News

Penn Researcher Awarded $8 Million to Advance Mental Health Diagnostics Using AI

LDI Senior Fellow Yong Chen Leads a Data Team in a New Five-Year NIH Project

April 28, 2025

By:

Hoag Levins

Print:

How can artificial intelligence (AI) and machine learning (ML) be used to develop more precise psychiatric phenotypes for mental health care diagnoses and treatment? That goal is what LDI Senior Fellow and Penn Medicine Professor Yong Chen, PhD, will be working on for the next five years in a project that has the potential to revolutionize the way mental health disorders are studied and treated.

Chen, who directs both the Penn Computer, Inference and Learning Lab (PennCIL) and the Center for Health AI and Synthesis of Evidence (CHASE), has received an $8 million grant from the National Institute of Mental Health, together with Yale University and the Mayo Clinic. He will lead the data coordinating center for a multidisciplinary collaborative research project—the more than $150 million Individually Measured Phenotypes to Advance Computational Translation in Mental Health (IMPACT-MH) initiative. Eight teams will be using behavioral measures and computational methods to define novel clinical patient profiles to support clinical decision-making in mental health care.

New Kind of Integrated Data

“The IMPACT-MH project is highly unique because it seeks to apply precision medicine concepts in the mental health field, where heterogeneity is especially challenging,” said Chen, a Professor of Biostatistics, Epidemiology, and Informatics at the Perelman School of Medicine. “The integration of multimodal data, including behavioral, clinical, and biological information, into computational phenotyping is a groundbreaking approach that promises to revolutionize patient care by making mental health diagnostics and treatment more personalized, precise, and data driven. This level of integration is not yet common in mental health research.”

“The project is distinct from our other work in its ambitious scale and focus,” Chen continued. “While PennCIL has extensive experience in computational methods, machine learning, and AI across various clinical fields, this project is unusual in its application to the complexities of mental disorders. It also stands out for its multi-institutional collaboration, aiming to set new standards for mental health data collection, representation, and analysis.”

Beyond the Observable

Currently, mental health clinicians face major hurdles due to overly broad diagnostic categories and a lack of clearly defined patient profiles, making treatment decisions more difficult. They may struggle to capture the full picture using standard tools like checklists or interviews, relying heavily on the observable, but two patients with the same diagnosis may have different symptoms and respond differently to treatment.

A phenotype is the set of characteristics or traits of an individual. In mental health, this can include symptoms and behaviors, cognitive performance, biological markers such as brain imaging findings and hormone levels, and responses to treatment. A clinical phenotype is a detailed profile of how a person’s condition presents—not just in symptoms but across these various biological, psychological, and behavioral domains.

Tailored Treatment Approaches

“In a clinical situation, the finished IMPACT-MH product could be used to improve diagnosis, prognosis, and treatment personalization,” said Chen. “For example, clinicians could use these signatures to predict how individual patients might respond to specific treatments, monitor their progress more effectively, and make more informed decisions regarding the course of treatment. The goal is to foster personalized care, reduce heterogeneity in mental health diagnoses, and improve patient outcomes by tailoring treatment approaches based on new kinds of clinical signatures.”