Voice-Based Deep Learning Medical Diagnosis System for Parkinson's Disease Prediction
Asmae Ouhmida, Oumaima Terrada, Abdelhadi Raihani, Bouchaib Cherradi, Soufiane Hamida
Abstract
Nowadays, the biomedical signal processing area (MSP) is one of the most important research fields. It is often applied in medical diagnosis and early detection of neurological diseases. Thereby, the MSP is deployed in Parkinson’s disease (PD) detection from voice disorder. Therefore, Convolutional Neural Networks (CNN) and Artificial Neural Networks (ANN) are employed to classify healthy patients from PD ones, based on vocal features. We accomplished our study using two UCI Machine Learning repository databases, denoted database I and database II in the whole article. These datasets include 22 and 45 acoustic features, respectively. Accuracy, sensitivity, and specificity were calculated in order to qualify and evaluate the performance of the detection system. The experiment results reveal that the accuracy reached a rate of 93.10 % as the highest value when we applied the CNN model to database I.