Robot Assisted Exercise: Modelling the Recovery Process to Personalise Therapy

Sedda, G.

First

;Franzosi, R.;Mazzone, A.;Sanguineti, V.;Colombo, R.

2019-01-01

---

Abstract

Neurorehabilitation may greatly benefit from computational approaches. We present a computational model of the trial-by-trial dynamics of recovery through task-specific, robot-assisted exercise. The model assumes that recovery is driven by movement performance. The model explicitly addresses the extent to which training in one direction affects performance of subsequent movements in other directions. We fitted the model to data from a rehabilitation trial based on a task-specific exercise, involving reaching movements to different directions. The model reproduces the trial-by-trial speed and smoothness time series. These findings suggest that the model can be used to interpret the evolution of performance, and to formulate testable hypotheses on the recovery mechanisms, at the individual subjects' level. Therefore, it can be used to adaptively customize the robot-aided exercise based on each patient's direction-specific impairment.