Subject-specific musculoskeletal modeling in the evaluation of shoulder muscle and joint function

Author’s Accepted Manuscript SUBJECT-SPECIFIC MUSCULOSKELETAL MODELING IN THE EVALUATION OF SHOULDER MUSCLE AND JOINT FUNCTION Wen Wu, Peter V.S. Lee, Adam L. Bryant, Mary Galea, David C. Ackland www.elsevier.com/locate/jbiomech

PII: DOI: Reference:

S0021-9290(16)31010-7 http://dx.doi.org/10.1016/j.jbiomech.2016.09.025 BM7889

To appear in: Journal of Biomechanics Received date: 28 April 2016 Revised date: 30 July 2016 Accepted date: 16 September 2016 Cite this article as: Wen Wu, Peter V.S. Lee, Adam L. Bryant, Mary Galea and David C. Ackland, SUBJECT-SPECIFIC MUSCULOSKELETAL MODELING IN THE EVALUATION OF SHOULDER MUSCLE AND JOINT FUNCTION, Journal of Biomechanics, http://dx.doi.org/10.1016/j.jbiomech.2016.09.025 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

SUBJECT-SPECIFIC MUSCULOSKELETAL MODELING IN THE EVALUATION OF SHOULDER MUSCLE AND JOINT FUNCTION

Wen Wu, Peter V. S. Lee, Adam L. Bryant, Mary Galea, David C. Ackland Department of Mechanical Engineering, University of Melbourne, Parkville, Victoria 3010, AUSTRALIA Submitted as an original article to the Journal of Biomechanics Revision 1 Word count (Introduction to Discussion): 3,899

Address for correspondence: David C. Ackland Department of Mechanical Engineering University of Melbourne Parkville, Victoria 3010, AUSTRALIA Phone: +613 8344 0405 Fax: +613 9347 8784 Email: [email protected]

Key Words: upper limb, biomechanical model, musculotendon parameters, subject-specific

ABSTRACT Upper limb muscle force estimation using Hill-type muscle models depends on musculotendon parameter values, which cannot be readily measured non-invasively. Generic and scaled-generic parameters may be quickly and easily employed, but these approaches do not account for an individual subject’s joint torque capacity. The objective of the present study was to develop a subject-specific experimental testing and modeling framework to evaluate shoulder muscle and joint function during activities of daily living, and to assess the reliability of generic and scaled-generic musculotendon parameters in predicting muscle and joint function. Three-dimensional musculoskeletal models of the shoulders of 6 healthy subjects were developed to calculate muscle and glenohumeral joint loading during abduction, flexion, horizontal flexion, nose-touch and reaching using subject-specific, scaled-generic and generic musculotendon parameters. Muscle and glenohumeral joint forces calculated using generic and scaled-generic models were significantly different to those of subjectspecific models (p<0.05), and task dependent; however, scaled-generic model calculations of shoulder glenohumeral joint force demonstrated better agreement with those of subject specific models during abduction and flexion. Muscles in generic musculoskeletal models operated further from the plateau of their force-length curves than those of scaled-generic and subject-specific models, while muscles in subject-specific models operated over a wider region of their force length curves than those of the generic or scaled-generic models, reflecting diversity of subject shoulder strength. The findings of this study suggest that generic and scaled-generic musculotendon parameters may not provide sufficient accuracy in prediction of shoulder muscle and joint loading when compared to models that employ subject-specific parameter-estimation approaches.

INTRODUCTION The human neuromuscular system has a remarkable ability to modulate shoulder muscle function during the diverse range of daily tasks of the upper-limb that most of us accomplish with ease. The inability to directly measure muscle loading non-invasively in humans poses a major challenge in understanding neuromuscular control of upper limb movement. The shoulder complex is a highly mechanically redundant system: over 26 muscle-tendon units span the glenohumeral joint alone (Garner and Pandy, 2001). As a consequence, it is impossible to determine a unique set of muscle contributions to joint movement, since a net joint torque can be produced by an infinite combination of muscle forces. Computational modeling is an established strategy to estimate muscle and joint loading during upper limb movement, and has led to significant advances in our understanding of ergonomic practice (Dickerson, Chaffin and Hughes, 2007), shoulder implant behavior (Kontaxis and Johnson, 2009), surgical planning (Magermans et al., 2004) and rehabilitation (Saul et al., 2011). In Hill-type muscle models, musculotendon force is defined by values of the musculotendon parameters, including optimum fiber length, tendon slack length, maximum isometric muscle force and maximum fiber shortening velocity, which vary with subject size, gender, age and neuromuscular condition (Thelen, 2003, Piazza, 2006). Musculotendon parameters employed in musculoskeletal models may be generic, scaled from a generic parameter dataset (scaled-generic) or subject-specific. Generic musculotendon parameters are typically taken from the literature, often from multiple cadaveric measurements, and are representative of a generic subject (van Drongelen et al., 2005; Dubowsky et al., 2008; Nikooyan et al., 2010). While the use of generic musculotendon parameters provides some convenience in estimating muscle and joint loading using a musculoskeletal model, determining the most appropriate parameters for a subject, or

cohort, can be difficult due to the wide range of reported values, and the sensitivity of model behavior to small changes in parameter values (Scovil and Ronsky, 2006; Ackland, Lin and Pandy, 2012a; Navacchia et al., 2016). Scaled-generic musculotendon parameters are obtained by scaling a set of generic musculotendon parameters to a given subject, and are thought to improve on estimates of muscle and joint loading using generic musculoskeletal models. One of the most widely used scaling strategies is to resize both the optimal fiber length and tendon slack length by a constant ratio to match the predicted total muscle-tendon unit length at a given joint position (Delp et al., 2007). This method attempts to reproduce the actuator operating range on the force-length curve defined in the generic subject model. An alternative strategy is to use optimization to calculate the optimal fiber length and tendon slack length by minimizing differences in fiber operating range between the generic and scaled model through a range of joint motions (Winby, Lloyd and Kirk, 2008; Modenese et al., 2015). However, since these strategies do not take into consideration a subject’s capacity to generate joint torque, which can vary greatly with task and from person to person (Holzbaur et al., 2007a), some parameters values may be sub-optimal and potentially result in erroneous muscle and joint loading prediction. The degree to which this undesirable phenomenon occurs is currently not well understood. Subject-specific musculotendon parameters may be calculated using musculotendon parameter optimization, or model parameter calibration, which estimate a subject’s musculotendon parameter values using a series of functional calibration tasks performed in vivo, including isometric and isokinetic contractions (Garner and Pandy, 2003). Parameters are optimized to the subject’s individual torque capacity, resulting in more subjectrepresentative musculoskeletal model behaviour; however, calibration tasks can be difficult to set up experimentally, and time-consuming and tiring for the subject. The specific aims of

this study were twofold. Firstly, to develop a subject testing protocol and musculoskeletal modeling framework to non-invasively evaluate shoulder muscle and glenohumeral joint forces during activities of daily living; and secondly, to benchmark model predictions of shoulder muscle and glenohumeral joint loading using generic, scaled-generic musculotendon parameters with gold-standard subject-specific musculotendon parameters. This study seeks to elucidate the reliability of generic and scaled-generic musculotendon parameters in musculoskeletal model predictions of human shoulder muscle and joint function. The results will assist in development of accurate and representative musculoskeletal models of the upper limbs of human subjects.

MATERIALS AND METHODS Experimental protocol Upper limb experiments were performed on six healthy male participants, mean age: 28.3 years (range: 25–38 years), mean mass: 68.2 kg (range: 56-85kg), mean height 1.73 m (range: 1.70-1.75m). Joint motion and muscle electromyographic (EMG) data were recorded simultaneously during five tasks with the dominant upper limb (Table 1). These included abduction, flexion and horizontal-flexion in the coronal, sagittal and transverse planes, respectively, as well as reaching and nose touching. These tasks form the basis of many activities of daily living including eating, dressing and personal grooming. In addition, four maximal voluntary isokinetic contractions (MVIKCs) at two speeds and six maximum voluntary isometric contractions (MVIMCs) were performed on a Biodex dynamometer (Biodex Pro 4, Shirley, NY) (Table 1). A number of practice trials on the dynamometer were taken by subjects for familiarization before the tests. Subjects were seated and secured for all trials, and each task was repeated three times. Rest periods of at least 30 seconds were given between trials.

During testing, three-dimensional locations of retro-reflective markers attached to each subject’s upper limb were measured using an 8-camera, video-based motion capture system (Vicon, Oxford Metrics Ltd., Oxford). Retro-reflective markers were attached to the bony landmarks recommended by the International Society of Biomechanics (Wu et al., 2005), with additional markers attached to the acromion process, olecranon process, middle of the dorsal side of wrist, and the inferior aspect of the 11th rib. A marker cluster was attached to the scapula spine using a customized 3D-printed clamp, and used to track scapula motion (van Andel et al., 2009). Marker trajectories were sampled at 200 Hz, and filtered using a fourth-order, low-pass Butterworth filter with a 10 Hz cut-off frequency. A two-stage calibration method was used to minimize the skin artefact of the scapula during upper limb motion (Brochard, Lempereur and Rémy-Néris, 2011). Pairs of pre-amplified EMG surface electrodes (Cometa, Bareggio, Italy) were attached to the upper limb to record activity from eleven muscles including the trapezius (upper and lower), deltoid (anterior, middle, and posterior), infraspinatus, pectoralis major (upper and lower), middle latissimus dorsi, biceps brachii and triceps brachii (Dickerson, Hughes and Chaffin, 2008). Skin overlying these muscles was prepared via shaving and debridement of the superficial layer. EMG data were sampled at 2,000 Hz, filtered using a second order high-pass Butterworth filter with cut-off frequency of 25 Hz, and rectified (Langenderfer et al., 2005). Muscle EMG data normalized to maximum values from the MVIMCs (Halaki and Ginn, 2012) were used to assess the timing of muscle activation. Approval for the study was obtained from the relevant ethics committee and all participants provided written informed consent prior to participation. Musculoskeletal modeling A generic 5-segment, 10-degree-of-freedom (DOF) musculoskeletal model of the upper limb was developed in OpenSim (Figure 1) (Delp et al., 2007) and initially used to

perform inverse kinematics and inverse dynamics to obtain joint angles and joint moments from the measured marker trajectory data. The forearm and wrist were modelled as one rigid segment that articulated with the elbow via a 2-DOF universal joint, the glenohumeral and acromioclavicular joints were modelled as 3-DOF constrained ball and socket joints, while the sternoclavicular joint was modelled as a 2-DOF universal joint. Segment mass and length properties were scaled to each subject’s anthropometry using marker data and subject mass information (Delp et al., 2007). The model was actuated by 26 Hill-type muscle-tendon units representing the major axioscapular, axiohumeral and scapulohumeral muscle groups. Muscle paths were initially determined for the neutral arm configuration using the anatomy of the Visible Human Male (Garner and Pandy, 2001). An optimization procedure was then developed to determine optimal muscle paths and wrapping ‘via point’ locations in order to match the model’s muscle moment arms with those measured on 8 cadaveric upper extremities (see Supplementary Material for details). Muscle moment arms were subsequently scaled to each subject’s measured anthropometry (Hamner, Seth and Delp, 2010). Subject-specific musculoskeletal models were then developed using the structure of the generic upper limb model. Segment mass, anthropometry and inertial properties, as well as musculotendon unit lengths (  ) and moment arms, were firstly scaled to subject segment lengths using the locations of bony landmarks (Delp et al., 2007). Musculotendon parameters for each model were then determined using a two-stage optimization process. In the first stage, optimal fiber lengths ( ! ) and tendon slack lengths ("# ) were calculated, where tendon strain was assumed negligible for the movements reported in the present study. For a given muscle, the maximum muscle force, $ ! , during each subject’s MVIMC was a function of the muscle’s force-length curve, as described previously (Zajac, 1989) $ ! = %" (  , &, $ ! ,  ! , "' )

(2)

where,   was calculated from the model at each joint position, and muscle pennation angle, &, was obtained from a previous study (Langenderfer et al., 2004). Maximum fiber force, $ ! , was estimated from $ ! = +-

./

(3)

01/

where specific tension, +- , was assumed to be 865 kPa (Garner and Pandy, 2001; Holzbaur, Murray and Delp, 2005). Muscle volume, 2! , was obtained from regression relations described by Holzbaur et al (2007) (Holzbaur et al., 2007b). Simulated shoulder joint torques during MVIMCs (345"5! ) were calculated by summing the product of $"! and moment arm values for each muscle.  ! and "# were then calculated by minimising the objective function "5! 6( ! , "# ) = ∑@ − 345:;< > 5A48345 :;<

where, 6 is the objective function, and 345

?

(4)

represents the experimental torque values for the

B th MVIMC trial through to the total number (N) of MVIMC trails. The interior-point optimizer in Matlab and the ‘GlobalSearch’ method was used to ensure the global minimum was reached. "# values were constrained between 0 and the minimal   throughout all motions.  ! values were constrained between 50% and 200% of optimal fiber length (Langenderfer et al., 2004), and normalized fiber length, which is a function of  ! and "# , was constrained to values no larger than 1.25. In the second stage of the parameter optimization, the maximum shortening velocity (2 ! ) of each muscle was calculated. Given the parameters obtained in the first parameter optimization stage, the dynamic maximum muscle force during each subject’s MVIKC was subject to the muscle’s force-velocity curve as described previously (Zajac, 1989), ̇ , 2!> $ ! = %C 8 

(5)

̇  is the velocity of the muscle-tendon unit and calculated from the model at each where,   joint position and angular velocity. 2 ! was optimized using Matlab’s interior-point optimizer

in the same manner as described in stage one. This was achieved by minimizing the square of difference between the simulated MVIKC torques and experimentally measured MVIKC torques. To assist in convergence, the ratio between 2 ! and  ! for each muscle was assumed the same (Delp et al., 2007). The calculated musculotendon parameters for each subject were used to create goldstandard subject-specific musculoskeletal models with which to benchmark generic and scaled-generic musculoskeletal models (Figure 2). For the generic models, one set of musculotendon parameters defined by the heaviest test subject (i.e. body mass 85kg, height 1.73m) were employed to ensure the models created were ‘strong’ enough to represent all subsequent subjects. In the case of scaled-generic models,  ! and "' were scaled using two approaches. First  ! and "' were scaled in equal proportion to match the muscle-tendon unit length with the upper limb in the neutral position (scaled-generic-1) (Delp et al., 2007). Second, a least squares optimization described by Modenese et al. (2016) was used to calculate  ! and "' by mapping the normalized actuator operating range in the generic model to the scaled model over the entire range of measured joint positions (scaled-generic-2). Subject-specific, generic, and scaled generic models (scaled-generic-1 and scaled-generic-2) were used to calculate muscle and glenohumeral joint forces for each subject during abduction, flexion, horizontal flexion, reaching and nose touching. Muscle forces were calculated using static optimisation by minimizing the sum of the squares of muscle activations (Anderson and Pandy, 2001) (Figure 2). An anisotropic elliptical shape ratio was used as a constraint on the calculated glenohumeral joint force to simulate rotator cuff muscle co-contraction (Lippitt and Matsen, 1993). Resultant and orthogonal components of the glenohumeral joint force were calculated in Matlab using the OpenSim application programming interface (API), and were expressed with respect to the scapula coordinate system (see Supplementary Material for details).

Data analysis Two-way repeated-values analysis of variance (ANOVA) was used to assess the effect of model type on calculations of muscle and joint loading. Specifically, multiple twoway ANOVAs were performed to compare each muscle force and joint force component predicted by the generic and subject-specific models for each task. The dependent variables in each ANOVA were muscle force or glenohumeral joint force, while the independent variables were joint position and model type (generic vs subject-specific). The two-way ANOVAs were repeated in a similar manner to compare scaled-generic-1 and subjectspecific model outputs, as well as scaled-generic-2 and subject-specific model outputs. Analysis of variance was also used to assess interactions between the independent variables. Level of significance was defined as p < 0.05. All muscle and joint forces were normalized and expressed as a percentage of subject body weight.

RESULTS Muscle forces calculated using generic and scaled-generic musculoskeletal models were significantly different to those calculated using subject-specific models for all muscles (Figure 3), with differences in force calculations task-dependent and influenced by joint position (Table 2). For example, during abduction, there were significant differences in middle deltoid, supraspinatus, infraspinatus, and subscapularis forces between generic and subject-specific models, as well as between scaled-generic and subject-specific models (all p<0.001); however, during flexion, there were no significant differences in middle deltoid (p=0.789) and supraspinatus forces (p=0.268) between scaled-generic-2 models and subjectspecific models (Table 2). Across all tasks, there were vastly more significant differences in muscle forces calculated between generic and subject-specific models, than there were between scaled-generic and subject-specific models. Neither models scaled using scaled-

generic-1 nor scaled-generic-2 showed notably better capacity to predict subject-specific muscle forces. With the exception of infraspinatus and glenohumeral joint force calculations during abduction and flexion, respectively, there were no significant interaction effects demonstrated between model type and joint position (p>0.05) (Table 2). Generic musculoskeletal models tended to underestimate muscle force calculations in all tasks. For example, the subscapularis muscle forces calculated using generic models were significantly lower than those calculated using subject-specific models [mean difference: 2.27%BW (p<0.001), 2.92%BW (p<0.001), 4.15%BW (p<0.001), 1.75%BW (p<0.001), 1.03%BW (p<0.001) for abduction, flexion, horizontal flexion, nose touching and reaching, respectively]. In most cases there were greater differences with subject-specific model muscle force calculations when using generic model calculations compared to when using scaledgeneric model calculations. There were some exceptions, however. For example, during reaching, the differences in calculated infraspinatus force between the subject-specific models and that of the generic, scaled-generic-1, and scaled-generic-2 models was 0.08%BW (p=0.824), 2.96%BW (p<0.001) and 2.63%BW (p<0.001), respectively (Table 2). Glenohumeral joint forces calculated with generic and scaled-generic models showed significant differences to those calculated using subject-specific models (Figure 4 and Table 2), as did muscle contributions to the glenohumeral joint force (see Supplementary Material). The resultant glenohumeral joint force, and its compressive, anterior and superior components, was most poorly predicted using generic models, exhibiting significant differences to subject-specific model calculations during abduction [-12.63%BW (p<0.001), 11.98%BW (p<0.001), -4.04%BW (p<0.001) and 2.87%BW (p<0.001), respectively] and flexion [-13.77%BW (p<0.001), -13.41%BW (p<0.001), -5.12%BW (p<0.001), 1.55%BW (p=0.005),

respectively]

(see

Supplementary

Material).

Scaled-generic-2

models

demonstrated better glenohumeral joint force agreement with subject-specific models than

generic models. In the case of flexion, there were no significant differences in the calculated resultant, compressive and inferior glenohumeral joint force between scaled-generic-2 and subject-specific models (p>0.05). Muscles in generic musculoskeletal models tended to operate further down the ascending region of their force-length curve than those of scaled-generic and subject-specific models during abduction and flexion (Figure 5). In general, muscles in both scaled-generic models operated on similar regions of their force-length curves. Muscles in the subjectspecific models operated over a wider region of their force length curves than those of the generic or scaled-generic models.

DISCUSSION Developing subject-specific musculoskeletal models using Hill-type muscle-tendon actuators can be challenging, since musculotendon parameters are known to vary greatly between individuals and cannot be measured in vivo. The objective of this study was to develop a subject-specific testing protocol and modeling framework for evaluating shoulder muscle and joint function during activities of daily living, and to assess the use of generic, scaled-generic and subject-specific musculotendon parameters on model calculations of muscle and joint loading. Our study provides evidence that musculoskeletal models with generic or scaled-generic musculotendon parameters generate muscle force predictions that are substantially different to those of calibrated models with subject-specific musculotendon parameters. These differences in muscle force calculations may be due to the inability of the generic and scaled generic models to represent each subject’s maximum isometric muscle force capacity. Un-scaled $ ! values were employed in the generic and scaled-generic type models, while $ ! in subject-specific models were calibrated to each subject’s joint torque capacity. This study also demonstrates that joint kinematics has a significant influence on

calculated muscle and joint forces, since muscle force requirements were shown to vary greatly with shoulder elevation and rotation through different tasks; however, there was generally little interaction between the effects of joint position and model type. Generic and scaled-generic upper limb models predicted different operating regions on muscle force-length curves than those of subject-specific models (Figure 5), resulting in disparate muscle force predictions. Muscles in generic models tended to operate further from the plateau down the ascending region of the force-length curve than those in subject-specific models, since their optimal muscle fiber lengths and tendon slack lengths were not calibrated to each subject and were assumed constant. As the active force generating capacity of a muscle decreased substantially when its normalized fiber length approaches 0.5 or 1.5, muscles in generic models risk running out of active muscle strength, particularly when a subject’s segment lengths, and therefore muscle-tendon lengths, are substantially different to those of the generic subject. While subject-specific models exhibited greater variation in muscle-tendon parameters and muscle force-length curve operating regions, reflecting diversity of measured strength and torque capacity of each subject, scaled-generic models tended to maintain a consistent operating region on the force-length curve of muscles across all subjects. This was the primarily reason for the muscle force magnitude differences observed between the subject-specific and scaled-generic models (Figure 5). The present study showed that using both scaled-generic-1 and scaled-generic-2 musculotendon parameter, resultant muscle and glenohumeral joint force predictions, as well as individual muscle contributions to glenohumeral joint forces, were similar across all movements (see Table 2 and Figure 4). The two scaled-generic models were shown to have similar operating regions on each muscle’s force-length curve for all tasks, resulting in comparable muscle force generating capacity (Figure 5). For a given subject, the scaledgeneric-1 method resized  ! and "' values from the generic model to match the subject using

estimates of the subject’s   . Values of   were estimated by scaling the generic model segments to the subject’s anthropometry with the upper limb in the neutral position. A limitation of this method is that when bones of muscle attachments are scaled by different factors, a disparity between the  ! and "' may arise, resulting in different operating regions on the force-length curve to that of the generic subject. The scaled-generic-2 method overcomes this shortcoming by optimising  ! and "' over a wide range of joint positions, and was shown to greatly improve glenohumeral joint force estimates during flexion toward subject-specific values. Glenohumeral joint forces calculated using scaled-generic models were similar in magnitude to those calculated using subject-specific models, and were substantially different to generic model calculations. For example, the average glenohumeral joint force magnitude difference between the subject-specific model and the scaled-generic-1 model calculations was -0.23%BW during nose-touching, while a 6.25%BW difference was reported between subject-specific and generic model calculations. In a number of tasks, glenohumeral joint force calculations were less sensitive to model type than muscle force calculations (Table 2) and muscle contributions to glenohumeral joint force calculations (Supplementary Material). This is likely due to the glenohumeral joint force calculation from the sum of muscle force vectors, as well as dynamic coupling, which may attenuate high sensitivity to model type observed in some muscles. The objective function constraint to maintain the glenohumeral joint force within the boundary of the glenoid, employed to simulate muscle rotator cuff coactivation and glenohumeral joint stability by concavity compression, may have also influenced between-model differences in muscle recruitment patterns, since this constraint has a substantial influence on model predictions of muscle and joint function (see Supplementary Material).

The subject-specific musculotendon parameters and model calculations in the present study are in reasonable agreement with previously reported studies. For example, the calculated average optimal muscle fiber length of middle deltoid, subscapularis and infraspinatus was 13.18cm, 16.32cm and 13.23cm, respectively (Table 4), while previous measurements were 12.97cm, 13.32cm and 10.66cm, respectively (Langenderfer et al., 2004). Our subject-specific model predictions of muscle force compare to those of previously reported generic models(Karlsson and Peterson, 1992; Yanagawa et al., 2008). For instance, our peak total deltoid and infraspinatus forces during abduction (48%BW and 22%BW, respectively) are similar to those of Yanagawa et al. (2008) (50%BW and 14%BW, respectively) and Karlsson and Peterson (1992) (37%BW and 25%BW, respectively). The timing of our calculated muscle forces was in reasonable agreement with our measured muscle EMG (Figure 3), and the pattern of muscle recruitment observed is indicative of prominent co-contraction effects that may ultimately act to stabilize the glenohumeral joint. For example, while the subscapularis had an adduction moment arm, we observed its significant activation during abduction

(Supplementary Material). Our calculated

glenohumeral joint force magnitudes are in close agreement with instrumented shoulder implant data for abduction and flexion (Bergmann et al., 2007) (Supplementary Material). There are several limitations that should be noted. The scapulothoracic muscles were not included in the musculotendon parameter optimization, since the glenohumeral joint torque was assumed to be primarily governed by the activity of the glenohumeral and axiohumeral muscles. In addition, the ligaments of the shoulder girdle were not included in the model for the simulated activities of daily living; however, previous studies indicate that the glenohumeral ligaments do not contribute significantly to the joint stability and joint moment except for at the end-range of joint motion (Engin, 1980). We neglected shoulder translations and modeled the glenohumeral joint as a ball-and-socket articulation; however,

translation of the natural shoulder during elevation has been reported as 2mm (Graichen et al., 2000), and is unlikely to substantially influence shoulder muscle force-length properties. Finally, only selective isometric and isokinetic contractions were used in model calibration tasks to minimize subject testing time and fatigue, and a different set of tasks may improve parameter estimation and muscle force calculation. In

conclusion,

this

study

presented

a

subject-specific

experimental

and

musculoskeletal modeling framework for evaluation of shoulder muscle and joint function during activities of daily living. Significant differences in calculated muscle and glenohumeral joint forces were observed between generic, scaled-generic and subject-specific models during different activities of daily living. The findings suggest that the use of generic and scaled-generic musculotendon parameters do not reproduce the muscle loading predicted by subject-specific parameters, however, scaled-generic models provide substantially better prediction of glenohumeral joint loading than generic models. The present study suggests that accurate models of a subject’s muscle and joint function should employ a series of functional calibration tasks in order to encapsulate the force-producing capacity of each muscle-tendon unit. The subject-specific musculotendon parameter estimation approaches reported may be used to improve reliability of generic and scalable upper limb musculoskeletal model predictions of muscle and joint loading.

ACKNOWLEDGEMENTS This study was partially funded by a Victorian Orthopaedic Research Trust grant. Bryant is a recipient of an NHMRC Career Development Fellowship (R.D.Wright Biomedical, # 1053521).

CONFLICT OF INTEREST The authors do not have any financial or personal relationships with other people or organizations that could inappropriately influence this manuscript.

REFERENCES Ackland, D.C., Lin, Y.C. and Pandy, M.G., 2012a. Sensitivity of model predictions of muscle function to changes in moment arms and muscle-tendon properties: A Monte-Carlo analysis. Journal of Biomechanics, 45, 1463–1471. Ackland, D.C., Pak, P., Richardson, M. and Pandy, M.G., 2008. Moment arms of the muscles crossing the anatomical shoulder. Journal of Anatomy, 213, 383–390. Ackland, D.C., Richardson, M. and Pandy, M.G., 2012b. Axial rotation moment arms of the shoulder musculature after reverse total shoulder arthroplasty. The Journal of bone and joint surgery. American volume, 94, 1886–95. van Andel, C., van Hutten, K., Eversdijk, M., Veeger, D. and Harlaar, J., 2009. Recording scapular motion using an acromion marker cluster. Gait and Posture, 29, 123–128. Anderson, F.C. and Pandy, M.G., 2001. Static and dynamic optimization solutions for gait are practically equivalent. Journal of Biomechanics, 34, 153–161. Bergmann, G., Graichen, F., Bender, A., Kääb, M., Rohlmann, A. and Westerhoff, P., 2007. In vivo glenohumeral contact forces-Measurements in the first patient 7 months postoperatively. Journal of Biomechanics, 40, 2139–2149. Brochard, S., Lempereur, M. and Rémy-Néris, O., 2011. Double calibration: An accurate, reliable and easy-to-use method for 3D scapular motion analysis. Journal of Biomechanics, 44, 751–754. Delp, S.L., Anderson, F.C., Arnold, A.S., Loan, P., Habib, A., John, C.T., Guendelman, E. and Thelen, D.G., 2007. OpenSim: Open-source software to create and analyze dynamic simulations of movement. IEEE Transactions on Biomedical Engineering, 54, 1940– 1950. Dickerson, C.R., Chaffin, D.B. and Hughes, R.E., 2007. A mathematical musculoskeletal shoulder model for proactive ergonomic analysis. Computer methods in biomechanics and biomedical engineering, 10, 389–400. Dickerson, C.R., Hughes, R.E. and Chaffin, D.B., 2008. Experimental evaluation of a computational shoulder musculoskeletal model. Clinical Biomechanics, 23, 886–894. van Drongelen, S., van der Woude, L.H., Janssen, T.W., Angenot, E.L., Chadwick, E.K. and Veeger, H.E., 2005. Mechanical load on the upper extremity during wheelchair ADL. Archives of Physical Medicine and Rehabilitation, 86, 1214–1220. Dubowsky, S.R., Rasmussen, J., Sisto, S.A. and Langrana, N.A., 2008. Validation of a musculoskeletal model of wheelchair propulsion and its application to minimizing shoulder joint forces. Journal of Biomechanics, 41, 2981–2988. Engin, A., 1980. On the biomechanics of the shoulder complex. Journal of Biomechanics, 13, 575–590. Garner, B.A. and Pandy, M.G., 2001. Musculoskeletal model of the upper limb based on the visible human male dataset. Computer methods in biomechanics and biomedical engineering, 4, 93–126. Garner, B.A. and Pandy, M.G., 2003. Estimation of musculotendon properties in the human upper limb. Annals of Biomedical Engineering, 31, 207–220.

Graichen, H., Stammberger, T., Bonel, H., Karl-Hans Englmeier, Reiser, M. and Eckstein, F., 2000. Glenohumeral translation during active and passive elevation of the shoulder - A 3D open-MRI study. Journal of Biomechanics, 33, 609–613. Halaki, M. and Ginn, K. a., 2012. Normalization of EMG Signals: To Normalize or Not to Normalize and What to Normalize to? Computational Intelligence in Electromyography Analysis - A Perspective on Current Applications and Future Challenges, 175–194. Hamner, S.R., Seth, A. and Delp, S.L., 2010. Muscle contributions to propulsion and support during running. Journal of Biomechanics, 43, 2709–2716. Holzbaur, K.R.S., Delp, S.L., Gold, G.E. and Murray, W.M., 2007a. Moment-generating capacity of upper limb muscles in healthy adults. Journal of Biomechanics, 40, 2442– 2449. Holzbaur, K.R.S., Murray, W.M. and Delp, S.L., 2005. A model of the upper extremity for simulating musculoskeletal surgery and analyzing neuromuscular control. Annals of Biomedical Engineering, 33, 829–840. Holzbaur, K.R., Murray, W.M., Gold, G.E. and Delp, S.L., 2007b. Upper limb muscle volumes in adult subjects. J Biomech, 40, 742–749. Karlsson, D. and Peterson, B., 1992. Towards a model for force predictions in the human shoulder. Journal of Biomechanics, 25, 189–199. Kontaxis, A. and Johnson, G.R., 2009. The biomechanics of reverse anatomy shoulder replacement - A modelling study. Clinical Biomechanics, 24, 254–260. Kuechle, D.K., Newman, S.R., Itoi, E., Morrey, B.F. and An, K.N., 1997. Shoulder muscle moment arms during horizontal flexion and elevation. Journal of Shoulder and Elbow Surgery, 6, 429–439. Langenderfer, J., Jerabek, S.A., Thangamani, V.B., Kuhn, J.E. and Hughes, R.E., 2004. Musculoskeletal parameters of muscles crossing the shoulder and elbow and the e ect of sarcomere length sample size on estimation of optimal muscle length. Brand, 19, 664– 670. Langenderfer, J., LaScalza, S., Mell, A., Carpenter, J.E., Kuhn, J.E. and Hughes, R.E., 2005. An EMG-driven model of the upper extremity and estimation of long head biceps force. Computers in Biology and Medicine, 35, 25–39. Lippitt, S. and Matsen, F., 1993. Mechanisms of glenohumeral joint stability. Clinical orthopaedics and related research, 291, 20–28. Magermans, D.J., Chadwick, E.K.J., Veeger, H.E.J., Rozing, P.M. and Van Der Helm, F.C.T., 2004. Effectiveness of tendon transfers for massive rotator cuff tears: A simulation study. Clinical Biomechanics, 19, 116–122. Modenese, L., Ceseracciu, E., Reggiani, M. and Lloyd, D.G., 2015. Estimation of musculotendon parameters for scaled and subject specific musculoskeletal models using an optimization technique. Journal of Biomechanics, 49, 141–148. Navacchia, A., Myers, C.A., Rullkoetter, P.J. and Shelburne, K.B., 2016. Prediction of In Vivo Knee Joint Loads Using a Global Probabilistic Analysis. Journal of Biomechanical Engineering, 138, 031002. Nikooyan, A.A., Veeger, H.E.J., Westerhoff, P., Graichen, F., Bergmann, G. and van der

Helm, F.C.T., 2010. Validation of the Delft Shoulder and Elbow Model using in-vivo glenohumeral joint contact forces. Journal of Biomechanics, 43, 3007–3014. Saul, K.R., Hayon, S., Smith, T.L., Tuohy, C.J. and Mannava, S., 2011. Postural dependence of passive tension in the supraspinatus following rotator cuff repair: A simulation analysis. Clinical Biomechanics, 26, 804–810. Scovil, C.Y. and Ronsky, J.L., 2006. Sensitivity of a Hill-based muscle model to perturbations in model parameters. Journal of Biomechanics, 39, 2055–2063. Winby, C.R., Lloyd, D.G. and Kirk, T.B., 2008. Evaluation of different analytical methods for subject-specific scaling of musculotendon parameters. Journal of Biomechanics, 41, 1682–1688. Wu, G., Van Der Helm, F.C.T., Veeger, H.E.J., Makhsous, M., Van Roy, P., Anglin, C., Nagels, J., Karduna, A.R., McQuade, K., Wang, X., Werner, F.W. and Buchholz, B., 2005. ISB recommendation on definitions of joint coordinate systems of various joints for the reporting of human joint motion - Part II: Shoulder, elbow, wrist and hand. Journal of Biomechanics, 38, 981–992. Yanagawa, T., Goodwin, C.J., Shelburne, K.B., Giphart, J.E., Torry, M.R. and Pandy, M.G., 2008. Contributions of the individual muscles of the shoulder to glenohumeral joint stability during abduction. Journal of biomechanical engineering, 130, 021024. Zajac, F.E., 1989. Muscle and tendon: properties, models, scaling, and application to biomechanics and motor control. Critical reviews in biomedical engineering, 17, 359– 411.

FIGURE CAPTIONS Figure 1:

Anterior (A) and posterior view (B) of the musculoskeletal shoulder model illustrating the major bony segments and muscle paths.

Figure 2:

Schematic diagram of modelling framework used in the present study. Shoulder muscle and glenohumeral joint forces for six subjects were computed during five upper limb tasks using musculoskeletal models that employed subject-specific,

generic,

scaled-generic-1

and

scaled-generic-2

musculotendon parameters. The moment arms, muscle-tendon lengths (  ), segment lengths and mass properties of the models were scaled to each subject’s mass and anthropometry measurements. A two-stage parameter optimization process was used to compute subject-specific musculotendon parameters from a series of maximum voluntary isometric contractions (MVIMCs) and maximal voluntary isokinetic contractions (MVIKCs). A generic set of musculotendon parameters were obtained and scaled in two ways: by maintaining constant ratios of optimal fiber length ( ! ) and tendon slack length ("# ) with the shoulder in the neutral position (scaled-generic-1) and by maintaining constant ratios of  ! and "# through a range of dynamic upper limb motions (scaled-generic-2). Figure 3:

Average forces calculated for selected muscles during abduction, flexion, horizontal-flexion, nose touching and reaching. Data are shown for the anterior deltoid, middle deltoid, supraspinatus, infraspinatus, subscapularis, and middle pectoralis major. Solid black lines represent data from subjectspecific models; dashed lines represent data from generic models; grey lines represent data from scaled-generic-1 models (Delp et al., 2007); thin black

lines indicate data scaled-generic-2 models (Modenese et al., 2015); while Solid blue bars indicate muscle activity recorded from EMG. Figure 4:

Average resultant glenohumeral joint force and joint force components in the compressive, anterior and inferior directions during abduction and flexion. Solid black lines represent data from subject-specific models; dashed lines represent data from generic models; grey lines represent data from scaledgeneric-1 models 1(Delp et al., 2007); while thin black lines represent data from scaled-generic-2 models (Modenese et al., 2015).

Figure 5:

Box-plots indicating muscle fiber operating ranges of all subjects using generic (G), scaled-generic-1(SG-1) (Delp et al., 2007), scaled-generic-2(SG-2) (Modenese et al., 2015), and subject-specific (SS) models for anterior deltoid, middle deltoid, supraspinatus, infraspinatus, subscapularis, middle pectoralis major during abduction and flexion. The central mark, edges of the box, and box whiskers represent the median, the 25th and 75th percentiles, and the maximum/minimum data points, respectively. Black curves indicate the maximum normalized active muscle force, while gray curves represent the active passive muscle force.

Figure 1

A

B

Generic musculoskeletal model structure

Figure 2

Two-stage musculotendon parameter optimisation

Scaling of muscle moment arms, l mt, segment mass, and anthropometry

Scaling of l0

m and

lst

Range of measured shoulder positions

Scaling of l0mand lst

Neutral shoulder position

MVIMCs and MVIKCs experiments

Subject mass and anthropometry measurements

Scaled-generic -2 musculotendon parameters

Scaled-generic -1 musculotendon parameters

Generic musculotendon parameters

Musculotendon parameters from one subject

Subject-specific musculotendon parameters

Glenohumeral joint forces

Muscle forces

Musculoskeletal model calculations:

Subject motion tasks: Abduction Flexion Horizontal Flexion Nose touch Reaching

Figure 3

Abduction Muscle force (BW)

Anterior deltoid

Flexion Muscle force (BW)

Subscapularis

Middle PecMajor

0.4

0.4

0.4

0.4

0.4

0.35

0.35

0.35

0.35

0.35

0.3

0.3

0.3

0.3

0.3

0.3

0.25

0.25

0.25

0.25

0.25

0.25

0.2

0.2

0.2

0.2

0.2

0.2

0.15

0.15

0.15

0.15

0.15

0.15

0.1

0.1

0.1

0.1

0.1

0.1

0.05

0.05

0.05

0.05

0.05

0.05

0

20 40 60 80 Abduction angle (deg)

0

20 40 60 80 Abduction angle (deg)

0

20 40 60 80 Abduction angle (deg)

0

20 40 60 80 Abduction angle (deg)

0

20 40 60 80 Abduction angle (deg)

0.4

0.4

0.4

0.4

0.4

0.4

0.35

0.35

0.35

0.35

0.35

0.35

0.3

0.3

0.3

0.3

0.3

0.3

0.25

0.25

0.25

0.25

0.25

0.25

0.2

0.2

0.2

0.2

0.2

0.2

0.15

0.15

0.15

0.15

0.15

0.15

0.1

0.1

0.1

0.1

0.1

0.1

0.05

0.05

0.05

0.05

0.05

0.05

0

20 40 60 80 Flexion angle (deg)

0

20 40 60 80 Flexion angle (deg)

0

20 40 60 80 Flexion angle (deg)

0

20 40 60 80 Flexion angle (deg)

0.4

0.4

0.4

0.4

0.4

0.35

0.35

0.35

0.35

0.35

0.35

0.3

0.3

0.3

0.3

0.3

0.3

0.25

0.25

0.25

0.25

0.25

0.25

0.2

0.2

0.2

0.2

0.2

0.2

0.15

0.15

0.15

0.15

0.15

0.15

0.1

0.1

0.1

0.1

0.1

0.1

0.05

0.05

0.05

0.05

0.05

0.05

0 30 45 60 75 90 Horizontal flexion angle (deg)

0 30 45 60 75 90 Horizontal flexion angle (deg)

0 30 45 60 75 90 Horizontal flexion angle (deg)

0 30 45 60 75 90 Horizontal flexion angle (deg)

0.4

0.4

0.4

0.4

0.4

0.35

0.35

0.35

0.35

0.35

0.35

0.3

0.3

0.3

0.3

0.3

0.3

0.25

0.25

0.25

0.25

0.25

0.25

0.2

0.2

0.2

0.2

0.2

0.2

0.15

0.15

0.15

0.15

0.15

0.15

0.1

0.1

0.1

0.1

0.1

0.1

0.05

0.05

0.05

0.05

0.05

0.05

0

0

0

0

0

1 Time (s)

2

0

1 Time (s)

2

0

1 Time (s)

2

0

1 Time (s)

2

0

1 Time (s)

2

0

0.4

0.4

0.4

0.4

0.4

0.4

0.35

0.35

0.35

0.35

0.35

0.35

0.3

0.3

0.3

0.3

0.3

0.3

0.25

0.25

0.25

0.25

0.25

0.25

0.2

0.2

0.2

0.2

0.2

0.2

0.15

0.15

0.15

0.15

0.15

0.15

0.1

0.1

0.1

0.1

0.1

0.1

0.05

0.05

0.05

0.05

0.05

0.05

0

0

1 Time (s)

2

0

0

1 Time (s)

2

0

0

1 Time (s)

2

0

0

1 Time (s)

2

0

0

1 Time (s)

2

20 40 60 80 Flexion angle (deg)

0 30 45 60 75 90 Horizontal flexion angle (deg)

0.4

0

20 40 60 80 Abduction angle (deg)

0

20 40 60 80 Flexion angle (deg)

0.4

0 30 45 60 75 90 Horizontal flexion angle (deg)

Nose touch Muscle force (BW)

Infraspinatus

0.4

0

Reach Muscle force (BW)

Supraspinatus

0.35

0

Horizontal flexion Muscle force (BW)

Middle deltoid

0

0

1 Time (s)

2

0

1 Time (s)

2

0.3 0.2 0.1 0

0.3

0.2

0.1

0

0.5 0.4 0.3 0.2 0.1 0

0.5

0.4

0.3

0.2

0.1

0

Flexion angle (deg)

0.6

0.6

80

0.7

0.7

60

0.8

0.8

40

0.9

0.9

20

0.4

0.4

Abduction angle (deg)

0.5

0.5

80

0.6

0.6

60

0.7

0.7

40

0.8

0.8

20

0.9

Resultant force

0.9

Figure 4

Abduction Muscle force (BW)

Flexion Muscle force (BW)

20

20

60

80

60

Flexion angle (deg)

40

80

Abduction angle (deg)

40

Compressive force

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

20

20

60

80

60

Flexion angle (deg)

40

80

Abduction angle (deg)

40

Anterior force

!"5

!"4

!"3

!"2

!"1

0

0.1

0.2

0.3

0.4

!"5

!"4

!"3

!"2

!"1

0

0.1

0.2

0.3

0.4

20

20

60

80

60

Flexion angle (deg)

40

80

Abduction angle (deg)

40

Superior force

Abduction

Muscle fiber force (F )

0.4

SG-2

SS

SG-2

SG-1

G

SS

0.2

0.4

0.6

0.8

1

0

o

0.4

0.6 0.8 1 1.2 m Muscle fiber length (lo )

Anterior deltoid

Muscle fiber force (Fm)

0

0.2

0.4

0.6

0.8

1

SG-1

G

m o

Flexion

Figure 5

0.6 0.8 1 1.2 m Muscle fiber length (lo )

Middle deltoid

0.4

0

0.2

0.4

0.6

0.8

1

0.6 0.8 1 1.2 m Muscle fiber length (lo )

Supraspinatus

0.4

0

0.2

0.4

0.6

0.8

1

0.6 0.8 1 1.2 m Muscle fiber length (lo )

Infraspinatus

0.4

0

0.2

0.4

0.6

0.8

1

0.6 0.8 1 1.2 m Muscle fiber length (lo )

Subscapularis

Muscle fiber force (Fm)

o

Muscle fiber force (Fm) o

Muscle fiber force (Fm) o

Muscle fiber force (Fm) o

0.4

0

0.2

0.4

0.6

0.8

1

0.6 0.8 1 1.2 m Muscle fiber length (lo )

Middle PecMajor

Table 1. Descriptions of the maximum voluntary isometric contraction tasks, maximum voluntary isokinetic contraction tasks, and upper-limb activities of daily living, including abduction, flexion, horizontal-flexion, nose touching and reaching. The neutral position was defined with the upper-limb at 0º of abduction, the elbow fully extended and the palm facing inward. The x, y, and z axis are the global upper limb coordinates defined by the ISB (Wu et al., 2005), where the x, y and z axes are aligned with the anterior, superior and lateral directions (see Supplementary Material for details). ‘Longaxis’ refers to the direction aligned with the long axis of the humerus.

Activity of daily living

Start

Abducti on

Neutral position

Flexion

Neutral position

End Humerus abducted to 120° in the coronal plane; elbow fully extended; palm facing down Humerus flexed to 120° in the sagittal plane; elbow fully extended; palm facing inward

Task completing time (s)

Rotatio nal axis

4

-x

4

+z

Humerus Humerus flexed to abducted to 90° in the 90° in the sagittal Horizon coronal plane; tal plane; elbow flexion elbow fully fully extended;pa extended; lm facing palm down facing down Thumb touching the nose; elbow naturally Nose Neutral placed; touching position the humerus proximat ely elevated 30° Hand reaching out in the midsagitt al plane, same Reachin Neutral height as g position nose, palm facing down; elbow fully extended

4

+y

4

-------

4

-------

Isokinet Start / End ic task Abducti on

Adducti on

Flexion

Extensio n

External rotation Internal rotation

External rotation Internal rotation

End / Start

Humerus abducted to 120° in the Neutral coronal position plane; elbow fully extended Humerus flexed to 120° in the Neutral sagittal position plane; elbow fully extended Humerus Humerus fully fully externall internally y rotated, rotated, abducted abducted to to 30° in 30° in the the coronal coronal plane; plane; elbow 90° elbow flexed 90° flexed Humerus Humerus fully fully internally externall rotated, y rotated,

Fast velocity (°/s)

Slow velocity (°/s) -x

30

90 +x

+z

30

90

30

90

± Longaxis

30

90

± Longaxis

-z

abducted to abducted 90° in the to 90° in coronal the plane; coronal elbow 90° plane; flexed elbow 90° flexed Isometr ic task 1

2

3

4

5

6

Position

Contracti Contracti on 1 on 2

Humerus abducted to 30° in the coronal Abduction Adduction ±z plane; elbow fully extended; Humerus abducted to 90° in the coronal Abduction Adduction ±z plane; elbow fully extended; Humerus flexed to 30° in the sagittal plane; Flexion Extension ±z elbow fully extended; Humerus flexed to 90° in the sagittal plane; Flexion Extension ±z elbow fully extended; Humerus abducted to 30° in the coronal Internal External ± Longplane; elbow 90° rotation rotation axis flexed Humerus abducted to Internal External ± Long90° in the coronal rotation rotation axis plane; elbow 90° flexed

Table 2.

Results of two-way repeated-values analysis of variance (ANOVA) calculations. Dependent variables used in the ANOVA calculations were muscle and resultant glenohumeral (GH) joint forces, while the independent variables were joint position and model type [generic (G) vs subject-specific (SS), scaled-generic-1 (SG-1) vs SS, and scaled-generic-2 (SG-2) vs SS]. Significant differences in muscle and glenohumeral joint forces as a result of the model type, joint position, and interaction effects, are denoted by ‘T’, ’P’ and ’I’, respectively. Mean differences in muscle and glenohumeral joint force (%BW) between model types are given in addition as well as their p-values and standard deviations (parentheses). Data are provided for abduction, flexion, horizontal-flexion, nose touching, and reaching. Muscle force

Resultant GH joint force

Abduc tion

G vs SS SG-1 vs SS SG-2 vs SS

Middle deltoid -1.39 (1.32) p < 0.001, T P 2.71 (1.71) p < 0.001, T P 1.36 (1.43) p < 0.001, T P -

Supraspinatus -0.83 (1.22) p < 0.001, T P -1.38 (1.12) p < 0.001, T P -1.21 (1.06) p < 0.001, T P -

Infraspinatus -2.02 (4.91) p < 0.001, T P 5.01 (3.04) p < 0.001, T P I 3.60 (2.74) p < 0.001, T P -

Subscapularis -2.27 (4.18) p < 0.001, T P 1.09 (1.06) p < 0.001, T P 2.47 (1.06) p < 0.001, T P -

-12.63 (7.08) p < 0.001, T P 1.52 (3.61) p = 0.022, T P 1.38 (2.81) p = 0.040, T P -

Flexio n

G vs SS SG-1 vs SS SG-2 vs SS

-0.51 (2.13) p = 0.060, - P 0.88 (0.69) p = 0.004, T P 0.08 (0.76) p = 0.789, - P -

-1.27 (2.66) p < 0.001, T P 0.97 (4.01) p = 0.023, T P 0.38 (2.38) p = 0.268, - P -

-0.66 (4.42) p = 0.080, - P 2.42 (4.35) p < 0.001, T P 2.19 (4.05) p < 0.001, T P -

-2.92 (2.34) p < 0.001, T P 0.97 (2.00) p < 0.001, T P 0.82 (1.78) p = 0.005, T P -

-13.77 (4.17) p < 0.001, T P I 5.90 (12.34) p < 0.001, T P 0.80 (5.64) p = 0.244, - P -

Horiz ontal flexion

G vs SS SG-1 vs SS SG-2 vs SS

-1.03 (2.66) p = 0.004, T P 1.15 (6.04) p = 0.035, T P -0.10 (6.06) p = 0.841, - P -

-2.31 (2.47) p < 0.001, T P 2.21 (3.71) p < 0.001, T P I 1.88 (3.15) p < 0.001, T P -

0.31 (3.35) p = 0.637, - - 1.36 (4.30) p = 0.054, - P 2.39 (3.11) p = 0.001, T - -

-4.15 (2.85) p < 0.001, T P 0.86 (2.80) p = 0.014, T P 1.28 (2.52) p < 0.001, T P -

-16.34 (12.16) p<0.001, T P 10.42 (7.25) p < 0.001, T P 6.20 (4.66) p < 0.001, T P -

Nose touchi ng

G vs SS SG-1 vs SS SG-2 vs SS

-0.93 (1.32) p < 0.001, T P 0.56 (1.31) p < 0.001, T - 0.56 (1.28) p < 0.001, T - -

-0.14 (0.94) p = 0.271, - P -0.93 (0.89) p < 0.001, T P -1.02 (0.82) p < 0.001, T P -

-0.16 (2.64) p = 0.328, - P 1.14 (1.98) p < 0.001, T P 1.35 (1.92) p < 0.001, T P -

-1.75 (1.33) p < 0.001, T P 0.57 (1.06) p < 0.001, T P 0.42 (1.07) p < 0.001, T P -

-6.25 (1.87) p < 0.001, T P 0.15 (1.30) p = 0.681, - P -0.23 (1.43) p = 0.526, - P -

Reach ing

G vs SS SG-1 vs SS SG-2 vs SS

-0.85 (2.39) p < 0.001, T P 1.19 (2.30) p < 0.001, T P 0.54 (2.29) p = 0.033, T P -

-1.25 (1.27) p < 0.001, T P -1.43 (1.43) p < 0.001, T - -1.31 (1.47) p < 0.001, T - -

-0.08 (4.80) p = 0.824, - P 2.96 (4.17) p < 0.001, T P 2.63 (4.09) p < 0.001, T P -

-1.03 (1.40) p < 0.001, T P 0.70 (1.36) p = 0.005, T P 0.81 (1.48) p = 0.001, T P -

-10.91 (4.90) p < 0.001, T P -0.85 (2.66) p = 0.323, - P -2.40 (2.22) p = 0.003, T P -

Table 3.

Muscle

Anterior deltoid Middle deltoid Posterior deltoid Supraspin atus Infraspinat us Subscapul aris Teres minor Teres major Superior pec. major Middle pec. major Inferior pec. major Latissimus dorsi Coracobra chialis

Averaged musculotendon parameter data for shoulder muscles including optimal muscle forces, optimal fiber lengths, tendon slack lengths, physiological cross-sectional areas (PCSA) and muscle volumes. Mean values and standard deviations across the 6 subjects are provided. Optimal muscle force (N)

Optimal fiber length (cm)

Tendon slack length (cm)

PCSA (cm2)

Muscle volume (cm3)

S Me T an D

Mean STD

Mean STD

Mean STD

Mean STD

556.76 109.72

17.52 2.64

3.13 2.37

1098.4 214.23

13.18 1.58

4.67 1.11

944.68 196.79

12.28 1.39

9.75 1.35

410.73 83.02

11.65 2.32

2.48 1.91

864.58 213.82

13.23 2.46

3.37 1.23

10

944.28 194.36

16.32 0.87

0.75 0.11

10. 92

605.41 185.79

4.53 1.14

10.38 1.49

7

234.9 80.92

13.88 3.87

5.37 2.76

983.44 277.02

10.25 3.41

4.8 2.96

14 0.88

9.55 1.41

446.73 112.97

18.13 1.99

9.88 2.81

1129.7 297.11

23.17 7.21

7.7 5.31

306.93 105.68

8.32 2.03

6.15 2.36

699.7 143.3

6.4 4 12. 7 10. 92 4.7 5

2.7 2 11. 37 8.0 9 5.1 6 13. 06 3.5 5

1. 39 2. 71 2. 49 1. 05 2. 71 2. 46 2. 35 1. 02 3. 51 1. 81 1. 43 3. 76 1. 34

111.8 28.22 8 167.6 42.16 2 134.3 33.93 8 54.39 13.69 129.2 32.52 5 179.0 45.35 4 30.28 7.64 35.61 8.97 109.8 27.61 113.6 28.66 8 92.41 23.37 285.7 71.81 27.43 6.95