Sept. 29, 2026 — A recently published study shows that using artificial intelligence to track which pathologies radiology residents saw each day made it possible to identify gaps in their clinical exposure and provide supplemental, targeted teaching cases. This approach helped ensure that residents were exposed to the full breadth of pathology needed for training, the study authors said.
Led by NYU Langone Health researchers, the study addressed a longstanding challenge in radiology education: residents traditionally learn to diagnose disease based on the real patient cases they encounter during their assigned clinical workdays, which may not cover the full range of important conditions. By monitoring daily case exposure with AI and adding targeted teaching cases, the researchers improved the breadth of pathology seen by residents without reducing their experience with real patient cases.
Published online recently in Academic Radiology, the team’s work found that the AI could identify disease exposure gaps for radiology residents and suggest the specific patient case pathologies they need to see with more than 90 percent accuracy.
“If a resident sees 30 cases in a day, 29 will be routine cases like normal exams or common pathology such as fractures, and maybe one patient will have a less common condition, such as a rare form of autoimmune arthritis,” said study author Vinay Prabhu, MD, MS, an associate professor in the Department of Radiology at NYU Grossman School of Medicine. “As it stands, doctors in training in radiology and other specialties are likely not seeing the breadth of case types that they will encounter throughout the rest of their careers.”
“This represents a fundamental shift in how we train radiologists — moving from a one-size-fits-all model to one that automatically addresses each resident's specific learning needs," said Michael P. Recht, MD, chair of the Department of Radiology at NYU Langone. “No other system provides this level of personalized, data-driven training where residents work, and we think this tool will be much sought after in the field.”
Research Study
Prior to the study, the research team built a curriculum listing important conditions that radiology residents should see during their first three years of training. Faculty experts in five imaging specialties — abdominal, musculoskeletal, brain, pediatric, and chest imaging — identified these conditions based on preparation materials for exams that residents must pass to practice as board-certified radiologists. They also set target numbers for how many times residents should encounter each condition, as well as rankings of what pathologies were more or less important to see.
The study authors then used a chatbot, ChatGPT-4o, to read the summary sections of residents' daily clinical reports and select three to five teaching cases each night for each resident, prioritizing conditions they had seldom seen. The AI-picked cases appeared on residents' workstations alongside their regular clinical work, and they discussed findings with supervising physicians — mimicking real clinical practice.
Current strategies to address exposure gaps can be cumbersome for a training organization, typically including lecture-based courses, faculty-shared teaching files, and self-directed supplemental learning (e.g., textbooks, question banks, videos), the authors said. These methods can also vary between trainees and programs and often fail to capture real clinical scenarios.
"Although existing metrics help ensure that residents review a sufficient number of cases during residency, they do not reliably measure whether trainees are exposed to an adequate variety of pathologies,” said study co-author Matthew G. Young, DO, also an associate professor in the Department of Radiology. “Our precision education tool is an important proof of concept for an automated pathology tracker that could meaningfully enhance radiology training. Looking ahead, we hope to expand the tool to additional subspecialties, including cardiac imaging, nuclear medicine, breast imaging, and interventional radiology."
The study was funded in part by a grant from the Committee of Interns & Residents/Service Employees International Union Healthcare (CIR/SEIU) Patient Care Trust Fund (#3182).
Along with Drs. Prabhu, Young, and Recht, study authors from the Department of Radiology at NYU Langone were Antonio Verdone, Erin Alaia, Anna Chen, Luoyao Chen, Sumit Chopra, Ryan Cummings, Jay Karajgikar, Jane Ko, Renata La Rocca Vieira, Shailee Lala, Evan Stein, Naomi Strubel, Danielle Toussie, and William Walter. A study author from Visage Imaging in Berlin, Germany, was Malte Westerhoff.
October 23, 2026 
