Tom Konikoff, MD
Project Title
Multimodal AI for Early Detection of Pancreatic Cancer in New-Onset Diabetes
About the Investigator
Dr. Tom Konikoff is a gastroenterologist and lecturer at Tel Aviv University Gray Faculty of Medical and Health Sciences whose research focuses on applying artificial intelligence to improve early detection of gastrointestinal cancers, particularly pancreatic cancer. He completed his postdoctoral research fellowship in early detection of pancreatic cancer at the Mayo Clinic, where he gained extensive experience in AI-driven analysis of large-scale clinical, imaging, and pathology data. He has led multiple grant-funded projects, established an AI-focused research group at Rabin Medical Center, and developed patented AI-based clinical decision-support tools for risk prediction and computer-aided diagnosis.
About the Research
Pancreatic cancer is one of the deadliest cancers because it is usually found late, when treatment options are limited. However, many patients develop problems with blood sugar, including new-onset diabetes, in the years before their cancer is diagnosed. This means that among the many people who develop diabetes each year, a small number are actually in an early, hidden stage of pancreatic cancer. The challenge is to find those few high-risk individuals without exposing everyone else to unnecessary, costly, and potentially harmful tests.
In this project, Dr. Konikoff and his team will use data from Clalit Health Services, Israel’s largest health organization, which follows millions of people over time and is a mirror to the population. They will build a large database of patients who developed new-onset diabetes and then apply artificial intelligence to look for patterns that predict who will develop pancreatic cancer within three years.
Unlike previous tools that rely only on basic clinical information, their models will be able to combine several types of data at once, including medical history, imaging scans, and doctors’ notes. The team will then test these models in a separate group of patients to ensure that they work reliably in real-world conditions.
If successful, this research will provide a practical tool that health systems can use to identify a smaller, high-risk group of people with new-onset diabetes who should be offered focused pancreatic cancer screening and, in the future, blood tests for very early detection.
