CWT-based transfer learning model with optimal channel selection for the detection of MDD using EEG signals

Major Depressive Disorder (MDD) is one of the most prevalent and severe psychiatric conditions that pose significant diagnostic challenges due to the complex and subtle nature of brain dynamics. Conventional EEG analysis methods often fall short in capturing these nuanced temporal–frequency dynamics. To address this limitation, we propose a deep learning framework that integrates Continuous Wavelet Transform (CWT), scalogram representations, and transfer learning with the VGG16 architecture. The rationale for employing CWT and scalograms lies in their ability to preserve both time and frequency information, enabling the extraction of clinically meaningful features of brain dynamics, while transfer learning with VGG16 leverages prior knowledge from large-scale image datasets, improving feature generalization and reducing the computational burden of training on limited EEG data. By combining these techniques with robust preprocessing and channel selection, the proposed model effectively identifies distinct spectral power alterations, particularly in theta and alpha bands, associated with MDD. The experimental results demonstrate that the proposed model outperformed traditional methods, achieving a more precise detection of MDD. These findings highlight the promise of combining advanced signal processing with deep learning for more reliable, non-invasive diagnosis of MDD through EEG signals, and support its clinical application.