Exploring MODMA Database for Major Depressive Disorder Diagnostic

Depression is a prevalent psychiatric disorder associated with diagnostic challenges, as conventional assessment tools such as the DSM-5 and HAMD may produce inconsistent results. This study investigates the potential of EEG and audio biomarkers to enhance diagnostic accuracy using the MODMA dataset. We employed a comprehensive pipeline encompassing preprocessing, feature extraction, feature selection using mRMR, and classification using SVM for both EEG and audio signals. The EEG-based classification achieved an accuracy of 76%, while the audio-based models reached 80.90% accuracy. These results suggest that speech features may offer a slight advantage in discriminating individuals with major depressive disorder (MDD).