Reproducible Research: Computational Design of Personalized Clinical Treatments for Walking Impairments Caused by Knee Osteoarthritis and Stroke

Background: Personalized computational neuromusculoskeletal models have great potential for optimizing the design of clinical treatments for movement impairments. While many software tools address specific parts of the model personalization and treatment optimization processes, they typically require significant programming experience to use and do not cover the full breadth of these two processes. Furthermore, published neuromusculoskeletal modeling studies typically do not provide all the methodological details needed for others to reproduce the work. Consequently, researchers seeking to develop skills in the model personalization and treatment optimization processes face a steep learning curve due to the lack of detailed training materials that demonstrate both processes for real-life clinical problems using real-life subject movement data. Methods: This article presents detailed training tutorials for the model personalization and treatment optimization processes using two real-life clinical problems and the Neuromusculoskeletal Modeling (NMSM) Pipeline. The first clinical problem involves the design of personalized gait modifications and high tibial osteotomy surgery for an individual with bilateral medial knee osteoarthritis, where the goal is to reduce the peak adduction moment in both knees to a specified target level. The second clinical problem involves the design of a synergy-based functional electrical stimulation prescription for an individual post-stroke with impaired walking function, where the goal is to equalize the propulsive and braking impulses between the two legs. Both tutorials were implemented as course projects given to novice users in a combined undergraduate/graduate mechanical engineering course. Results: Both tutorials produced personalized neuromusculoskeletal models and associated dynamically consistent tracking optimizations that closely reproduced subject-specific experimental joint angles, joint moments, ground reaction forces and moments, and (if applicable) muscle activations measured during walking. Subsequent design optimizations predicted personalized treatments that achieved target values of peak knee adduction moments or propulsive and braking impulses. Conclusions: The detailed step-by-step tutorials presented with this article are the first to walk users through the entire process of creating personalized neuromusculoskeletal models and then using them to design personalized treatments for two subjects with two different clinical problems. These tutorials can be used to introduce new users to the NMSM Pipeline and as projects in neuromusculoskeletal modeling courses.