Long short-term memory (LSTM) recurrent neural network for muscle activity detection.
Marco Ghislieri, Giacinto Luigi Cerone, Marco Knaflitz and 1 others
PMID 34674720WHAT IT FOUND
A new AI method detects muscle activity from surface electromyography more accurately than standard techniques, especially in noisy signals.
It requires no manual threshold setting and shows promise for gait analysis and rehabilitation, though validation was limited to walking in a small group.
Key findings
01The LSTM-based detector achieved higher F1-scores than standard and statistical methods on both simulated (0.95 vs 0.87/0.76) and real (0.91 vs 0.76/0.74) data.
02The LSTM method maintained better performance consistency across varying signal-to-noise ratios, with minimal degradation even at low SNR levels.
03Onset and offset timing errors for the LSTM method were significantly smaller (under 6 ms average bias) compared to the other detectors.
STILL TO COME
How it was doneWhat they foundWhat it means for PTsWhat it means for OTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The real data sample size was small (20 subjects). Validation was restricted to walking; performance on other movements or pathological conditions like stroke or spinal cord injury was not tested. The ground truth for real data relied on manual segmentation by experts, which is inherently subjective and time-consuming. The study did not evaluate the method's performance in the presence of spurious background spikes common in some neurological conditions.
The easy way to misread this
Do not assume this method is ready for immediate clinical deployment or generalizable to all patient populations. The validation was limited to a small group of subjects during walking only, and the 'ground truth' was based on expert manual segmentation, which has its own subjectivity.