KIMA - Korea International Medical Association


Seoul National University Hospital Gangnam Center, AI can screen colon adenoma without biopsy

August 31,2020

South Korean researchers have developed a system for diagnosing colon polyps by analyzing endoscopic pictures with artificial intelligence (AI). Professor Eun-Hyo Jin of the Seoul National University Hospital Gangnam Center, Professor Joo-sung Kim of Seoul National University Hospital Gastroenterology Department, and a joint research team led by Hee-chan Kim of the Medical engineering department (photo 1) announced in June the results of a study incorporating AI based on the images of 2,150 polyps found in the colonoscopy that conducted a health examination. The research team identified the adenoma with an accuracy of 86.7% as a result of putting the polyp image in the trained AI system. In addition, 22 endoscopic doctors were divided into groups of beginners, endoscopy specialists, and endoscopy specialists who received optical diagnostic training according to their proficiency, and compared the group using the AI system with the group not using it. Assisting the AI system, the reading accuracy increased from 82.5% to 88.5%. In particular, 11.8% accuracy was higher for the beginner doctors who did not have much experience in endoscopy. There are many types of polyps found by colonoscopy. Among them, it is very important to find and remove adenoma, a polyp that can progress to cancer. Currently, diagnosis of adenoma was possible only through tissue biopsy of polyps removed during endoscopy. The artificial intelligence system developed this time can detect and diagnose adenoma with only an endoscopic image. Accordingly, the research team said that it is expected to reduce the time and cost required for pathology examination and reduce unnecessary tissue biopsy. Professor Eun-Hyo Jin said, “AI-based diagnostic systems will be widely introduced in the endoscopy field sooner or later,' and added, 'It is the first study to confirm the possibility of diagnosis through AI as an auxiliary diagnostic method in actual clinical trials as it helps with reading accuracy.' The results of this study were published in the latest issue of Gastroenterology, a global journal, and counted among the first in the field of digestive organs.