A virtual reality-based multimodal framework for adolescent depression screening using machine learning.

Area of Science:

• Neuroscience
• Psychiatry
• Biomedical Engineering

Background:

• Adolescent major depressive disorder (MDD) is a growing global concern.
• Current MDD screening relies on subjective self-reports, limiting diagnostic accuracy.
• Virtual reality (VR) integrated with physiological sensing presents an opportunity for objective MDD diagnostics.

Purpose of the Study:

• To investigate the efficacy of a VR-based multimodal system for diagnosing adolescent MDD.
• To identify objective physiological biomarkers indicative of MDD in adolescents.
• To develop a machine learning model for classifying MDD status using physiological data.

Main Methods:

• A case-control study involving 51 adolescents with first-episode MDD and 64 healthy controls.
• Participants engaged in a 10-minute VR emotional task while electroencephalography (EEG), eye-tracking (ET), and heart rate variability (HRV) were recorded.
• Statistical analysis identified key physiological differences, and a support vector machine (SVM) model was trained for MDD classification.

Main Results:

• Adolescents with MDD exhibited higher EEG theta/beta ratios and HRV LF/HF ratios (p <.05).
• Reduced saccade counts and longer fixation durations were observed in the MDD group.
• The SVM model achieved 81.7% classification accuracy with an Area Under the Curve (AUC) of 0.921.

Conclusions:

• The VR-based multimodal system effectively identified physiological biomarkers for adolescent MDD.
• The system demonstrated strong diagnostic performance, supporting its use in early mental health screening.
• Future research may focus on integrating this system into wearable or mobile platforms for scalable, real-world application.