Zongwei Zhou, an assistant professor of computer science, has received a four-year, nearly $2.4 million R01 grant from the National Institutes of Health (NIH) to develop AI algorithms to detect three types of abdominal cancers on computed tomography (CT) scans, enabling earlier diagnosis and treatment. This is Zhou’s second NIH R01 grant in two years.
Cancers of the pancreas, hepatobiliary (liver/gallbladder) system, and upper gastrointestinal (GI) tract are among the deadliest in the U.S. Causing over 113,000 deaths each year, their lethality is in part due to the fact that there is no effective, widely adopted screening for these diseases, and as such are typically found too late to be cured.
However, CT imaging is already routine, with an estimated 30–40 million abdominal CTs performed annually in the U.S. during emergency and outpatient care. These scans create a practical opportunity for earlier cancer detection without requiring additional imaging visits.
“Our vision is to turn these routine scans into an early-warning safety net to help find small tumors sooner and reduce unnecessary invasive procedures by improving diagnostic precision, flagging only those patients who truly warrant biopsy or surgery based on confirmatory imaging and risk factors,” explains Zhou, who serves as the contact principal investigator on the project.
The project team also includes multiple PIs Curtis P. Langlotz and Akshay Chaudhari—both faculty in Stanford University’s Departments of Radiology and Biomedical Data Science—as well as Johns Hopkins Bloomberg Distinguished Professor of Computational Cognitive Science Alan Yuille. Together they will develop and validate generalizable, efficient AI algorithms capable of detecting very small cancers in these abdominal areas from two types of routine CT scans: indicated scans—ordered for symptoms, abnormal labs, or known high risk—and opportunistic scans, acquired for other, unrelated reasons like trauma evaluation, kidney stone assessment, or vascular screening. The researchers’ goal is to achieve high sensitivity and specificity to minimize both missed detections and false alerts that could trigger unnecessary invasive procedures.
After evaluating performance and clinical impact through retrospective trials and studies, the team plans to test its algorithms on CT datasets curated from Johns Hopkins, Stanford, and the University of California, San Francisco, as well as public international collections with diverse patient demographics and imaging protocols.
This project’s success will move the medical AI field toward a more predictive and preventive model of care—one in which AI can identify at-risk individuals from routine CT data and prompt early, targeted screening.
“We envision a world where we can predict, well in advance, who is at high risk for pancreatic and related cancers, enabling personalized screening and timely early detection before symptoms appear,” Zhou says. “This project represents the first major step toward realizing that vision by transforming routine CT imaging into a proactive, population-scale early-detection tool, with the potential to extend its benefits of earlier intervention, fewer missed cancers, and fewer unnecessary procedures to other types of cancer.”
Additional co-investigators on this project include Kai Ding and Heng Li, both faculty in the Department of Radiation Oncology and Molecular Radiation Sciences at the Johns Hopkins University School of Medicine, and Yang Yang and Kang Wang, both faculty in the Department of Radiology and Biomedical Imaging at the UCSF School of Medicine.