The ability to predict optimal treatments computationally for individual patients would be a significant step forward in the field of medicine. For most conditions or diseases, treatment plans are prescribed by doctors based on their experience, intuition, and generally accepted standards. Many conditions, however, do not have a single best treatment that works for everyone. In these cases, attempting to find the most effective treatment is often a trial and error process. In order to change this paradigm, a method of tailoring treatments to specific individuals is needed.

Gait rehabilitation for individuals who have had a stroke is one case where subject-specific treatments are needed. Fast functional electrical stimulation (fastFES) has shown promise as a post-stroke gait rehabilitation technique, but there are currently many unknowns in the process. This study lays out some underlying principles and provides a framework for determining optimal fastFES parameters at a subject-specific level.

First, basic principles of optimal control of human gait are explored, especially as they relate to prediction of foot-ground contact forces. Prediction of these forces is needed to predict new gait patterns that would arise from proposed interventions. A simplified model of the human lower body is applied to a trajectory tracking problem to determine which aspects of the problem formulation hinder or encourage computational speed and robustness. Specifically, ground contact constraints, generalized coordinate choice, and explicit versus implicit dynamics are tested. The exploration revealed that ground contact constraints may not be beneficial, whereas ground to foot generalized coordinates and implicit dynamics may be.

Second, a full body, three dimensional, subject-specific model of an individual post-stroke is created to explore ways to improve and customize the design of fastFES training schemes. Skeletal, muscular, ground contact force, and neural control parameters are calibrated to movement data collected from a specific individual poststroke who had undergone standard fastFES training. This model is then used in an optimal control problem to predict optimal treatment parameters for fastFES training. The optimizations predicted that extending gastrocnemius medialis stimulation duration or stimulating semimembranosus and soleus instead may produce more symmetric propulsive force compared to current methods.