PsychSens: A Framework for Automated and Pervasive Psychiatric Comorbidity Screening Using Multimodal Wearable Sensor Data

Each year, millions of individuals in the U.S. experience mental illness, yet as of March 2024, over 122 million people reside in areas with shortages in the mental health workforce. Machine learning-based screening using wearable sensors offers a timely, low-intervention approach to mental health support. However, current methods typically focus on single disorders, neglecting the more complex and clinically relevant challenge of psychiatric comorbidity—an oversight that limits their practical utility. This proposal addresses that critical gap by leveraging multimodal sensor data to detect comorbid conditions, with the goal of reducing mental health burdens among college students through early warning systems and precise interventions. The accompanying educational plan enhances undergraduate and graduate curricula with hands-on machine learning projects centered on sensor-based mental health analysis, and extends these opportunities to K–12 students through summer research camps. PsychSens will significantly advance sequential neural machine learning models for automated, pervasive screening of psychiatric comorbidities across three key dimensions: (1) It introduces the first hierarchical state-space model capable of efficiently and robustly processing ultra-long sensor data. (2) It enables personalized, privacy-aware risk stratification through novel sociodemographic encoding and individual-level differentially private model training. (3) It supports automated biomarker discovery via innovative sequential influence functions. By bridging the gap in psychiatric comorbidity screening through wearable sensor data, PsychSens aims to enable more accurate diagnoses and effective treatments—aiming ultimately to reduce suicidality and to improve mental health outcomes. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.