Biomechanical Analysis and AI-Guided Robotic-Assisted Therapy for Post-Stroke Upper Limb Rehabilitation

Project Abstract

This study presents a comprehensive investigation into the integration of biomechanical analysis with artificial intelligence (AI)-guided robotic-assisted therapy to enhance upper limb rehabilitation in post-stroke patients. By combining quantitative biomechanical metrics with adaptive AI-driven control strategies, the research aims to optimize motor recovery, reduce compensatory movements, and accelerate functional restoration. The abstract outlines a multi-phase approach first, a biomechanical assessment to characterize kinematics, kinetics, muscle activation patterns, and joint load distributions during standardized reaching and grasp tasks in a cohort of chronic stroke survivors and matched healthy controls. Surface electromyography (sEMG), motion capture, and force sensors form the core measurement suite, enabling precise profiling of impairment severity, movement speed-accuracy trade-offs, and synergy patterns across the shoulder, elbow, and wrist. In the second phase, rehabilitation episodes are delivered via a lightweight, portable robotic exoskeleton with haptic feedback and assist-as-needed control populated by an embedded AI module. The AI system leverages a hybrid learning framework combining supervised learning for initial personalization with reinforcement learning for real-time adaptation, thereby tuning assistance levels, trajectory guidance, and resistance to maximize motor learning while ensuring patient safety. A key objective is to minimize abnormal synergies and reliance on proximal joints by promoting task-specific, functionally relevant training that mirrors daily activities. Third, the study introduces a closed-loop optimization strategy that continuously updates patient models using online data streams to adjust therapy parameters, such as assist-as-needed thresholds, impedance settings, and timing of corrective cues. The performance metrics comprise clinically validated scales (e.g., Fugl-Meyer Assessment for the Upper Extremity, Box and Block Test), kinematic indices (trajectory smoothness, end-point error, movement variability), kinetic measures (joint torques, interaction forces), and neurophysiological indicators (co-contraction ratios, muscle activation onset). A randomized controlled trial design compares the AI-guided robotic therapy to conventional robotic assistance and conventional therapy, with follow-ups at one, three, and six months to evaluate retention and transfer to activities of daily living. Statistical analyses include mixed-effects models to account for inter-subject variability and repeated measures, along with effect size estimation and clinical relevance thresholds. Anticipated outcomes include improved movement quality, faster task completion, greater independence in daily activities, and enhanced neuroplastic changes indicated by sustained improvements in motor synergies and cognitive-motor integration. The study also explores design implications for home-based rehabilitation, emphasizing portability, user-friendly interfaces, adaptive safety features, and data privacy. Potential limitations encompass heterogeneous lesion characteristics, variability in patient engagement, and the need for long-term follow-up to assess durability of gains. By bridging biomechanical insight with AI-enabled personalization, the research seeks to establish an end-to-end rehabilitation paradigm that adapts to individual recovery trajectories, thereby facilitating durable restoration of upper limb function and quality of life for post-stroke individuals.

Project Overview

What This Project Is About

---

This project looks at how to understand and improve arm movement after a stroke using two ideas: measuring how the arm moves (biomechanics) and using smart robots with artificial intelligence to guide therapy. It aims to find ways to make rehabilitation more effective and engaging for patients.

---

---

The Problem It Addresses

---

Many people recovering from a stroke have limited arm movement, which slows daily activities and independence. Traditional therapy can be repetitive and hard to personalize. The project explores how precise measurements and AI-powered robots can tailor therapy to individual needs, potentially speeding recovery and reducing caregiver load.

---

---

Objectives of the Project

---

1. Explain how to measure arm movement in a user-friendly way.
2. Investigate simple AI ideas that guide robotic therapy sessions.
3. Develop a basic protocol for a robot-assisted therapy session.
4. Test whether the approach improves movement compared to standard exercises.
5. Identify factors that affect success, such as motivation and feedback.

---

---

What You Will Do Step by Step

---

1. Review simple literature on biomechanics and robot-assisted therapy.
2. Select affordable sensors to track arm motion (e.g., position and speed).
3. Prototype a basic robotic assist device or simulator with AI guidance.
4. Design short therapy sessions and collect movement data.
5. Analyze data to see changes in range of motion and smoothness.
6. Assess user experience and safety considerations with patients or volunteers.
7. Refine the therapy protocol based on results and feedback.
8. Prepare a concise report and presentation of findings.

---

---

Expected Outcome

A practical, easy-to-use approach for robot-assisted therapy that adapts to individual patients, with preliminary evidence of improved arm movement and enjoyment in therapy sessions. The work could inform future, larger studies and clinical tools.