Downstate Researchers Pave the Way for AI in Identifying Stroke Lesions

By Office of the President | May 23, 2024

SUNY Downstate researchers, led by Neurology Resident Jung-Hyun Lee, M.D., and Professor of Physiology and Pharmacology William W. Lytton, M.D., are investigating the role of artificial intelligence (AI), particularly the GPT-4 model, in identifying brain lesions resulting from strokes. This research is crucial given the severe consequences of strokes, including long-term disabilities or death, which stem from interruptions in brain blood flow. Locating these lesions post-stroke is essential for predicting potential long-term effects and determining optimal treatment strategies. However, access to conventional diagnostic tools like brain scans and neurological exams is often limited, especially in underserved urban or rural areas.

The study explores whether AI, specifically GPT-4, could serve as a viable alternative for lesion localization when traditional methods are unavailable. Analyzing texts from health records and neurological exams of 46 stroke cases, researchers tasked GPT-4 with identifying lesion presence, location, and details. While GPT-4 demonstrated promising sensitivity and specificity in identifying lesion location and side of the brain, overall accuracy remained at 41 percent, indicating that further refinement is necessary before clinical implementation. Nonetheless, this research underscores AI’s potential in addressing healthcare disparities, particularly in global health contexts where access to specialized neurological care is limited, given GPT-4’s language versatility.

I extend my congratulations to Drs. Lee and Lytton for their insightful exploration into the potential of AI. The study, published in the March 27, 2024, online issue of Neurology Clinical Practice, an official journal of the American Academy of Neurology, can be accessed here and highlights the promising role of AI in addressing healthcare disparities, especially in regions with limited access to specialized care.

Kudos on this vital contribution to advancing medical research!