Rehabilitation monitoring and assessment with sensors and artificial intelligence (AI) provide an opportunity to improve current rehabilitation practices. However, the adoption of such a technology in practice still remains a challenge. This work presents an interactive approach that supports collaboration between a therapist and an AI based system for rehabilitation assessment. An AI-based system can automatically identify salient features of assessment to generate patient-specific analysis for a therapist. After reviewing patient-specific analysis, a therapist can provide feedback to tune an imperfect system.
In two evaluations with therapists, we found that the patient-specific analysis supports significantly higher agreement of therapists' assessment than a traditional system without any analysis. In addition, therapists can provide feature-based feedback to tune a system, which significantly improves its performance from 0.8377 to 0.9116 average F1-scores on three exercises. This work discusses the potential of a human and AI collaborative system that supports more accurate decision making while learning from each other's strength.
Towards Efficient Annotations for a Human-AI Collaborative, Clinical Decision Support System
Interactive Hybrid Intelligence Systems for Human-AI/Robot Collaboration