Automated Identification of Depressive Disorder using Multivariate Iterative Filtering

Abstract

This paper proposes a multivariate iterative filtering (MIF)-based framework to classify individuals as healthy or having major depressive disorder (MDD) using multichannel electroencephalogram (EEG) signals. MIF has been employed to decompose EEG signals into narrow-band oscillatory components, from which the instantaneous amplitude and instantaneous frequency are extracted using the Hilbert transform to construct the joint time-frequency (JTF) images. These JTF images have been given as input to a pretrained convolutional neural network (CNN) model, leveraging transfer learning to extract discriminative features for classification. The performance of the proposed framework has been evaluated on a publicly available EEG dataset, which has recordings from healthy controls and MDD patients under eye-closed (EC) and eye-open (EO) conditions. The proposed method obtained classification accuracies of 86.70% and 96.88% and sensitivities of 94.33% and 95.47% for EC and EO conditions, respectively, using a 10 -fold cross-validation procedure. The results show that the proposed MIF-based framework with deep learning algorithm provides an efficient method for MDD diagnosis. © 2026 IEEE.