We are pleased to announce that Stanford Psychiatry’s Kilian Pohl, Professor of Psychiatry and Behavioral Sciences, and, by courtesy, of Electrical Engineering, has received a grant from the National Institute of Mental Health titled “Explainable AI for Predicting Incidents of Depression in People with HIV.” Dr. Pohl is joined on this grant by Co-PI Robert Paul, Professor at University of Missouri–St. Louis and Director of the Missouri Institute of Mental Health.
The prevalence of Major Depressive Disorder (MDD) among people with HIV is 20-50% — three times higher than the prevalence of MDD in the general population. Depression is a leading risk factor for continuity of care among people with HIV, including adherence to anti-retroviral therapy. Overlapping symptoms of depression with those of HIV and side effects of anti-retroviral therapy (e.g., psychomotor slowing, difficulty thinking, sleep disruption, fatigue, apathy, and other comorbidities (e.g., anxiety)) complicate clinical decision making and patient care. In addition to depression being underdiagnosed, the study team’s preliminary machine learning analyses indicate bidirectional associations between HIV and depression: (1) depressive symptoms predict chronic immune dysregulation among virally suppressed people with HIV; and (2) elevated levels of peripheral inflammatory markers despite sustained viral control predict depression incidence and persistence. These findings suggest an urgent need for novel approaches to accurately identify and manage depression in people with HIV.
The study team will examine the interplay between overlapping constructs, moving away from conventional a priori measurement selection and small single-study, group-level analyses, to integrative, multimodal analyses leveraging existing data acquired prospectively across multiple longitudinal studies. They will then create a pioneering, multimodal, explainable AI (xAI) framework that uses knowledge graphs to holistically capture individual-level neurobiological, cognitive, sociodemographic, and environmental determinants of depression in people with HIV.
“To better understand the underlying drivers of depression among people with HIV who are receiving antiretroviral therapy, we will create an explainable artificial intelligence framework that forecasts depression in individuals from multi-modal, prospective data,” say Dr. Pohl and Dr. Paul. “By sharing our validated explainable AI tools publicly, this research promises immediate, tangible benefits to HIV clinical care, with the potential of enabling precise identification and management of depression risk. Doing so will help refine clinical care strategies to reduce the clinical burden of depression in people with HIV.”
Dr. Pohl co-directs the AI for Mental Health initiative, which aims to transform research, diagnosis, and treatment of psychiatric and behavioral disorders by creating and using responsible AI. He also leads the Computational Neuroscience Laboratory (CNSLAB), which develops AI methods to identify phenotypes that advance personalized medicine and the prevention of neuropsychiatric disorders. The multidisciplinary research team focuses on MRI studies of substance abuse, depression, sleep, HIV, and adolescent brain development. Recent publications related to this work include “Integrating Anatomical Priors into a Causal Diffusion Model”, published in the journal IEEE Transactions on Medical Imaging, and “Amyloid-β PET Positivity Among Cognitively Impaired People with HIV Over Age 60,” published in the Journal of Acquired Immune Deficiency Syndromes.
More Information
For more details about this project, visit NIH RePORTER.