DiffMDD: A Diffusion-Based Deep Learning Framework for MDD Diagnosis Using EEG.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Major Depression Disorder (MDD) is a prevalent mental health condition with significant global impact.
• Electroencephalography (EEG) is a valuable non-invasive tool for MDD diagnosis, but faces challenges with data noise and limited dataset sizes.
• Overfitting is a common issue in deep learning models for MDD diagnosis due to data quality and size limitations.

Purpose of the Study:

• To develop and validate DiffMDD, a diffusion-based deep learning framework for accurate MDD diagnosis using EEG.
• To address challenges in EEG data quality and quantity for improved deep learning model performance.
• To enhance the robustness and generalization capabilities of MDD diagnostic models.

Main Methods:

• Proposed DiffMDD framework incorporating a Forward Diffusion Noisy Training Module to extract noise-irrelevant features.
• Implemented a Reverse Diffusion Data Augmentation Module to increase data size and diversity.
• Re-trained a classifier on the augmented EEG dataset for MDD diagnosis.

Main Results:

• DiffMDD demonstrated improved model robustness against EEG noise.
• Data augmentation significantly increased dataset size and diversity, aiding generalized feature learning.
• The framework achieved state-of-the-art performance on two public MDD diagnosis datasets.

Conclusions:

• DiffMDD effectively addresses key challenges in EEG-based MDD diagnosis.
• The proposed diffusion-based framework enhances diagnostic accuracy and model generalization.
• This approach offers a promising advancement for the early and accurate detection of MDD.