Detecting polyps out of direct vision

The adenoma detection rate (ADR) correlates directly with the amount of mucosal surface inspected by the endoscopist. In many cases, considerable areas of the mucosa remain unexplored, due to the limited angle of view of the cameras used in current endoscopes. The aim of this project was therefore to use different approaches to increase the amount of surface inspected.

One approach involves incorporating rear-viewing cameras that can recognize polyps using artificial intelligence (AI). A simple example may clarify this: vehicle drivers can’t look forward and into both side mirrors at the same time. Commercial systems have therefore been developed in the automotive sector that show a symbol on the side mirror during driving to alert the driver to vehicles passing at the side. Theoretically, this assistant function is fully transferable to colonoscopy. The examiner can then focus on the conventional image looking “forward,” while an AI system analyzes the “rear view” for adenomas and presents the appropriate warning signals.

Another approach involves inspecting the mucosa using different types of camera that operate in the nonvisible spectrum — that is, with light frequencies that the examiner cannot see. We believe that this new approach could similarly help increase the ADR and consequently reduce the risk of colon cancer.

Research for this project is funded by the Interdisciplinary Center for Clinical Research (Interdisziplinäres Zentrum für Klinische Forschung, IZKF) and by the “Forschung hilft!”(Research Helps!) foundation.

The project has been developed in collaboration with Adrian Krenzer and Prof. Dr. Frank Puppe, Head of Chair for Artificial Intelligence and Applied Informatics, Julius Maximilians University of Würzburg.