Classification of EMG Signals for Assessment of Neuromuscular Disorder using Empirical Mode Decomposition and Logistic Regression

• Published in: 2019 International Conference on Applied and Engineering Mathematics (ICAEM) pp. 237 - 243
• Main Authors: Khan, Muhammad Umar, Aziz, Sumair, Bilal, Muhammad, Aamir, Muhammad Bilal
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2019
• Subjects: amyotrophic lateral sclerosis (ALS); Diseases; Electromyography; electromyography (EMG); Empirical mode decomposition; empirical mode decomposition (EMD); Feature extraction; logistic regression (LR); Neuromuscular; support vector machine (SVM); Support vector machines
• DOI: 10.1109/ICAEM.2019.8853684

The electromyographic (EMG) signal generated in muscle fibers has been the topic under substantial research in immediate past years as it provides fairly large amount of information for assessment of neuromuscular diseases particularly amyotrophic lateral sclerosis (ALS). Besides this, the design of an accurate and computationally efficient diagnostic system remains a challenge due to variation of EMG signals taken from different muscles with different level of needle insertion. This study offers a complete framework for accurate classification of EMG signals which includes denoising by empirical mode decomposition (EMD), feature extraction from both the time and frequency domains and classification by logistic regression (LR) and support vector machine (SVM). The presented work efficiently discriminates between EMG signal of healthy subject and patient with ALS disease independent of which muscle is used for EMG signal acquisition and what insertion level of needle is. Performance evaluation measures such as sensitivity, specificity, F-measure, total classification accuracy and area under ROC curve (AVC) are used to validate the performance of both classifiers. LR classification technique shows superlative performance with a classification accuracy of 95.1%. These results shows the competence of proposed diagnostic system for classification of EMG signals. Moreover, the proposed method can be used in clinical applications for diagnoses of neuromuscular diseases.