18 a k 19 a k E E-12 2.E E E E E E E E E E E E E E E E E E E E E E+5 2.9E E E E+1 3.7E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+2-3.6E E E E E E E E+7-6.4E E E E E E E E E E+7-5.7E E E E E E E+9 2.8E E E E E+1-6.8E E E E E E E E E E E E E E E E E E E E E E E E E E E E+6-1.9E E E E E E E E E E E E E E E E E E E+9-3.3E E E E E E E E E E E E E E E E E E E+9 1.6E E E+4 6.4E E-1

73 E E+15-6.E E E E E E E E E E E E E+8-2.7E E+5-3.1E E E E+2-3.6E E E E E+1-2.7E E E E E E E E E E E E E E E E E E E E E E+5 2.6E E E E E E E E+1 3.2E E E E E E E E E E E E E+2-4.5E E E E E E E E E+6 8.9E E E E E E E E E E E E E E E E E E E E E+3 5.1E E E E E E E E E E E E E E E E+2 2.6E E E+8 1.6E E E E E E E E E+9 2.9E E+2-1.4E E E E E E E E+1 8.5E E+3-5.2E E E E E E E+9-6.1E E E+1-5.4E E E E E E E E E+6-5.6E E E E E E E+14-2.E E E E+3 3.7E E E E+2 1.5E E E E E E E E E E+2-1.3E E E E E E E E E E E E E E E E E E E E E E E E+9-5.4E+6 2.4E E E E E E E E E E E E E E E E E E E+9 5.E+5 1.8E E E E E E E E E E E+4 1.5E E E E E E E E E+8 7.9E E E E E E E E E E E E E E E E E E E E+8 1.3E E E E E E E E E E E E+7

74 E E E E E E E+1-2.2E E E E E E E E E E E E E E E E E E E E E E+3 4.4E E E E E E E E+7-1.1E E E E E E E E E E E E E E E E E E E E E E+3-6.6E E E E E E E E+1-2.3E+7-1.1E E E E E E E E E E E E E E E E E E E E E+4-1.8E E E E E E E E+9 6.1E E E E E E E E E E+1-2.6E+7 2.1E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+1 1.2E E+2-1.6E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+2-1.4E E E E E E E E E E E E+31-4.E E E E E E E+6-1.6E E E E E E E E E E E+2-5.E E E E E E E E E E E E E E E E E E E E E E+3 1.E E E E E E E+4-2.9E E E E E E E E E E E E E E E E E E E+8-1.3E E+ 2.74E E E E E E E E E E E E E E E E E E+8 9.7E E+ 3.49E-5

75 E E E E E E E E E E E E E E E E E E+2-1.6E E E E E+22 6.E E E E E E+ 1.39E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+8-7.2E+3-1.3E+ 7.33E E E E E E E E E E E E E E E E E E E E+ 1.93E E E E E E E E E E E E E E E E E E+7 3.E E E E E E+21 2.E E E E E E E E E E E E E E E+3 8.6E E E E E E E E E E E-1 4.3E E+28-2.E E E E E E E+3 1.2E E E E E E E E E E E E E E E E E E E E+3 1.3E E E E E E E E E+8-7.8E E+ 8.63E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E+ 1.88E E E E E E E E E E E E E E E E E E E E+ 1.8E E E E E E E E E E E E E E E E E E E E+ 8.78E E E E E E E E E+3-3.2E E-5

76 Appendix E Moment in the Lower Back Center of mass of the segments, segment masses and inertial parameters used in the calculation of moment in the lower back are taken from De Leva (1996) and given in Table D-1. Table E-1 Body segment parameters taken from De Leva (1996 ) 69

77 Moment graphs at the low back (L4 level) for the second lift trial of the stoop and squat lift for all load conditions and subjects are given. 7

78 71

79 72

80 73

81 74

82 75

83 76

84 77

85 78