Identification of Upper Gastrointestinal Diseases during Screening Gastroscopy through Deep Convolutional Neural Network Algorithm
Hang Yang,Yu Wu,Bo Yang,Min Wu,Jun Zhou,Qin Liu,Yifei Lin,Shilin Li,Xue Li,Jie Zhang,Rui Wang,Qianrong Xie,Jingqi Li,Yue Luo,Mengjie Tu,Xiao Wang,Haitao Lan,Xuesong Bai,Huaping Wu,Fanwei Zeng,Hong Zhao,Zhang Yi,Fanxin Zeng
DOI: https://doi.org/10.1016/j.gie.2022.06.011
IF: 10.396
2022-06-18
Gastrointestinal Endoscopy
Abstract:Background and Aims The clinical application of gastrointestinal endoscopy for the diagnosis of multiple diseases using artificial intelligence (AI) has been limited by its high false positive rates. There is an unmet need to develop a Gastrointestinal Endoscopy AI-assisted Diagnosis System (GEADS) to improve diagnostic accuracy and clinical utility. Methods In this retrospective, multicenter study, a convolutional neural network was trained to assess upper gastrointestinal diseases based on 26,228 endoscopic images from Dazhou Central Hospital that were randomly assigned (3:1:1) to the training dataset, validation dataset, and test dataset, respectively. To validate the model, 6 external independent datasets comprising 51,372 images of upper gastrointestinal diseases were collected. Besides, one prospective dataset comprising 27,975 images were collected. The performance of GEADS was compared with endoscopists with two professional degrees of expertise: expert, novice. 8 endoscopists in the expert group with more than 5 years of experience, 3 endoscopists in the novice group with 1 to 5 years of experience. Results The GEADS model achieved an accuracy of 0.918 (95% confidence interval [CI]: 0.914-0.922), with an F1 score of 0.884 (95% CI: 0.879-0.889), recall of 0.873 (95% CI: 0.868-0.878) and a precision of 0.890 (95% CI: 0.885-0.895) in the internal validation dataset. In the external validation datasets and one prospective validation dataset, the diagnostic accuracy of GEADS ranged from 0.841 (95% CI: 0.834-0.848) to 0.949 (95% CI: 0.935-0.963). With the help of GEADS, the diagnosing accuracies of novice and expert endoscopists were significantly improved ( P < 0.001). Conclusions The AI system can assist endoscopists in improving the accuracy of diagnosing upper gastrointestinal diseases.
gastroenterology & hepatology