AI Model Detects Hidden Disease Signals in Routine CT Scans

Computed tomography (CT) scans are often performed to answer a specific diagnostic question, which can leave other potentially important findings outside the immediate clinical focus. As a result, opportunities to recognize chronic disease or future health risks may be missed. New findings now show that an artificial intelligence model can extract broader health information from routine CT imaging.

The model, developed at Penn Medicine, is called Percival and was designed to analyze CT scans for disease-related patterns and clinical outcomes beyond the original reason imaging was performed. Unlike models that must be directed toward a specific condition or endpoint, Percival can operate in the background to detect potentially relevant findings that might otherwise go unrecognized.

Percival was trained using data from the Penn Medicine BioBank, which contains information from more than 350,000 people collected across the University of Pennsylvania Health System over two decades. Researchers used approximately 400,000 CT scans from about 50,000 patients to refine the model, which remains in preclinical testing.

In performance analyses, the Penn Medicine team found that routine CT scans contained clinically useful information that was not immediately apparent. Percival identified signs of chronic conditions including atrial fibrillation and chronic kidney disease. Compared with the AI model TotalSegmentator, it also more reliably flagged conditions including pancreatic cancer, cirrhosis, and bacterial pneumonia.

The study was published in NPJ Digital Medicine on September 29, 2026. The team is planning to test Percival early next year with the goal of moving the model toward regular clinical use. Developers are also examining whether larger datasets and models spanning multiple institutions could improve Percival’s robustness and predictive performance.

“Our teams have developed something that learns broadly from imaging data and uncovers connections between scans, lab values, people’s physiology, and clinical outcomes to potentially give more accurate assessments of health, or even ring alarm bells early. This is work that points toward a future in which medical imaging becomes a foundation for more personalized and proactive healthcare,” said Walter Witschey, PhD, senior author, associate professor of Radiology, and associate vice chair of Research IT.

“We’re hoping to elevate ‘opportunistic screening,’ where results are found beyond the original purpose of a scan, such as finding the early traces of cancer on a scan that was maybe assessing pneumonia. We believe the totality of what’s in a CT scan can be made much more useful with help from a model like this,” added Witschey.

Related Links
Penn Medicine