Researchers at Seoul National University Hospital Gangnam Center have developed an artificial intelligence (AI) system that can pick out adenomas in colon polyps by analyzing endoscopic images instead of conducting a biopsy.
An adenoma is a type of tumor that may affect various organs, such as the mucosa of the stomach, small intestine, and colon. While benign and non-cancerous, adenomas can be turned into adenocarcinomas, which are malignant or cancerous. There are many types of colon polyps found during a colonoscopy, and it is essential to find and remove adenoma that can develop into cancer.
Until now, hospitals have confirmed adenomas through tissue biopsies of polyps removed during the endoscopy process.
|From left, Professors Jin Eun-hyo, Kim Joo-Sung, and Kim Hee-chan|
To simplify the method, the research team, led by Professors Jin Eun-hyo, Kim Joo-Sung, and Kim Hee-chan at the hospital, combined 2,150 images of polyps found in colonoscopy in their AI platform. As a result, the AI platform was able to identify adenomas with an accuracy of 86.7 percent.
The team divided 22 endoscopic doctors into two groups – one group using the AI platform and the other not using it – and compared their findings of adenoma accurately.
It found that doctors could increase their reading accuracy from 82.5 to 88.5 percent by using the platform. Notably, the system significantly improved the accuracy of novice doctors with little experience in endoscopy by 11.8 percent.
“AI-based diagnostic assistance systems will soon be widely adopted in the field of endoscopy,” Professor Jin said. “This is the first study to confirm the possibility of using AI as a diagnostic aid in actual clinical practice.”
The journal Gastroenterology has published the results of the study.