Discrimination of Depression Levels Using Machine Learning Methods on EEG Signals

• Published in: Iranian Conference on Electrical Engineering pp. 1765 - 1769
• Main Authors: Mohammadi, Yousef, Hajian, Mojtaba, Moradi, Mohammad Hassan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.04.2019
• Subjects: Complexity theory; Depression; EEG; Electroencephalography; Entropy; Feature extraction; Fractal Dimensions; Fractals; Fuzzy Entropy; Fuzzy Function; Nonlinear Systems; Support vector machines; Time series analysis
• ISSN: 2642-9527
• DOI: 10.1109/IranianCEE.2019.8786540

Depression is a mental disorder which has direct effects on electroencephalography (EEG) of patients, that made EEG analysis a beneficial way for a depression diagnosis. A precise system which can diagnose the depression levels based on the EEG signal would be useful support. This paper presents a machine learning approach to discriminate the depressed subjects to four different levels of depression, according to the Beck depression inventory (BDI-II) scores, besides the separability of different levels is investigated. In this way, we also proposed a fuzzy function based on neural network (FFNN) classifier. Our dataset contains EEG signals recorded from 60 depressed subjects with different levels of depression, under resting state, and EEG analysis was done using nonlinear features including fuzzy entropy (FuzzyEn), Katz fractal dimension (KFD) and fuzzy fractal dimension (FFD). The results indicate that KFD has a better capability in the prediction of the depression level. The proposed fuzzy classifier has demonstrated significant supremacy compared to support vector machine (SVM) in almost all experiments.