HONG XIE, QUISHI WANG, PRAKRITI PODDAR
DOI: https://doi.org/Mental health disorders remain a critical global health concern, yet traditional diagnostic approaches often rely on subjective assessments, leading to variability and delays in treatment. This study explores the integration of artificial intelligence (AI)-powered computer vision within healthcare-based digital care platforms to enhance the accuracy and accessibility of mental health diagnostics. Video data from individuals with depression, anxiety, stress-related disorders, and healthy controls were analyzed to extract behavioral and visual parameters, including blink rate, gaze fixation, micro-expression frequency, and head movement variability. Statistical analyses demonstrated significant group differences, with strong correlations between extracted features and clinical scales (PHQ-9, GAD-7, PSS). Advanced AI models, particularly BiLSTM and Transformer architectures, achieved superior predictive performance, with accuracies exceeding 90% and AUC-ROC values above 0.94. These results highlight the capacity of AI-driven systems to detect subtle, clinically relevant patterns that complement traditional assessments. While ethical and interpretability challenges remain, the findings underscore the promise of AI-powered computer vision as a transformative tool for early detection, continuous monitoring, and scalable delivery of digital mental healthcare.
