Automatic detection of major depressive disorder using electrodermal activity.

Area of Science:

• Psychiatry and Neuroscience
• Biomedical Engineering
• Computational Biology

Background:

• Major depressive disorder (MDD) is a leading cause of global disability.
• Current diagnostic methods for depression lack objectivity and efficiency, relying on clinical interviews and self-report scales.
• There is a need for more objective and efficient methods for depression screening.

Purpose of the Study:

• To develop and evaluate a machine learning approach for screening major depressive disorder (MDD) using electrodermal activity (EDA).
• To assess the feasibility of using physiological signal changes during autonomic arousal and recovery for depression detection.

Main Methods:

• Electrodermal activity (EDA) was recorded from 30 MDD patients and 37 healthy controls during five experimental phases.
• Features were extracted from EDA data, including differential features between phases.
• Support Vector Machine Recursive Feature Elimination (SVM-RFE) was used for feature selection, followed by classification with a decision tree.

Main Results:

• A decision tree classifier achieved 74% accuracy, 74% sensitivity, and 71% specificity in detecting MDD.
• Key features for classification included differential EDA measures and those from stress and relaxation tasks.
• The findings indicate that EDA patterns during autonomic challenges are informative for MDD screening.

Conclusions:

• Machine learning-based screening for MDD using electrodermal activity is feasible.
• Monitoring physiological responses during stress and recovery phases enhances the discriminative power for depression detection.
• This approach offers a potential objective and efficient tool for augmenting current depression diagnostic practices.