Application of Convolutional Neural Networks in the Diagnosis of Helicobacter pylori Infection Based on Endoscopic Images

Abstract

Background and aims

The role of artificial intelligence in the diagnosis of Helicobacter pylori gastritis based on endoscopic images has not been evaluated. We constructed a convolutional neural network (CNN), and evaluated its ability to diagnose H. pylori infection.

Methods

A 22-layer, deep CNN was pre-trained and fine-tuned on a dataset of 32,208 images either positive or negative for H. pylori (first CNN). Another CNN was trained using images classified according to 8 anatomical locations (secondary CNN). A separate test data set (11,481 images from 397 patients) was evaluated by the CNN, and 23 endoscopists, independently.

Results

The sensitivity, specificity, accuracy, and diagnostic time were 81.9%, 83.4%, 83.1%, and 198 s, respectively, for the first CNN, and 88.9%, 87.4%, 87.7%, and 194 s, respectively, for the secondary CNN. These values for the 23 endoscopists were 79.0%, 83.2%, 82.4%, and 230 ± 65 min (85.2%, 89.3%, 88.6%, and 253 ± 92 min by 6 board-certified endoscopists), respectively. The secondary CNN had a significantly higher accuracy than endoscopists (by 5.3%; 95% CI, 0.3–10.2).

Conclusion

H. pylori gastritis could be diagnosed based on endoscopic images using CNN with higher accuracy and in a considerably shorter time compared to manual diagnosis by endoscopists.

Highlights

• •

  We compared the diagnostic ability for H. pylori gastritis between a convolutional neural network (CNN) and endoscopists.

• •

  Diagnostic ability of CNN was greater than that of endoscopists in general and similar to that of experienced endoscopists.

• •

  The diagnostic time of CNN was considerably shorter than that of manual diagnosis by endoscopists.

In Japan, H. pylori infection is common, and its detection during endoscopic examination is desirable. However, a diagnosis based on endoscopic findings requires training, and its accuracy depends on the endoscopist's skill. We showed that the diagnostic ability of a convolutional neural network (CNN) for H. pylori infection was comparable to that of experienced endoscopists, and the time required for the diagnosis was considerably shorter. Thus, the use of CNNs for diagnosis of H. pylori infection will reduce endoscopists' workload. Moreover, the procedure can be performed completely “online,” thereby addressing the problem of inadequate number of physicians in remote locations.