Deep learning using electroencephalogram (EEG) data for diagnosing and predicting SSRI response in major depressive...

Area of Science:

• Neuroscience
• Computational Psychiatry
• Medical Diagnostics

Background:

• Major Depressive Disorder (MDD) presents a significant global health burden.
• Current MDD diagnosis relies on subjective methods, leading to high rates of non-response to first-line treatments.

Purpose of the Study:

• To investigate the efficacy of deep learning (DL) algorithms applied to electroencephalogram (EEG) data for diagnosing MDD.
• To predict treatment outcomes for selective serotonin reuptake inhibitor (SSRI) therapy in MDD patients.

Main Methods:

• Utilized six independent datasets comprising 146 healthy subjects and 203 MDD patients.
• Trained and tested DL models on unseen EEG data to assess diagnostic and predictive accuracy.
• Employed Gradient-weighted Class Activation Mapping (Grad-CAM) for interpreting classification features.

Main Results:

• Achieved 67.5% average accuracy (70% best fold) in distinguishing MDD patients from controls.
• Demonstrated 79% average accuracy (85% best fold) in predicting SSRI treatment responders.
• Identified frontal and parietal alpha activity as key EEG markers; model-guided selection improved SSRI response rates.

Conclusions:

• EEG-based DL models show clinical promise for stratified MDD treatment and personalized therapy selection.
• Integrating objective neurophysiological markers can enhance treatment allocation in psychiatry.
• This approach has the potential to reduce ineffective treatments and improve patient outcomes.