Subject-specific computer simulation studies have been used to describe the function of individual muscles in healthy and post-stroke gait, and the results of musculoskeletal simulations can be useful for understanding an individual’s response to rehabilitation. Subject-specific information about muscle force and volitional activation can help to improve the predictive power of computer generated musculoskeletal models, but require accurate and reliable measurement techniques to obtain. The goals of this dissertation are to enhance our ability to create subject-specific models of individuals post-stroke and improve our understanding of muscle coordination in individuals post-stroke.

As part of this dissertation, we identified available compensatory strategies for muscle weakness during gait by simulating activation deficits in multiple muscle groups. Subject-specific simulations were created to determine the changes in modeled activation and contribution to knee joint and center of mass acceleration of the ankle plantar flexor muscles in patients post-stroke after a targeted function electrical stimulation intervention, as well as to identify relationships between simulation results and clinical gait variables. Next, an adjustment equation was developed for the burst superimposition test to estimate the maximum force generating ability (MFGA) of the plantar flexor muscles in a post-stroke population. This new adjustment equation was then used to assess the ability of muscle volume obtained through magnetic resonance imaging (MRI) to estimate the MFGA of the plantar flexor muscle group for individuals post-stroke. Lastly, we created musculoskeletal simulations of individuals post-stroke with subject-specific muscle force and activation data.

We found that musculoskeletal models were unable to recreate normal gait patterns with simultaneous impairment of the plantar flexor, dorsiflexor, and hamstrings muscle groups. Other muscle groups were unable to assist the dorsiflexor muscles during early swing, which suggests that rehabilitation or assistive devices may be required to correct foot drop. Simulations of individuals post-stroke created both pre- and post-intervention showed a new pattern of model-predicted activation for the plantar flexor muscles after training, suggesting that the subjects activated these muscles with more appropriate timing following the intervention. Functionally, the simulations predicted that the plantar flexors provided greater contribution to knee flexion acceleration after training, which is important for increasing swing phase knee flexion and foot clearance. While muscle volume obtained via MRI provided information on the overall size of muscle, it was found to overestimate the force generating ability of the paretic limb plantar flexors and, therefore, does not provide an accurate, detailed description of the overall force producing capability of muscle. Finally, the inclusion of subject-specific muscle data resulted in greater modelpredicted force and activation levels for the hip and knee flexors, which agree with previously reported compensation patterns. The timing of muscle activation predicted by the model agreed more closely with the timing of EMG for the plantar flexor and hamstring muscles when subject-specific parameters were used.

By identifying how muscles can interact, clinicians may be able to develop specific strategies for using gait retraining and orthotic assistance to best address an individual’s needs. The central activation ratio predicted using our adjusted burst superimposition technique has a strong relationship with the maximum volitional force of contraction, and may be a more accurate and useful measure than MRIderived muscle volume when evaluating muscle weakness in persons post-stroke. Lastly, the results of this dissertation suggest that subject-specific isometric force and activation data may affect the accuracy of model predictions and should be used when building musculoskeletal models of individuals post-stroke.