How to retrain movement after neurologic injury: a computational rationale for incorporating robot (or therapist) assistance

This paper develops an adaptive Markov model of sensory motor control, and then uses the model to examine the putative role of external mechanical assistance from a robotic device or therapist in promoting neurologic recovery. The model assumes that: 1) the CNS probabilistically interprets proprioceptive information in real time in order to generate motor output; 2) sensory-motor pathways become more reliable with repetitive activation in a sort of Hebbian learning; 3) normal sensory input sometimes elicits abnormal motor output following neurologic injury due to disrupted neural organization. The model predicts the best movement recovery when an external trainer intervenes to correct errant movements on an "as-needed" basis, compared to no or continual assistance. The model thus provides a computational rationale for incorporating mechanical assistance on an as-needed basis during neurorehabilitation therapy.