Towards AI-Driven Clinical Decision Support for Post-stroke Gait Rehabilitation
摘要
3D clinical gait analysis for neurological rehabilitation is rapidly democratizing and offers an opportunity to support clinicians in identifying gait deficits. While simple difference metrics between pathological and normal can be calculated, there is a complex reasoning process that experts perform to identify treatment targets. We here attempt to capture some of this reasoning process by training algorithms to predict expert-derived gait deficit labels. This can in the future lead to clinical decision support systems for selecting gait treatment targets in an automated and targeted manner.