Penn Medicine Researchers Develop AI to Detect Hidden Health Issues in CT Scans
An artificial intelligence model named Percival, developed by Penn Medicine, can analyze CT scans to proactively identify chronic diseases and future health risks.
Penn Medicine researchers have developed a new artificial intelligence model named Percival, which analyzes data from computed tomography (CT) scans to uncover hidden health issues beyond the original diagnosis. Published in the journal NPJ Digital Medicine, the study demonstrates that the model can detect chronic diseases and potential health risks with details that might escape the human eye.
Trained on a Vast Dataset
Percival was trained using data from more than 350,000 individuals collected over two decades within the Penn Medicine BioBank. The researchers used approximately 400,000 CT scans from about 50,000 patients to enhance the model's capabilities.
Opportunistic Screening Capacity
The model can catch early signs of diseases like cancer even in scans taken for a different purpose, such as pneumonia. This 'opportunistic screening' method aims to make medical imaging a foundation for personalized and proactive healthcare.
Early Diagnosis of Diseases
Percival showed high success in detecting conditions such as atrial fibrillation, chronic kidney disease, pancreatic cancer, and cirrhosis. Experts state that this technology will enable cardiovascular and metabolic complications to be noticed at an early stage.
Future Plans and Open Source
The developers plan to begin testing processes next year to make the model ready for clinical use. Percival's code and technical specifications were shared as open source to support scientific collaboration.