Voice signal processing for detecting possible early signs of Parkinson’s disease in patients with rapid eye movement sleep behavior disorder

Abstract

In this study we introduced a method for early detecting of Parkinson’s disease (PD) in patients with rapid eye movement sleep behavior disorder (RBD). Patients suffering from RBD are at extremely high risk (> 80%) for developing PD as well as other related neurodegenerative disorders. The database used in this study contains 30 PD patients in the very early stages, 50 RBD patient and 50 healthy subjects (HS). First, we created a model with a maximal accuracy of 85% of discrimination between PD and HS by testing different combinations of acoustic features along with different kernels of SVM and leave one subject out validation scheme. Based on that model, we tested 50 RBD patients in order to see whether they will belong to PD or HS groups. As a result we found 66% of RBD patients were classified as PD. Based on these foundlings we confirmed the existence of a correlation between RBD patients and early PD patients using speech analysis and thus, early PD signs can be reliably captured. These results will lead to the development of an embedded system for detecting the possible early signs of PD and other neurodegenerative diseases.