Post written by Hisatomo Ikehara, MD, PhD, from the Division of Gastroenterology and Hepatology, Department of Medicine, Nihon University School of Medicine, Tokyo, Japan.

We have developed an e-learning system to learn the magnifying endoscopic findings of early gastric cancer that can be taken on the web. And we have already reported the effectiveness of this e-learning system. However, e-learning for types of gastric lesions were unclear, so we decided to do additional analysis on this point.

Based on the results of this study, a high learning effect was obtained for depressed lesions lesions under 10 mm. However, with the elevated type and flat type, the learning was poor, and the learning effect was poor even with lesions of 10 mm or more. In addition, the results of this study showed that AUC significantly increased when the combination of demarcation line (DL), microvascular pattern (MV), and microsurface pattern (MS) was evaluated. AUC of MS is higher than DL and MV.

This result clarified the point for improving the e-learning system. In addition, when the gastric lesion was positive for WOS, the accuracy was very low. However, it is interesting that learning has shown that accuracy is significantly increased by learning WOS through e-learning. On the other hand, the examination result of kappa value showed that the agreement rate of each participant’s answer was low. Currently, AI diagnosis of endoscopic images using deep learning is developing rapidly. We look forward to the possibility that endoscopy with AI will be a gospel for endoscopists.

Read the full article online.

The information presented in Endoscopedia reflects the opinions of the authors and does not represent the position of the American Society for Gastrointestinal Endoscopy (ASGE). ASGE expressly disclaims any warranties or guarantees, expressed or implied, and is not liable for damages of any kind in connection with the material, information, or procedures set forth.