Discovering the neuronal dynamics in major depressive disorder using Hidden Markov Model

Area of Science:

• Neuroscience
• Computational Psychiatry
• Biomedical Engineering

Background:

• Major Depressive Disorder (MDD) poses a significant global health challenge, with current treatments lacking robust neurological evidence.
• Electroencephalography (EEG) offers a method to record neural activity and potentially provide objective evidence for MDD.

Purpose of the Study:

• To propose and validate a probabilistic graphical model for decoding neural dynamics in MDD patients and healthy controls.
• To investigate the potential of EEG-based neural dynamics analysis for understanding MDD pathogenesis.

Main Methods:

• Utilized a Hidden Markov Model with Multivariate Autoregressive observation (HMM-MAR) for neural dynamics decoding.
• Applied the HMM-MAR model to the MODMA dataset, comprising resting-state and task-state EEG data from 53 participants (24 MDD, 29 HC).

Main Results:

• The model's state time courses correlated with Patient Health Questionnaire-9 (PHQ-9) scores and differentiated MDD from healthy controls.
• Observed Markov property in neuronal dynamics during sad face stimuli, with coherence and power spectrum analyses aligning with prior MDD research.

Conclusions:

• The HMM-MAR model demonstrates potential in capturing and interpreting neuronal dynamics from EEG signals for brain disease pathogenesis.
• Offers superior spatiotemporal interpretability compared to black-box machine learning or deep learning models in MDD research.