Key Takeaways
- Researchers at The Ohio State University received grants from Meta and access to its latest artificial intelligence (AI) models to support three medical research projects.
- They will study how AI could help medical experts more efficiently read chest X-rays, spot signs of future heart failure and guide the diagnosis and care of people with Alzheimer's disease.
COLUMBUS, Ohio – Three teams of researchers from The Ohio State University College of Medicine and The Ohio State University received funding and access to the latest Muse AI models from Meta Superintelligence Labs (MSL) to advance medical research. MSL is the artificial intelligence research and development division of Meta.
The three research projects are:
Using AI to help radiologists read chest X-rays
Demand for chest X-rays has outpaced the number of radiologists available to read them. Most AI models that create reports look only at the image, but radiologists also consider why patients had the exam, their medical history and earlier exams. Dufan Wu, PhD, associate professor of radiology in The Ohio State University College of Medicine, will use Meta AI models to draft chest X-ray reports based on all this information. Wu will check the AI generated reports to make sure they are reliable. Wu and Peter Kvam, PhD, associate professor in the Department of Psychology at The Ohio State University, will also study how radiologists use AI drafts in a simulated clinical setting. The goal is to figure out when radiologists reject correct AI information and when they accept incorrect information. Knowing this could help radiologists use AI tools safely and efficiently and help researchers develop reliable AI tools.
Identifying heart failure risk earlier
Heart failure with preserved ejection fraction (HFpEF), which accounts for about half of all heart failure cases, can be difficult to recognize early because its symptoms overlap with those of obesity and other common conditions. Abdominal CT scans taken for unrelated medical reasons may already contain information about a patient's future cardiovascular risk, including patterns of body fat distribution, liver fat, muscle composition and calcium buildup in blood vessels.
Ohio State researchers will use Meta's latest AI models to investigate whether routine abdominal CT scans, combined with medical records and ECG data when available, can help predict future HFpEF risk. Because the approach uses information already collected during routine care, patients would not need additional imaging or tests for the initial assessment. The model will also provide a summary of the findings supporting each risk estimate and be designed to work when some patient information is unavailable.
The project is led by Yuan Xue, PhD, an AI researcher and assistant professor at The Ohio State University, Yuchi Han, MD, a cardiovascular disease specialist at The Ohio State University Wexner Medical Center, professor of Internal Medicine and Chair of Excellence in Cardiovascular Medicine, and Xia Ning, PhD, a professor at The Ohio State University, holding appointments in the Colleges of Engineering, Medicine and Pharmacy.
Clinician-guided agents for Alzheimer's diagnosis
Alzheimer's disease and related dementias affect an estimated 7.4 million Americans age 65 and older, yet diagnosis often follows years of decline. The diagnostic process happens step by step: a clinician searches the record, weighs competing hypotheses, chooses the next most informative test and revises. Electronic health records are not built to support this process because they can't easily reconstruct how a patient's cognition has changed over time or suggest which assessment would most reduce uncertainty.
Ning, Xue and Douglas Scharre, MD, cognitive and behavioral neurologist and medical director of Ohio State's Center for Cognitive and Memory Disorders, are developing a clinician-in-the-loop multi-agent system built on Meta's open-weight Muse models as a locally deployable baseline. The multi-agent system will reconstruct patients' cognitive histories, reason over clinical evidence, identify diagnostic uncertainties and recommend the next most informative clinical actions, all while keeping clinicians at the center of decision-making. The work draws on curated records from Ohio State's Memory Disorders Clinic, which has over 5,600 patient visits a year.
