Comparing resting state and task-based EEG using machine learning to predict vulnerability to depression in a...

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Major Depressive Disorder (MDD) poses a significant societal burden, driving interest in its prediction and understanding.
• Electroencephalography (EEG) is a key neural measure for studying MDD mechanisms, yet studies often focus on either resting-state (rs-EEG) or task-based EEG, not both.
• Comparing the efficacy of rs-EEG and task-based EEG for predicting depression vulnerability is crucial for advancing diagnostics.

Purpose of the Study:

• To compare the predictive power of resting-state EEG (rs-EEG) and task-based EEG for depression vulnerability.
• To identify neural markers associated with varying levels of depression vulnerability.
• To explore the utility of machine learning models and evolutionary algorithms for EEG-based depression diagnostics.

Main Methods:

• Collected questionnaire and EEG data from 40 participants with varying depression vulnerability scores.
• Analyzed both resting-state EEG (rs-EEG) and task-based EEG data from a sustained attention to response task.
• Applied Long Short-Term Memory (LSTM) and 1D-Convolutional Neural Networks (CNNs) for prediction, and evolutionary algorithms to identify key EEG biomarkers.

Main Results:

• Individuals more vulnerable to depression showed distinct EEG amplitude patterns in frontal, occipital, temporal, and parietal regions for both rs-EEG and task-based data.
• A 1D-CNN model achieved 98.06% accuracy using rs-EEG data, outperforming an LSTM model's 91.42% accuracy with task-based delta waves.
• Evolutionary algorithms identified Higuchi fractal dimension, phase lag index, correlation, and coherence as important rs-EEG biomarkers for predicting depression vulnerability.

Conclusions:

• Resting-state EEG (rs-EEG) demonstrates superior performance for predicting depression vulnerability compared to task-based EEG.
• Task-based EEG may provide more nuanced insights into cognitive mechanisms underlying depression, such as rumination.
• The study highlights the potential of rs-EEG combined with machine learning and specific biomarkers for developing future EEG-based diagnostics for MDD.