MDD-TSVM: A novel semisupervised-based method for major depressive disorder detection using electroencephalogram...

Area of Science:

• Neuroscience
• Machine Learning
• Psychiatry

Background:

• Major Depressive Disorder (MDD) is a prevalent mental illness causing significant distress and potential suicide risk.
• Current clinical detection of MDD relies on electroencephalogram (EEG) signals and supervised learning, facing challenges with data labeling costs, subjectivity, and a prevalence of unlabeled data.
• Existing methods struggle with the inherent limitations of supervised learning in real-world clinical scenarios.

Purpose of the Study:

• To develop a novel semi-supervised learning method for accurate automatic detection of Major Depressive Disorder (MDD) using EEG signals.
• To address the limitations of supervised learning approaches, specifically the reliance on costly and subjective labeled data and the underutilization of abundant unlabeled EEG data.
• To improve the efficiency and accessibility of MDD diagnosis through advanced machine learning techniques.

Main Methods:

• A novel semi-supervised learning approach, termed MDD-TSVM, was developed, building upon the Transductive Support Vector Machine (TSVM) framework.
• The core innovation involves modifying the TSVM objective function by splitting the unlabeled data penalty term into two distinct pseudo-labeled components: one for MDD and one for non-MDD.
• This adaptation allows the model to effectively leverage both labeled and unlabeled EEG datasets while simultaneously mitigating the issue of class imbalance.

Main Results:

• The proposed MDD-TSVM method demonstrated superior performance in identifying MDD patients compared to existing state-of-the-art techniques.
• Achieved a high F1 score of 0.85 ± 0.05 and an accuracy of 0.89 ± 0.03 in classifying individuals with MDD.
• The method effectively utilized both labeled and unlabeled EEG data, showcasing the advantages of the semi-supervised approach.

Conclusions:

• The MDD-TSVM method presents a significant advancement in the automatic detection of Major Depressive Disorder using EEG signals.
• The semi-supervised approach effectively overcomes the limitations of traditional supervised methods by utilizing unlabeled data and addressing class imbalance.
• This research offers a promising, more efficient, and potentially less subjective tool for MDD diagnosis in clinical settings.