Parkinson's Illness Identification Based on Decision Support System Using EEG Signals
Year: 2026
Type: conference-paper
Abstract: Parkinson's illness is a highly prevalent neurological illness that is distinguished through the reduction of dopamine release in the striatum and the deterioration of neurons that produce dopamine in the nigrostriatal pathway. Parkinson's illness is the worldwide second most widespread neurological illness and now lacks a viable remedy. Scientists are primarily concerned with prolonging the advancement of the disease, and to achieve this goal, it is imperative to identify Parkinson's disease at the earliest stage feasible. This research introduces a decision support system that is intended to enable the detection of Parkinson's illness utilizing Discrete Wavelet Transforms (DWT) based on electroencephalogram (EEG) signals. First, these EEG data are preprocessed before the DWT method is utilized to divide them into five frequency subbands. The decomposed sub-bands are utilized to compute several features, which are then picked based on their importance level in identifying Parkinson's disease using the binary gray wolf optimization process. Several classifiers are employed to determine suitable feature sets, such as support vector machines (SVM), decision trees (DT), logistic regression, k-nearest neighbors(K-NN), and random forests (RF). The proposed methodology is assessed using the University of SanDiego dataset)15 individuals with PD were assessed in both ON /OFF Medication states and 16 healthy controls (HC) at the University of San Diego. The KNN classifier produces an F1-score of 98.35% for HC VS OFF medication PD, with accuracy of 98.02%, specificity of 98.19%, sensitivity of 97.91%, and precision of 98.54%. The ON-medication PD VS HC results showed an accuracy of 98.58%, specificity of 99.10%, sensitivity of 98.76%, precision of 98.99%, and an F1-score of 98.37%. The OFF-medication PD VS ON-medication PD results showed an accuracy of 95.89%, specificity of 97.27%, sensitivity of 95.76%, precision of 97.21%, and an F1-score of 95.73% with the RF. The proposed method would be highly advantageous for neurologists in their diagnosis procedures and for existing clinical processes.
Authors: Zahraa Awad Ghani, Firas Sabar Miften
Institution: Thi Qar University
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: hybrid
OpenAlex ID: W7154994010