Deep learning system compared with expert endoscopists in predicting early gastric cancer and its invasion depth and differentiation status (with videos)

Current Status and Future Perspective of Artificial Intelligence in the Management of Peptic Ulcer Bleeding: A Review of Recent Literature

“…As endoscopic biopsy may not performed at the index endoscopy, follow-up endoscopy for these patients is essential because the treatment of benign or malignant gastric ulcers is entirely different [33]. Artificial intelligence systems are being studied for differentiating benign vs. malignant gastric ulcers [34][35][36] and early vs. advanced gastric cancer [35] or for improving endoscopic depth prediction of early cancer [37,38]. Cho et al [35] developed a convolutional neural network (CNN)-based model to automatically classify gastric neoplasms from endoscopic images.…”

“…The overall accuracy was significantly improved from 45.9% to 95.9% [36]. Wu et al [38] developed a deep learning-based AI system covering early gastric cancer diagnosis, cancer invasion depth prediction, and differentiation status. A human-machine competition involving an AI system and 46 expert endoscopists from 19 provinces in China was performed.…”

Artificial Intelligence and Deep Learning for Upper Gastrointestinal Neoplasia

“…In addition, the same system showed sensitivity rates for detecting gastric neoplasia and diagnosing early gastric cancers (EGCs) of 87.81% and 100%, respectively, significantly higher than those of endoscopists (83.51% and 87.13%, respectively). 67 Accuracy rates of the system for predicting EGC invasion depth and differentiation status were comparable to those of the endoscopists. 67 When passing from preclinical to clinical validation studies, not only bias related to deep AI validation should be considered, because other types of bias purely related to the blinding of the operator and the randomization process may occur.…”

“…67 Accuracy rates of the system for predicting EGC invasion depth and differentiation status were comparable to those of the endoscopists. 67 When passing from preclinical to clinical validation studies, not only bias related to deep AI validation should be considered, because other types of bias purely related to the blinding of the operator and the randomization process may occur.…”

“…To date, one system (EndoAngel; Wuhan University, Wuhan, China) has been extensively validated with a more rigorous methodologic approach, for which 4 published clinical trials are currently available. 64,[66][67][68] Briefly, in these various clinical trials, this system has been shown to increase the inspection time of the gastric mucosa (5.40 minutes [SD, 3.82] vs 4.38 minutes [SD, 3.91]; P < .001), to reduce the percentage of blind spots during upper GI endoscopy (21% [95% CI, 1.6-40] vs 38.9% [95% CI, 0.8-68.3]; P < .001), and to reduce the miss rate 68 in a tandem methodology design (6.1% [95% CI, 1.6-17.9] vs 27.3% [95% CI, 15.5-43.0]; relative risk, 0.224 [95% CI, 0.068-0.744]; P ¼ 0.015). In addition, the same system showed sensitivity rates for detecting gastric neoplasia and diagnosing early gastric cancers (EGCs) of 87.81% and 100%, respectively, significantly higher than those of endoscopists (83.51% and 87.13%, respectively).…”