A new Cepstral-based biomarker of reward positivity evaluated in Parkinson’s disease detection

Document Type : Original Article

Authors

1 Faculty of biomedical engineering_Sahand university of technology_Tabriz_Iran

2 Computational Neuroscience Laboratory, Faculty of Biomedical Engineering, Sahand University of Technology, Tabriz, Iran

Abstract

Parkinson’s disease (PD) is one of the central nervous system disorders that affect dopaminergic neurons in the substantia nigra, leading to impairments in midbrain dopaminergic functions. The development of efficient detection is required to control the impairments. In previous PD detection works, either the detection cost or complexity is high or the robustness of the method to clinical parameters or individual differences is low. This article provides a reliable PD detection method proposing a new marker of reward positivity using a Cepstral decomposition of electroencephalogram (EEG) signals discriminating oscillation and excitation components and providing amplitude and phase information while minimizing the number of analyzed coefficients. Cepstral analysis has been used for extracting a more effective representation of spectral information of the quasi-periodic signals using source-filter separation. The capability of this method has been evaluated using 28 patients on both ON and OFF medication states and 28 healthy control individuals during the reinforcement-learning task. It has achieved an average accuracy rate of 99.79% by minimizing the real Cepstrum coefficients to 250 from 4000 ones. It has also obtained satisfactory results on medication states and frontal channels (85% channel reduction) indicating the efficiency, robustness, and cost-effectiveness of the method.

Keywords

Main Subjects