Dr Sharma has been at the forefront and a pioneer in improving the diagnosis and management of GI diseases and cancer, specifically GERD, Barrett’s esophagus, advanced imaging, artificial intelligence and endoscopic treatments. He is a world-renowned physician, teacher, and educator. Dr. Sharma has over 400 publications, including original articles and book chapters, and is regularly invited to present at major national and international meetings. In addition, he has published several important textbooks: Barrett’s esophagus and esophageal cancer; Rise of acid reflux in Asia; Gastrointestinal Cancers. Dr. Sharma has and continues to publish in several high-profile national and international journals, including the New England Journal of Medicine, Annals of Internal Medicine, Gastrointestinal Endoscopy, Gastroenterology, and American Journal of Gastroenterology. He completed his internal medicine residency at the Medical College of Wisconsin in Milwaukee and his gastroenterology fellowship at the University of Arizona in Tucson. LESS … MORE
Evaluating artificial intelligence (AI) systems for finding polyps during colonoscopy is important. This study looked at two different AI systems (CADe A and CADe B) used in Humanitas Research Hospital to detect false-positive (FP) alerts, which are when the AI incorrectly signals a polyp. Researchers compared the performance of these systems using a previously developed way to classify FP alerts.
Method: The study analyzed 40 colonoscopy videos per system, noting the number of FPs, why they happened, and the time taken by doctors to check the areas marked incorrectly by the AI.
Results: Both CADe A and CADe B systems showed similar results, with around 25 FPs per colonoscopy. Most FPs were due to bowel wall artifacts, about 89-86%, and around 10-14% were because of bowel content. Each false alert took about 0.2 seconds for CADe A and CADe B, and a small percentage needed extra time for closer examination by the doctors.
Conclusions: Using a standard way to label FP alerts, both AI systems performed similarly in finding false polyp signals during colonoscopy. This suggests that these systems can be reliable in clinical settings.
Read full research: Comparing the number and relevance of false activations between 2 artificial intelligence computer-aided detection systems: the NOISE study
