A machine learning based depression screening framework using temporal domain features of the electroencephalography...

Area of Science:

• Neuroscience
• Psychiatry
• Biomedical Engineering

Background:

• Depression is a prevalent mental health disorder requiring accurate diagnosis.
• Electroencephalography (EEG) shows potential for identifying neural markers of depression.
• Early detection is crucial for effective depression treatment and management.

Purpose of the Study:

• To develop and validate an EEG-based mechanism for detecting mental depressive disorders.
• To utilize the Multi-modal Open Dataset for Mental-disorder Analysis (MODMA) for depression detection.
• To enhance the accuracy and efficiency of depression classification using EEG data.

Main Methods:

• EEG data from 55 participants were analyzed using 3 electrodes in a resting-state condition.
• Twelve temporal domain features were extracted and processed through a novel feature selection algorithm.
• Classification was performed using Best-First (BF) Tree, k-nearest neighbor (KNN), and AdaBoost algorithms.

Main Results:

• A novel feature selection mechanism identified optimal attributes with high discriminative power.
• The highest classification accuracy of 96.36% was achieved using the BF-Tree classifier.
• The proposed method demonstrated superior performance compared to existing state-of-the-art depression classification schemes.

Conclusions:

• The developed EEG-based framework offers a highly accurate method for detecting mental depressive disorders.
• The approach utilizes a minimal number of electrodes and a concise feature vector, enhancing practicality.
• This framework has the potential to significantly support psychiatrists in clinical settings for depression diagnosis.