Japanese

Artificial Intelligence Using A Convolutional Neural Network for Diagnosis of Early Gastric Cancer Based on Magnifying Endoscopy with Narrow Band Imaging Hiroya Ueyama 1 , Yusuke Kato 2 , Atsushi Ikeda 1 , Noboru Yatagai 1 , Hiroyuki Komori 1 , Yoichi Akazawa 1 , Tsutomu Takeda 1 , Kohei Matsumoto 1 , Kumiko Ueda 1 , Kenshi Matsumoto 1 , Daisuke Asaoka 1 , Mariko Hojo 1 , Takashi Yao 3 , Tomohiro Tada 2,4 , Akihito Nagahara 1 1Department of Gastroenterology, Juntendo University, School of Medicine, Tokyo 2AI Medical Service Inc., Tokyo 3Department of Human Pathology, Juntendo University Graduate School of Medicine, Tokyo 4Tada Tomohiro Institute of Gastroenterology and Proctology, Saitama, Japan Keyword: 人工知能 , AI , 胃癌 , 拡大内視鏡 , NBI pp.433-442
Published Date 2021/4/25
DOI https://doi.org/10.11477/mf.1403202299
  • Abstract
  • Look Inside
  • Reference

 We constructed an AI(artificial intelligence)-assisted CNN(convolutional neural network)CAD(computer-aided diagnosis)system, based on ME-NBI(magnifying endoscopy with narrow-band imaging)images to enable the diagnosis of EGC(early gastric cancer)and evaluated the diagnostic accuracy of this system. The AI-assisted CNN-CAD system(ResNet50)was trained and validated on a dataset of 5,574 ME-NBI images(3,797 EGCs and 1,777 non-cancerous mucosa and lesions). For the evaluation of diagnostic accuracy, a separate test dataset of 2,300 ME-NBI images(1,430 EGCs and 870 non-cancerous mucosa and lesions)was assessed using this system. The overall accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of the AI-assisted CNN-CAD system were 98.7%, 98%, 100%, 100%, and 96.8%, respectively. All the misdiagnosed images of EGCs were of low quality or of superficially depressed and intestinal-type intramucosal cancers that were difficult to distinguish from gastritis, even by experienced endoscopists. The AI-assisted CNN-CAD system for ME-NBI diagnosis of EGC had a considerable diagnostic ability. This system may possess substantial potential for clinical application in the future that could facilitate ME-NBI diagnosis of EGC in actual clinical practice.


Copyright © 2021, Igaku-Shoin Ltd. All rights reserved.

基本情報

電子版ISSN 1882-1219 印刷版ISSN 0536-2180 医学書院

関連文献

もっと見る

文献を共有