Artificial intelligence in the diagnosis of gastric precancerous conditions by image-enhanced endoscopy

Post written by Honggang Yu, MD, from the Department of Gastroenterology, the Key Laboratory of Hubei Province for Digestive System Disease, and the Hubei Provincial Clinical Research Center for Digestive Disease Minimally Invasive Incision, Renmin Hospital of Wuhan University, Wuhan, China.


In this study, we conducted a deep convolutional neural network (DCNN) system to detect gastric precancerous conditions by image-enhanced endoscopy.

Gastric atrophy (GA) and intestinal metaplasia (IM) play an important role in the development of gastric cancer and constitute the background of dysplasia and adenocarcinoma. Thus, diagnosing gastric precancerous conditions and regular follow-up had great potential for protecting patients from gastric cancer.

In this study, we combined AI and medicine to construct a DCNN system using diversified endoscopic images from 5 different hospitals that was able to achieve high diagnostic accuracy and PPV for detecting precancerous conditions. The performance of the proposed system is similar to that of experts and superior to that of non-experts. However, our study only focuses on detecting precancerous conditions in image-enhanced endoscopy, but the endoscopists evaluated the gastric mucosa first with WLE and subsequently with NBI/BLI in clinic. So, we will combine both techniques simultaneously in our future study. Otherwise, the prospective video test set was only obtained from Renmin Hospital of Wuhan University. Therefore, we plan to conduct multicenter prospective research and enroll more patients to further validate the DCNN system. Last but not least, low-grade dysplasia, an important premalignant lesion, is not taken into consideration in this study. The lesion pathologically confirmed as low-grade dysplasia was too small to develop an image classification model in image-enhanced endoscopy. Thus, a potential method for detecting low-grade dysplasia will be investigated in our future research.

Our DCNN system shows great potential in assisting with gastric precancerous conditions diagnosis in image-enhanced endoscopy and reducing the necessity to perform multiple biopsies in real clinical settings.


Figure 1. Framework of the computer-aided detection–based system. Videos were clipped into images and then inputted to the deep learning models. First, a deep convolutional neutral network model was used to recognize clear images. Then, in WLE, the stomach was classified into 26 parts, and the part observed would be recognized by diagnosis models (such as the anterior wall of antrum). In ME-NBI/BLI, gastric precancerous conditions would be detected by models (such as gastric atrophy with intestinal metaplasia). WLE, White-light endoscopy; A, anterior wall; ME-NBI/BLI, magnifying endoscopy with narrow-band imaging/blue laser imaging.

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.

Leave a Comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s