A Novel Diagnosis System Using Regularized Encoder-Decoder Based Generative Probabilistic Network for Parkinson's Disease

摘要

This paper presents a novel way to address the classification problem of Parkinsons disease (PD) using the regularized encoder-decoder based generative probabilistic network. The progressive nervous system disorder, called PD, affected people continuously suffers low moving speed due to appear in motor disturbances. Additionally, decrease the arms swinging while walking, postural instability, and huffing gaits also appear for PD patients. In line with these problems, identifying PD has become one of the most challenging tasks in the medical sector. In this paper, a deep learning architecture is proposed to classify the PD patient with a detailed comparative analysis. The proposed model can not only optimize medical decisions but also reduce financial costs. The proposed network comprises an unsupervised encoder-decoder representation learning process for extracting the efficient feature that enables the unsupervised generative model to learn the probability distribution over the encoded representation. Finally, an artificial neural network is considered to perform the classification task. The experiment is conducted on 188 people where 107 people are affected with PD. Results show that the proposed model can achieve 93.1% accuracy that confirms high performance as compared to others.