.png){kind=icon}
.png){kind=icon}
Return
Topological data analysis improves estimation of muscle fatigue and contraction level from surface electromyography data
.png){kind=icon}
PRE
AI
C
H
A
B
DOI：10.1088/1741-2552/ae6891.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Objective. Muscle fatigue affects individuals with neuromuscular disease and elite athletes alike. Muscle fatigue is difficult to reliably assess using non-invasive, real-time measures such as surface electromyography (sEMG) due to the coupling of fatigue and contraction level. The objective of this study was to explore the use of topological data analysis (TDA) with sEMG data as a method to assess muscle function. Approach. sEMG data was recorded from the first dorsal interosseous muscle across two experiments: 1) a sustained contraction experiment (n = 57) and 2) a discrete contraction level experiment (n = 18). We performed TDA on raw sEMG data and analyzed the same data with a traditional Fourier transform analysis towards the estimation of muscle fatigue and muscle contraction level. We analyzed the ability of these two approaches to 1) correlate to contraction time during the sustained contraction experiment, and 2) classify muscle contraction level from the discrete contraction experiment. Main Results. Vector TDA measures strongly correlated with contraction time during the sustained experiment on an individual trial level (up to R2 = 0.95). TDA leveraged the topological aspects of sEMG time series from this experiment to improve estimations of muscle contraction time using a deep network regression approach (mean R2 = 0.51). From a discrete contraction experiment at different contraction levels of measured maximum voluntary contractile force (20%-80%), implementation of KNN classification with frequency-based measures yielded 68% accuracy, while classification with TDA measures yielded 85% accuracy. Significance. These results suggest TDA could be used as a robust analysis approach to estimate muscle contraction level and muscle fatigue from sEMG data, although future work is needed for clinical implementation.
Keywords:
machine learning
deep learning
regression
median frequency
peripheral motor fatigue
rehabilitation
Journal
IF:
3.8
Papers:
4.0K
Citations:
1.4W
Organization
Cited Papers