Abstract TP155: Robotic Retraining of Finger Movements After Stroke

Abstract

Background: Robots aid motor rehabilitation, but there has been limited attention to recovery of finger movements. This study evaluated robotic assistance during finger movement training. Functional MRI (fMRI) was acquired at baseline to understand predictors of treatment gains. Methods: Patients with chronic stroke underwent a baseline fMRI scan, alternating rest with affected-side finger movements similar to those made during robotic therapy. Next, subjects received therapy 3 hr/wk for 3 weeks using FINGER (Finger Individuating Grasp Exercise Robot), with which subjects moved their paretic index and middle fingers to play a musical game similar to GuitarHero. FINGER used an assistance-as-needed algorithm to facilitate completion of grasping movements, which increased sensory feedback without altering voluntary motor output. Participants were randomized to receive High Assistance (to insure 85% success) or Low Assistance (55% success). Results: 30 subjects (mean age 58 yr; baseline Fugl-Meyer 46 out of 66; 37 mo post-stroke) completed the study. Significant gains were found in the primary outcome measure, change in Box & Blocks (B&B) score (23 to 25.5, p<0.0001). There was no difference between High and Low Assistance groups in the primary endpoint (p=0.65), though some secondary outcomes favored High Assistance. The fMRI scans found that greater treatment gains were associated with higher laterality index in primary sensory cortex, indicating greater boost in B&B score over time with higher pretreatment balance of activation towards ipsilesional rather than contralesional sensory cortex; laterality index in primary motor cortex lacked predictive value. Conclusions: Significant motor gains were found with a robotic device that targets finger movements. Sensory factors appear key: treatment content emphasized augmented sensory feedback, and the hemispheric balance of fMRI activation within sensory but not motor cortex predicted treatment gains. Together these findings suggest Hebbian rules of sensorimotor cortex plasticity during finger robotic therapy after stroke.