Colorectal cancer is the second most lethal cancer type in developed countries and more than 70% of the polyps, when present in one’s colon or rectum, can eventually change into tumor cells (mostly adenomas). Polyp’s visual detection during colonoscopy is the world’s gold standard in colorectal cancer diagnosis. However, today polyps detection rates during these preliminary visual inspections commonly fall below 80%, resulting in that almost 50% of diagnosed colorectal cancer were not detected during a first colonoscopy, and it is widely assumed that these results are significantly practicioner-dependant. Still, all existing solutions focus on contrast modification or vision angle widening for best inspection of the mucosa. Our software aims at significantly enhancing polyp’s detection rate by visually alerting the practitioner in real time during a colonoscopy procedure. The method is based on optimized reinforced activelearning algorithms for image to image analysis. It is adapted to real time video treatment (more than 25 fps analyzed) and aims at detection rates higher than 90% at the practitioner level.
Competitive advantages :
Keywords : Computer Aided Detection (CAD), Polyps, Cancerology, Colonoscopy, Endoscopy, Video analysis