MSDB-Mam: Dual-Branch Mamba Network With Multi-Scale Features for EEG-Based Depression Detection

Area of Science:

• Neuroscience
• Computational Psychiatry
• Biomedical Engineering

Background:

• Depression is a prevalent mental disorder with significant global impact.
• Electroencephalogram (EEG) offers a non-invasive window into brain activity for depression assessment.
• EEG signal complexity across spatial, temporal, and spectral domains presents detection challenges.

Purpose of the Study:

• To develop an advanced deep learning model for accurate, non-invasive depression detection using EEG.
• To address the challenges posed by the multi-dimensional complexity of EEG signals.
• To improve the efficacy of EEG-based depression detection systems.

Main Methods:

• Proposed a novel Multi-Scale Dual-Branch Mamba network (MSDB-Mam) for EEG feature extraction and fusion.
• Employed Multi-Scale Convolution (MSC) to capture diverse temporal and spatio-temporal patterns.
• Introduced a Dual-Branch Mamba (DB-Mam) architecture with Temporal-Spectral (TS-Mam) and Spatial-Temporal-Spectral (STS-Mam) branches to model complex dependencies.

Main Results:

• The MSDB-Mam network achieved high accuracy rates of 96.58% on the MODMA dataset and 96.66% on the PRED+CT dataset.
• The proposed method demonstrated superior performance compared to existing state-of-the-art approaches.
• The model effectively extracted and fused multi-dimensional EEG features for robust depression detection.

Conclusions:

• The MSDB-Mam network presents a highly effective and accurate method for EEG-based depression detection.
• The model's ability to handle complex, multi-dimensional EEG data signifies a significant advancement in the field.
• This approach holds promise for developing more accessible and objective diagnostic tools for depression.