ReactEMG Stroke: Healthy-to-Stroke Few-shot Adaptation for sEMG-Based Intent Detection

2026-01-29

AI Summary

This study addresses the practical challenges in surface electromyography (sEMG)-based intent recognition for stroke patients, which are hindered by high inter-subject variability, costly calibration procedures, and distribution shifts in sEMG signals. To overcome these limitations, this work proposes a novel few-shot transfer learning framework that leverages a model pre-trained on large-scale sEMG data from healthy individuals and adapts it to stroke patients with minimal labeled data. The approach explores three adaptation strategies: fine-tuning only the classification head, employing parameter-efficient LoRA adapters, and full end-to-end fine-tuning. Experimental results demonstrate substantial improvements over baseline methods, increasing intent transition accuracy from 0.42 to 0.61 and overall accuracy from 0.69 to 0.78, thereby outperforming both zero-shot transfer and models trained exclusively on patient data.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Surface electromyography (sEMG) is a promising control signal for assist-as-needed hand rehabilitation after stroke, but detecting intent from paretic muscles often requires lengthy, subject-specific calibration and remains brittle to variability. We propose a healthy-to-stroke adaptation pipeline that initializes an intent detector from a model pretrained on large-scale able-bodied sEMG, then fine-tunes it for each stroke participant using only a small amount of subject-specific data. Using a newly collected dataset from three individuals with chronic stroke, we compare adaptation strategies (head-only tuning, parameter-efficient LoRA adapters, and full end-to-end fine-tuning) and evaluate on held-out test sets that include realistic distribution shifts such as within-session drift, posture changes, and armband repositioning. Across conditions, healthy-pretrained adaptation consistently improves stroke intent detection relative to both zero-shot transfer and stroke-only training under the same data budget; the best adaptation methods improve average transition accuracy from 0.42 to 0.61 and raw accuracy from 0.69 to 0.78. These results suggest that transferring a reusable healthy-domain EMG representation can reduce calibration burden while improving robustness for real-time post-stroke intent detection.

Problem

• sEMG
• stroke rehabilitation
• intent detection
• few-shot adaptation
• calibration burden

Innovation

• few-shot adaptation
• sEMG-based intent detection
• healthy-to-stroke transfer
• parameter-efficient fine-tuning
• distribution shift robustness