tsta

فارسی

Home / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

Toward Precision Psychiatry: Differentiating Depression and Psychosis Using EEG-Based Machine Learning Models

Authors :

Vahid Asayesh¹ Mehdi Dehghani² Majid Torabi³ Sepideh Akhtari-Khosroshahi⁴ Maedeh Akhtari-Khosroshahi⁵ Sebelan Daneshvar⁶

1- NPCindex Research Company, Asayesh Psychiatric Clinic 2- NPCindex Research Company 3- NPCindex Research Company, Asayesh Psychiatric Clinic 4- NPCindex Research Company 5- NPCindex Research Company 6- Department of Electrical and Computer Engineering, Brunel University

Keywords :

Major Depressive Disorder،Psychosis،Machine Learning،Statistical Analysis،Differential Diagnosis

Abstract :

Accurate differential diagnosis of major depressive disorder, psychosis, and psychotic depression remains a major challenge in psychiatry due to overlapping clinical symptoms and the reliance on subjective assessment tools. This study aimed to identify objective EEG-based neuromarkers that distinguish between these three diagnostic groups using quantitative EEG features and machine learning. Resting-state EEG was collected from 60 medication-free adult patients clinically diagnosed via DSM-based psychiatric interviews. A total of 26 linear and nonlinear features, including relative power, cordance, alpha peak frequency, Lempel–Ziv complexity, largest Lyapunov exponent, and sample entropy, were extracted across 19 scalp channels. Statistical group comparisons with FDR correction identified significant feature–channel combinations, which were then ranked using the MRMR algorithm. Stepwise feature inclusion and classification with SVM, KNN, and MLP were performed. Results showed that 42 selected features achieved optimal accuracy, with KNN yielding 96.53±0.20% accuracy. These findings highlight the potential of EEG neuromarkers to support precision psychiatry and improve diagnostic specificity.