Automatic and Diagnosis-supporting Detection of Polyps in Colonoscopic Image Sequences

Colon cancer is the second leading cause of cancer death. Around 26,000 people die in Germany each year as a result. In 90 percent of these cases, special colon polyps serve as precursors to cancer. For this reason, periodic colonoscopy-based cancer screening for people 55 years and older is now covered under Germany´s statutory health insurance program. During a colonoscopy, the large intestine and rectum are examined with an endoscope. The effectiveness of this procedure depends heavily on the experience and attentiveness of the physician. Various studies show that 12 to 24 percent of the polyps go undetected, a rate that could be considerably lowered by instituting automatic polyp detection with image-generating processes. The real challenge is identifying small polyps and new formations that are flat and still under the surface.

As part of its own basic research, the Fraunhofer Institute for Integrated Circuits IIS developed an analysis process in which suspicious tissue areas with pigment and texture deviations are highlighted and automatically detected. This development laid the foundations for a method that supports the physician by automatically visualizing suspicious areas and documenting detected polyps. A further step in the “KoloPol“ project (german) will involve testing the functionality of such image detection methods under actual clinical conditions, including fine-tuning the ability to identify smaller and flatter polyps. With this software researchers hope to not only increase the polyp detection rate, but also reduce the time and effort associated with colon cancer screening.