Enhanced Predictive Modeling for Neuromuscular Disease Classification: A Comparative Assessment Using Gaussian Copula Denoising on Electromyographic Data

Abstract

This study presents a methodology for automatically detecting neuromuscular diseases through prepro-cessing and classifying electromyography (EMG) signals. The presented approach integrates Gaussian Copula-based denoising techniques with feature extraction and Random Forest classification. To assess the performance, the study performs a comprehensive evaluation of various denoising techniques, including Empirical Mode Decomposition (EMD), Variational Mode Decomposition (VMD), Wavelet Thresholding Denoising (WTD), and Gaussian Copula Denoising (GCD). The study also compares the effectiveness of several classification algorithms, such as Random Forest (RF), Convolutional Neural Networks (CNN), Multilayer Perceptron (MLP), and Decision Tree (DT). The methodology demonstrated exceptional per-formance, achieving an overall accuracy greater than 99% in distinguishing between healthy, myopathic, and neuropathic EMG signals. The proposed method's effectiveness is attributed to its noise reduction ca-pabilities, feature selection focusing on mean amplitude and amplitude range, and the Random Forest al-gorithm's adeptness in classifying EMG data. The study's findings underscore the proposed method's ac-curacy and effectiveness and highlight its potential to revolutionize clinical diagnostics of neuromuscular disorders, offering a powerful tool for more precise and timely interventions. Keywords: Electromyography; Denoising; Classification; Neuromuscular Diseases; Gaussian Copula; Random Forest; EMG; CNN.