Computational Neurorehabilitation

Abstract

Computational Neurorehabilitation is an emerging field at the intersection of Neurorehabilitation, Computational Neuroscience, Motor Control and Learning, and Statistical Learning. The overarching goals of Computational Neurorehabilitation are to understand and to further improve motor recovery following neurologic injury by mathematically modeling and simulating the neural processes underlying the change in behavior due to rehabilitation (1). This chapter is organized into three main sections. First, we review the overall framework of Computational Neurorehabilitation and argue that computational neurorehabilitation models belong to the general class of dynamical system models. Second, we discuss the three categories of plastic processes that have been incorporated in previous models: unsupervised, supervised, and reinforcement learning. Third, we discuss the two main types of models in Computational Neurorehabilitation: Qualitative “biological” models whose main goal is to advance our understanding of the neural mechanisms of recovery and Quantitative “predictive” models whose main goal is to predict long-term changes in functional outcomes for individual patients. We illustrate these two types of models by briefly reviewing a number of recent relevant qualitative and quantitative models. We conclude by suggesting future directions for the field.