TLDR: Northwestern University researchers have developed a new generative AI radiology tool that significantly enhances the analysis of X-rays and CT scans. This in-house developed system, which requires far less data for training, can generate highly complete and personalized reports, and is capable of flagging life-threatening conditions in real-time. The tool aims to alleviate radiologist backlogs and accelerate diagnoses, particularly in critical scenarios, serving as an augmentation to human expertise rather than a replacement.
A pioneering generative AI system, developed by a team of radiologists and engineers at Northwestern Medicine, is set to revolutionize the field of radiology by dramatically improving productivity and the early identification of critical medical conditions. This innovative tool, unlike large internet-trained models such as ChatGPT, was built from scratch using clinical data from within the Northwestern Medicine network, resulting in a lightweight, nimble, and highly specialized AI model for radiology.
The system has undergone real-world deployment across the 11-hospital Northwestern Medicine network, where it analyzed nearly 24,000 radiology reports over a five-month period in 2024. The study, published in JAMA Network Open, demonstrated an average 15.5% boost in radiograph report completion efficiency, with some radiologists experiencing gains as high as 40%, all without compromising accuracy. Unofficial follow-up work indicates potential efficiency gains of up to 80% and the tool’s applicability to CT scans.
Dr. Mozziyar Etemadi, a senior author on the study and an assistant professor of anesthesiology at Northwestern University Feinberg School of Medicine and of biomedical engineering at Northwestern’s McCormick School of Engineering, stated, “This is, to my knowledge, the first use of AI that demonstrably improves productivity, especially in healthcare. Even in other fields, I haven’t seen anything close to a 40% boost.”
The Northwestern AI system analyzes entire X-rays or CT scans and generates reports that are 95% complete and personalized to each patient, mirroring the radiologist’s own reporting style. These reports summarize key findings and provide a template for radiologists to review and finalize, significantly reducing documentation time. During the study period, the AI system saved over 63 hours of documentation time, equivalent to reducing coverage needs from 79 to 67 radiologist shifts.
Dr. Samir Abboud, co-author and chief of emergency radiology at Northwestern Medicine, emphasized the tool’s impact on critical cases: “For me and my colleagues, it’s not an exaggeration to say that it doubled our efficiency. It’s such a tremendous advantage and force multiplier. On any given day in the ER, we might have 100 images to review, and we don’t know which one holds a diagnosis that could save a life. This technology helps us triage faster — so we catch the most urgent cases sooner and get patients to treatment quicker.”
The development team stressed that the in-house approach allowed them to create a system that is faster, more accurate, and requires significantly less computing power than commercial alternatives. Dr. Jonathan Huang, first author and a third-year medical student at Feinberg with a Ph.D. in biomedical engineering, noted, “Our study shows that building custom AI models is well within reach of a typical health system, without reliance on expensive and opaque third-party tools like ChatGPT.”
Also Read:
- Generative AI Models Show Variable Performance in Pulmonary CT Lung Cancer Diagnosis
- Nebraska Medicine Leverages AI to Streamline Patient Communications, Reducing Human Intervention in Calls
Two patents have already been approved for the Northwestern Medicine technology, with others in various stages of the approval process. The tool is currently in the early stages of commercialization, promising a significant step forward in addressing the projected shortage of radiologists in the U.S., which is anticipated to reach up to 42,000 by 2033.
