NCI is still seeking applications (until May 2023) that develop artificial intelligence (AI)-based models using secondary data to improve the early detection and risk assessment of abdominal cancers. NCI is particularly interested in applications that feature collaborative AI-development methods, such as federated learning. This approach would make the model and code widely available without needing to access the original data.

Learn more about this notice on the NIH Grants & Funding website.

Dr. Lalitha Shankar, clinical trials branch chief of the NCI Division of Cancer Treatment and Diagnosis’s Cancer Imaging Program, says, “The potential of federated learning is very significant. It could enable development and testing of algorithms using radiology images in their clinical context from large clinical databases in healthcare systems.”

AI could augment cancer screening and holds great potential for the early diagnosis of a wide range of abdominal cancers, such as pancreatic, liver, and kidney cancers. Unfortunately, very little data are available for training AI models.

To overcome this sample-size challenge, researchers are increasingly turning to existing data sources to build cohorts for training AI models. For example, each year Americans undergo 22 million abdominal CT scans, often for reasons not directly related to cancer. Information from these scans could help train an AI model to identify at-risk patients before they show signs of disease, boosting the odds for successful cancer prevention and treatment.