Upper extremity post-stroke motion quality estimation with decision trees and bagging forests

Abstract

Stroke is one of the leading causes of long-term disability. Approximately two thirds of stroke survivors require long-term rehabilitation, which suggests the importance of understanding movement quality in real-world settings. To address this need, we have developed an approach that quantifies physical activity and also evaluates performance quality. Accelerometer and gyroscope sensor data are used to measure upper extremity movements and to develop a mathematical framework to relate objective sensor data to clinical performance metrics. In this article we employ two approaches to extract clinically meaningful quality measures from individuals post-stroke; we then compare the resulting predictive ability of the two approaches. Our findings indicate that Bootstrap Aggregating forest approaches may be superior to the computationally simpler decision trees for unstable data sets including those derived from individuals post-stroke.