One in eight adults with a disability suffer from walking impairments were amputation, osteoarthritis, rheumatoid arthritis, multiple sclerosis, spinal cord injury, stroke, and traumatic brain injury are the most common conditions responsible. Along with other movement impairment conditions such as cerebral palsy, Parkinson's disease, and orthopedic cancer, these conditions have been associated with a decreased quality of life, an increased risk of serious secondary health conditions (e.g., heart disease, diabetes), and an increase in economic burden (e.g., unemployment, health care). Therefore, improving treatment for walking impairment conditions is a high rehabilitation priority and an important public health problem.

Clinicians and researchers have explored various neurorehabilitation treatments in search of effective approaches for maximizing walking function recovery. However, personalizing the design and delivery of neurorehabilitation treatments to the needs of individual patients is a challenging data science problem. Although a vast array of disparate movement-related data are available to the clinician, these data have not resulted in highly effective neurorehabilitation treatments. A promising alternative is to base treatment design on objective computational walking models that obey laws of physics and principles of physiology. With this approach, engineering design optimization methods that have successfully transformed the design of airplanes, automobiles, and other products can be used to optimize the design of clinical interventions. For such an approach to work, the computational walking models must be personalized to the patient's unique anatomical, physiological, and neurological characteristics and must be able to predict via optimization the patient's walking function following a planned intervention. Although the necessary computational methods for both capabilities exist today in validated prototype form in Dr. B.J. Fregly's lab at Rice University, they are not packaged in a way that makes them readily accessible and easy to use, thereby preventing significant research progress in this important area.

This dissertation 1) developed a software infrastructure, 2) enhanced that infrastructure with metabolic cost modeling, and 3) applied that infrastructure to pelvic sarcoma surgery. We showed that it was possible to develop a cohesive framework to generate personalized neuromusculoskeletal walking models. This framework was further enhanced by adding metabolic cost modeling. We also found that model personalization improved metabolic cost estimates. The entire framework was then used to predict physically realistic post-surgery walking function for a simulated individual with a pelvic sarcoma. Although preserving the psoas muscle increases the surgery time, it is claimed to increase mobility post-surgery and rehabilitation. However, our walking predictions revealed that the strength of this muscle did not have a strong influence on post-surgery walking function. This thesis shows that our current infrastructure has the potential to positively influence surgical or rehabilitative decisions for a wide array of walking impairments.