Single-Channel EMG Classification With Ensemble-Empirical-Mode-Decomposition-Based ICA for Diagnosing Neuromuscular...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neurology

Background:

• Quantitative analysis of electromyography (EMG) signals is crucial for diagnosing neuromuscular disorders and related applications.
• Standard independent component analysis (ICA) methods struggle with extracting sources from single-channel EMG data.
• Existing techniques often require multi-channel recordings, limiting their application.

Purpose of the Study:

• To develop and validate a classification method for neuromuscular disorders using single-channel EMG data.
• To address the limitations of traditional ICA in processing low-channel EMG signals.
• To improve the accuracy and efficiency of EMG-based diagnostic tools.

Main Methods:

• Ensemble empirical mode decomposition (EEMD) to denoise and decompose single-channel EMG signals into intrinsic mode functions (IMFs).
• FastICA algorithm applied to separated IMFs for source extraction.
• Extraction of five time-domain features from the separated components.
• Classification using linear discriminant analysis (LDA) with a majority voting refinement.

Main Results:

• The proposed method successfully decomposes and separates sources from single-channel EMG signals.
• A high classification accuracy of 98% was achieved using the developed feature extraction and classification pipeline.
• The method demonstrates robust performance on a clinical EMG database.

Conclusions:

• The developed single-channel EMG analysis method offers an accurate and computationally efficient approach for classifying neuromuscular disorders.
• This technique shows promise for assisting clinicians in diagnosing neuromuscular conditions, potentially extending to real-world clinical settings.
• The findings support the advancement of non-invasive diagnostic tools for neurological conditions.