A real-time and convex model for the estimation of muscle force from surface electromyographic signals in the upper and lower limbs

AML-DECODER: Advanced Machine Learning for HD-sEMG Signal Classification-Decoding Lateral Epicondylitis in Forearm Muscles

BACKGROUND

Innovative algorithms for wearable devices and garments are critical for diagnosing and monitoring disease (such as lateral epicondylitis (LE)) progression. LE affects individuals across various professions and causes daily problems.

METHODS

We analyzed signals from the forearm muscles of 14 healthy controls and 14 LE patients using high-density surface electromyography. We discerned significant differences between groups by employing phase-amplitude coupling (PAC) features. Our study leveraged PAC, Daubechies wavelet with four vanishing moments (db4), and state-of-the-art techniques to train a neural network for the subject's label prediction.

RESULTS

Remarkably, PAC features achieved 100% specificity and sensitivity in predicting unseen subjects, while state-of-the-art features lagged with only 35.71% sensitivity and 28.57% specificity, and db4 with 78.57% sensitivity and 85.71 specificity. PAC significantly outperformed the state-of-the-art features (adj. p-value < 0.001) with a large effect size. However, no significant difference was found between PAC and db4 (adj. p-value = 0.147). Also, the Jeffries-Matusita (JM) distance of the PAC was significantly higher than other features (adj. p-value < 0.001), with a large effect size, suggesting PAC features as robust predictors of neuromuscular diseases, offering a profound understanding of disease pathology and new avenues for interpretation. We evaluated the generalization ability of the PAC model using 99.9% confidence intervals and Bayesian credible intervals to quantify prediction uncertainty across subjects. Both methods demonstrated high reliability, with an expected accuracy of 89% in larger, more diverse populations.

CONCLUSIONS

This study's implications might extend beyond LE, paving the way for enhanced diagnostic tools and deeper insights into the complexities of neuromuscular disorders.

Unraveling cEMG-wet sEMG Correlation Dynamics: Investigating Influential Factors

The electromyography (EMG) signal provides insight into neuromuscular activity which is used in medical and technological fields. Traditional needle electrodes and surface electrodes have several drawbacks making them less suitable for portable and long-term use. In contrast, emerging capacitive electrodes offer promising features over the existing electrodes. Yet, the full potential of capacitive electrodes remains untapped due to the lack of comprehensive design optimization for consistently reliable signal quality. This study highlights the complex interplay of factors influencing correlation in capacitive EMG (cEMG) and wet surface EMG (wet sEMG) signals. The study emphasizes the importance of the surface area of capacitive electrodes, muscle force, preprocessing, and sampling frequency in understanding and improving the correlation between cEMG and wet sEMG signals, providing valuable insights for future research and applications in the field. The study reveals that the electrode area has no significant effect on the correlation. However, the correlation significantly depends on the muscle force. In addition, removing artifacts from the cEMG signal increases the correlation, especially for lower force where artifacts are significant. Again, oversampling the EMG signal above 800 Hz does not have any impact on increasing the correlation but the correlation decreases with higher inter-electrode distance (IED). In this research, the highest correlation of 82.89% (normalized-91.62%) between cEMG and sEMG has been achieved for high muscle force with a plate area of 4 cm2. Therefore, the capacitive electrode can be an alternative for EMG signal acquisition.

A new force profile signal for a convex solution of muscle force estimation from electromyographic signals. ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY. IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY. ANNUAL INTERNATIONAL CONFERENCE 2023;2023:1-4. [PMID: 38082591 DOI: 10.1109/embc40787.2023.10340594] [Citation(s) in RCA: 0] [Impact Index Per Article: 0]

High-Density Surface Electromyography (HD-sEMG) is a non-invasive technique for measuring the electrical activity of a muscle with multiple, closely spaced electrodes. Estimation of muscle force is one of the applications of HD-sEMG. Usually, validating different EMG-Force models entails simple movements limited to laboratory settings. The validity of these models in more ecological conditions, requesting force production over a wide frequency band, remains unknown. In this study, we, therefore, compare the results of force prediction using four different types of input force profiles that can be representative of daily life activities, and we investigate whether the crest factor of these different input signals affects force prediction. For predicting the force from sEMG signals, we used our real-time and convex methods. HD-sEMG signals were recorded with 144 channels from the biceps brachii, brachioradialis, and triceps (long, lateral, and medial head) muscles of 24 healthy subjects during random signal, random phase, Schroeder phase, and minimum crest factor (crestmin) signal. The correlation and coefficient of determination (R2) between measured and predicted forces were calculated for the different force feedback profiles. The crestmin signal showed significantly better results based on statistical tests (P-value < 0.05), with correlation and R2 equal to 0.92±0.03 and 0.86±0.05, respectively. The results demonstrate that the crest factor of input signals is a crucial parameter that can impact the performance of EMG-Force models and must be considered during training.Clinical Relevance- This study demonstrates that lower crest factor multisine force profiles result in improved fitness for force prediction and can be used as an alternative to random signals.