AI-based detection of upper gastrointestinal adenomas in FAP patients

Familial adenomatous polyposis (FAP) is a condition marked by the development of hundreds of colon adenomas, starting in childhood and affecting 1 in 10,000 people in Europe. Without treatment, all FAP patients face a 100% risk of colon cancer, typically by ages 35–45. Up to 70% also develop duodenal adenomas, with a 4–10% risk of progressing to duodenal carcinoma based on adenoma size, growth, and dysplasia grade. Current cancer prevention guidelines recommend endoscopic removal of duodenal adenomas over 10 mm, but subjective assessments often lead to inaccuracies. Gastric adenomas occur in 10% of cases, posing a cancer risk if high-grade dysplasia is present.
AI-assisted detection offers promising improvements, particularly in distinguishing neoplastic changes from benign glandular cysts in the stomach. This could prevent overtreatment and associated complications while improving early cancer prevention. In this project we aim to develop AI algorithms for real-time detection and measurement of gastroduodenal adenomas. These include AI-based segmentation and size determination. The project will be performed in close collaboration with the Department of Gastroenterology and Hepatology, Heidelberg University Hospital, Germany.
It is funded by the German Cancer Aid (Deutsche Krebshilfe).