Keya Joshi, Feyza Yagli, Joban Clair, and Vildan Cavusoglu
Abstract:
Mental disorders are significant sources of global disease but are diagnosed inadequately using techniques often grounded on subjective judgment and lacking biological validation. Prevention tools such as cognitive-behavioral therapy, school programs, and mobile health apps are promising entry points for early intervention but are limited by constraints of scalability, long- term effectiveness, and access, particularly in low- and middle-income countries. Recently developed breakthroughs using artificial intelligence (AI) and physiological monitoring offer a promising complement. Electroencephalography (EEG) and Galvanic Skin Response (GSR), when supplemented with AI, can discern complex physiology patterns associated with stress, anxiety, and depression. While EEG is afflicted by issues of noise, lack of spatial resolution, and variability when used individually, multimodal combinations of EEG and other biosignals and GSR have shown rich promise of diagnosis. Reports detail machine learning models founded on these signals that achieve 79–95% accuracy rates for distinguishing between and predicting affective states and clinical populations, and predictive models permit pre-warning of risks of mental health issues before their onset. These findings describe the possibility of AI-driven multimodal physiological monitoring overcoming the problems of conventional diagnosis with scalable, objective tools with worldwide promise.