Tele-rehabilitation and machine learning-driven personalized gait retraining for post-stroke lower-limb motor impairment

Project Abstract

This study presents a novel tele-rehabilitation framework that integrates machine learning-driven personalized gait retraining to address post-stroke lower-limb motor impairment. The research aims to design, implement, and evaluate an end-to-end, remotely deliverable rehabilitation system that adapts to individual gait deficits, real-time performance, and progress over time. A multi-modal data collection strategy, combining wearable inertial measurement units, plantar pressure sensors, and video-based motion analysis, generates a comprehensive feature set for downstream personalized intervention. We develop a patient-specific normative gait model using supervised learning on a large dataset of healthy and post-stroke gait patterns, enabling the system to quantify deviations, identify dominant impairment mechanisms (e.g., spatiotemporal asymmetry, knee flexion impairment, ankle dorsiflexion weakness), and predict optimal retraining strategies. The tele-rehabilitation platform comprises a clinician dashboard for remote assessment, a patient mobile/desktop interface for guided sessions, and a cloud-based analytics engine that performs real-time anomaly detection, progress tracking, and adaptive exercise prescription. The core contribution lies in an intelligent, closed-loop gait retraining pipeline. First, baseline assessment establishes individual impairment signatures and home-based task difficulty. Second, the system prescribes a personalized program of gait tasks, assistive device adjustments, and feedback modalities (auditory, visual, and haptic). Third, real-time kinematic and kinetic feedback is fused with ML-driven coaching cues to correct gait deviations during treadmill or over-ground walking, whether in clinical facilities or at home via telepresence. Fourth, the framework continuously updates the personalization model as new data accrue, allowing gradual progression from fundamental motor tasks to functional mobility scenarios such as stair negotiation and obstacle clearance. Fifth, remote monitoring enables clinicians to adjust target outcomes and safely escalate training intensity. A mixed-methods evaluation design is employed. Quantitative outcomes include changes in gait symmetry indices, swing-stance duration ratios, knee and ankle kinematics, functional mobility scores (e.g., Timed Up and Go, 6-Minute Walk Test), and user engagement metrics. The ML components are validated against ground-truth motion capture benchmarks, with emphasis on model generalizability across varying stroke severities, ages, and comorbidities. Qualitative insights are drawn from semi-structured interviews with participants and clinicians to assess usability, perceived autonomy, motivation, and perceived barriers to tele-rehabilitation adoption. A pilot randomized controlled trial compares the tele-rehabilitation system against standard therapy, focusing on feasibility, safety, and preliminary efficacy over a 12-week period. Ethical considerations address data privacy, informed consent, and safeguarding remote supervision. The study anticipates that personalized, ML-driven gait retraining delivered through tele-rehabilitation will enhance motor recovery, increase adherence through flexible remote access, reduce travel burdens for patients with mobility limitations, and enable scalable rehabilitation delivery. Implications extend to post-stroke care models, rehabilitation equity, and the integration of intelligent assistive feedback within home-based neurorehabilitation ecosystems.

Project Overview

What This Project Is About

---

A straightforward look at how tele-rehabilitation and simple, computer-assisted coaching can help people relearn walking after a stroke. The project joins remote therapy sessions with a lightweight computer model that suggests personalized gait exercises based on a person’s progress.

---

---

The Problem It Addresses

---

---

---

Objectives of the Project

---

1. Identify how remote coaching plus simple data feedback can influence gait recovery.
2. Develop a plan for personalized exercise guidance that users can follow at home.
3. Explore basic machine learning ideas that adjust exercises based on user progress.
4. Test usability and safety of the tele-rehabilitation approach with a small group.

---

---

What You Will Do Step by Step

---

Step 1: Review current gait rehab methods and tele-rehab tools. Step 2: Design a user-friendly remote exercise program. Step 3: Collect data from participants during home sessions (e.g., steps, gait roughness). Step 4: Apply simple data analysis to track improvements. Step 5: Create basic guidance updates that adapt over time. Step 6: Assess participant feedback and safety.

---

---

---

---

Expected Outcome

An approachable, home-based gait retraining framework that uses remote coaching and simple adaptive suggestions. Anticipated benefits include improved walking speed, balance, and confidence, with greater accessibility to rehab services.