Loading
Life sciences · Preprint
arXiv · September 9, 2026
Early or partial results. Treat as a signal, not a conclusion.
This work demonstrates proof-of-concept that deep learning can decode five-finger motor intent from impaired-arm sEMG in stroke patients with reasonable accuracy, using a compact model suitable for embedded hardware. However, the study remains in the development phase: it shows signal processing and model performance on a retrospective dataset but does not yet provide clinical outcomes, prospective validation, or hardware device efficacy.
Machine learning model development and cross-validation study on retrospective longitudinal dataset. Stroke patients enrolled in PhysioMio bilateral longitudinal dataset with measurable residual motor activity in impaired arm; specific inclusion criteria and cohort size not stated in abstract.. Intervention: Deep neural network architectures (LSTM, CNN, GNN, CNN-Micro with knowledge distillation) trained to decode five-finger motor intent from impaired-arm sEMG.. Compared with: Multiple model architectures compared on the same dataset; LSTM, GNN, and CNN-Large treated as baselines against CNN-Micro optimized for embedded hardware..
LSTM achieved highest subset accuracy at 0.545; GNN achieved highest macro F1 at 0.706 and macro AUPRC at 0.776 CNN-Large single-split architecture attained 0.593 subset accuracy and 0.714 macro F1 Four-channel CNN-Micro with cross-channel knowledge distillation reached 0.5219 ± 0.0114 subset accuracy, 0.7612 ± 0.0038 finger accuracy, and 0.6095 ± 0.0058 macro F1 across five seeds
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
This work establishes a machine learning pipeline and validates signal processing methods for decoding finger intent from sEMG in stroke survivors, which could inform future assistive device development. However, clinicians should note that no functional rehabilitation outcomes, patient-level efficacy, or hardware device performance has yet been demonstrated; further prospective clinical validation is essential before adoption.
This is an early-stage machine learning development study on a stroke rehabilitation dataset, demonstrating feasibility of finger-intent decoding from sEMG but lacking clinical validation, prospective outcomes, or hardware deployment results.
As stated by the source record.
Quoted from the source exactly as published.
This work establishes a machine learning pipeline and validates signal processing methods for decoding finger intent from sEMG in stroke survivors, which could inform future assistive device development. However, clinicians should note that no functional rehabilitation outcomes, patient-level efficacy, or hardware device performance has yet been demonstrated; further prospective clinical validation is essential before adoption.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
Finger-specific motor intent is a clinically meaningful control signal for post-stroke neurorehabilitation, where residual muscle activity may remain measurable despite weak or incomplete movement. We study five-finger multilabel intent decoding from impaired-arm high-density surface electromyography (sEMG) in PhysioMio, a bilateral longitudinal dataset collected from stroke patients. A common processing protocol aligns movement labels, applies 20--450 Hz Butterworth filtering and Symlet-4 wavelet denoising, segments overlapping 200 ms windows, and extracts twelve time- and frequency-domain descriptors per channel. Direct LSTM, CNN, and GNN baselines reveal complementary behavior: the LSTM attains the highest subset accuracy (0.545), whereas the GNN attains the highest macro F1 (0.706) and macro AUPRC (0.776). Architecture search then identifies CNN-Large as the strongest single-split CNN, with 0.593 subset accuracy and 0.714 macro F1, while CNN-Micro provides a compact architecture for embedded inference. To match a four-sensor hardware design, we retrain CNN-Micro using channels associated with ECRB, ECRL, FDS, and FDP and exclude the ground electrode from model input. Across five seeds, cross-channel knowledge distillation improves the four-channel student over direct training, reaching $0.5219 \pm 0.0114$ subset accuracy, $0.7612 \pm 0.0038$ finger accuracy, and $0.6095 \pm 0.0058$ macro F1. The selected 123K-parameter model accepts nine windows of 48 features and has been exported to ONNX. These results establish a reproducible software path from post-stroke sEMG to compact five-finger intent prediction for subsequent hardware-in-the-loop evaluation.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.