Multidimensional feature analysis shows stratification in robotic-motor-training gains based on the level of pre-training motor impairment in stroke.
Journal:
Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference
PMID:
40039510
Abstract
Stroke involves heterogeneity in injury and ongoing endogenous recovery, which are seldom stratified before testing post-stroke robot assisted motor training (RAMT). Pretraining variations, especially sensory-motor differences may also affect the gains achieved from the RAMT. Moreover, one assessment test may not effectively characterize the baseline sensory-motor status or the RAMT gains. Pre-therapy stratification may help personalize therapy and increase therapy gains. Towards this goal, we propose a data-driven approach to assess multiple functional scores with t-distributed stochastic neighbor embedding and affinity propagation clustering, both for pre-therapy and RAMT gains. Data included behavioral scores from 27 people with chronic stroke who underwent RAMT for finger movement. Three clusters were observed at start-of-therapy (SoT), concurrent with the overall impairment level. Four clusters were observed for the RAMT gains, indicating specific improvements. The SoT clusters showed agreement with the RAMT gain clusters, suggesting that the pre-therapy state, assessed across multiple domains, could be useful in guiding RAMT interventions to improve outcomes.