UC Irvine

Stroke is a leading cause of adult disability and patient response to treatment is highly variable. To understand this heterogeneity, the anatomical integrity and functional activity of the post-stroke motor system has been well investigated. In contrast, remarkably limited attention has been paid to somatosensory system counterparts in terms of predicting motor outcomes. Proprioception is known to be an integral aspect of motor control, and many rehabilitation strategies are built upon a somatosensory-induced Hebbian plasticity framework. Unfortunately, clinical assessments of proprioception often fail to yield meaningful behavioral data and neural correlates to post-stroke proprioception function are poorly understood. Behavioral and neuroimaging assessments of the somatosensory system, specifically proprioception, may yield valuable insight to the heterogeneity in therapy-induced motor gains. Therefore, the current dissertation aimed to 1) develop an objective and sensitive proprioception assessment; 2) characterize the neural correlates of post-stroke proprioception dysfunction; and 3) identify predictors of motor gains from a 3-week course of robotic finger therapy, taking into equal consideration somatosensory- and motor-derived variables. A proprioception assessment designed with the Finger Individuating Grasp Exercise Robot (FINGER) was capable of detecting age-related and stroke-induced decline in finger proprioception and proved to be more sensitive than standard scales. Among a population of 30 subjects with chronic stroke, finger proprioception deficits were present contralesionally in 67% and bilaterally in 56%. Post-stroke proprioception status was best explained by anatomical injury to somatosensory networks and changes in cortical connectivity between ipsilesional primary motor cortex (iM1) and secondary somatosensory cortex (iS2). After a course of robotic therapy, subjects showed variable improvements in arm motor function. Behaviorally, baseline proprioception status best predicted treatment gains, outperforming baseline measure of motor behavior. Neurologically, a combined model of somatosensory network injury and iM1-iS2 cortical connectivity explained 56% of variance in treatment gains. The comprehensive approach described here demonstrates that proprioception is an integral aspect of post-stroke motor recovery. Importantly, these results are the first to directly support the concept of somatosensory-induced Hebbian-like learning within the context of robot-assisted motor rehabilitation for chronic stroke. The findings illustrate the importance of incorporating proprioception into rehabilitation strategies and clinical decision making.