OPTIMIZED SINGLE EEG CHANNEL SELECTION FOR DETECTING MAJOR DEPRESSIVE DISORDER

Shruthi Narayanan Vaniya, Ahsan Habib, Sheik Mohammed Shariful Islam, Maia Angelova, Chandan Karmakar

Keywords: Major depressive disorder, deep learning, electroencephalogram, EEG, single-channel

TL;DR: This paper presents a deep learning framework for EEG channel selection for the detection of major depressive disorder.

Abstract: Major depressive disorder (MDD) or depression is a chronic mental illness that significantly impacts individuals' well-being and is often diagnosed at advanced stages, increasing the risk of suicide. Current diagnostic practices, which rely heavily on subjective assessments and patient self-reports, are often hindered by challenges such as under-reporting and the failure to detect early, subtle symptoms. Early detection of MDD is crucial and requires monitoring vital signs in everyday living conditions. Electroencephalogram (EEG) is a valuable tool for monitoring brain activity, offering critical insights into MDD and its underlying neurological mechanisms. While traditional EEG systems typically involve multiple channels for recording, making them impractical for home-based monitoring, wearable sensors can effectively capture single-channel EEG data. However, generating meaningful features from this data poses challenges due to the need for specialized domain knowledge and significant computational power, which can hinder real-time processing. To address these issues, our study focuses on developing a deep learning model for the binary classification of MDD using single-channel EEG data. We focused on specific channels from various brain regions, including central (C3), frontal (Fp1), occipital (O1), temporal (T4), and parietal (P3). Our study found that the channels Fp1, C3, and O1 achieved an impressive accuracy of 88\% when analyzed using a Convolutional Neural Network (CNN) with leave-one-subject-out cross-validation. Our study highlights the potential of utilizing single-channel EEG data for reliable MDD diagnosis, providing a less intrusive and more convenient wearable solution for mental health assessment.

Primary Area: learning on time series and dynamical systems

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 10084