Prediction of Gait Neurodegenerative Diseases by Variational Mode Decomposition Using Machine Learning Algorithms

Abstract

Degeneration in brain cells is the cause of neurodegenerative illness, which cause motor function impairment. Anomaly in walking is one of this impairment’s adverse effects. Since sensor technologies and artificial intelligence applications have advanced in recent years, it is possible to predict a patient’s disease severity from their gait data using a model helps to grade the severity of the patient disease. The proposed research work is developed using the gait neurodegenerative data composed of gait signals collected from physionet database. The composite non stationary and nonlinear gait signals are decomposed into different modes of Intrinsic Mode Function (IMF) using Variational Mode Decomposition (VMD). IMFs ensure the signals retrieved are stationary and linear. Power Spectral Density (PSD) is used to choose best IMF. Feature extraction for the chosen IMF is done with Shannon entropy technique. Prediction of different neurodegenerative disease such as Parkinsons Disease (PD), Huntingtons Disease (HD), Amyotrophic Lateral Sclerosis (ALS) and healthy subjects are carried out by Multilayer Perceptron (MLP) with the parameters optimized with Genetic Algorithm (GA). The results of the experiments shows that the proposed model produces a better accuracy of 98.4% than the other existing algorithms for the gait dataset.