Neurodegenerative Disease Classification Using Gait Signal Features and Random Forest Classifier

Neurodegenerative diseases (NDDs), which occur due to progressive loss of function or structure of neurons, are a fundamental cause of abnormalities in gait dynamics and cause movement disorder in humans. Therefore, gait time-series signals can be used to predict NDDs like Parkinson's disease (PD), Huntington's disease (HD), and Amyotrophic Lateral Sclerosis (ALS). In this paper, linear, non-linear, statistical features of the gait signal and four physiological features of the human subjects were used to train random forest classifier to predict the NDDs. Kernel Principal Component Analysis (PCA) with Gaussian Kernel was employed to improve accuracy and prevent overfitting. The proposed method displayed an overall 95.26% prediction accuracy and provided a unique set of features that could be used to analyze NDDs further.