EMG Signal Analysis using Improved Wigner- Ville Distribution Model And Long Short- Term Memory For Amyotrophic Lateral Sclerosis Prediction

Amyotrophic Lateral Sclerosis (ALS) is a degenerative disease of the motor neuron function and physical ability that highlights the importance of early diagnosis to recognize ALS and improve overall outcomes. In this study, we propose a hybrid novel framework that utilizes improved Wigner-Ville Distribution (WVD) to improve time-frequency representations of electromyography (EMG) signals, convolutional neural networks (CNNs) to capture spacial features of EMG signals and long short-term memory (LSTM) networks to model dependencies in time. The improved WVD signals do provide a time-frequency representation that provides informative extraction of features suited to clinical features. CNNs can capture important patterns regarding the spatial features and complexity of EMG signals. The LSTM network makes use of the sequential dependencies in the EMG signal to enhance the classification of ALS. The hybrid framework conclusively outperforms existing conventional methods, proving to be a reliable and efficient means in clinical practice for ALS detection.