Japanese
English
- 有料閲覧
- Abstract 文献概要
- 1ページ目 Look Inside
- 参考文献 Reference
- サイト内被引用 Cited by
要旨●H. pylori感染胃炎の画像診断は,胃癌リスクの層別化に有効とされる.これを踏まえて本稿では,筆者らが研究中のコンピュータ画像認識技術を応用したH. pylori感染胃炎に対する人工知能診断について述べる.画像強調内視鏡を用いたコンピュータ支援診断(LCI-CAD)の正診率(accuracy)は,H. pylori未感染84.1%,現感染81.7%,除菌後78.6%であった.この結果は,同じ手法で作成した白色光のCAD(WLI-CAD)よりよい成績であり,加えて内視鏡専門医と同等の診断精度と考えられた.一方で,人工知能を用いた胃X線二重造影像によるH. pylori未感染と現感染の2分類では,感度86.7%,特異度91.7%の結果が得られた.人工知能を用いたH. pylori診断は,常に一定の診断精度が得られるだけでなく,診断速度も速い.さらに,プログラムを複製することも可能である.筆者らは本研究が胃癌リスクの層別化に応用されれば,早期胃癌の内視鏡スクリーニングやX線検診の診断支援へ貢献できるものと期待している.
*本論文中、QRコードを読み込む,もしくはURLにアクセスいただくことで動画を再生できます(公開期限:2024年4月).
The diagnosis of H. pylori(Helicobacter pylori)infection by clinical imaging can be effective in stratifying the risk of gastric cancer. In this article, we describe our results of AI(artificial intelligence)using a computer image recognition technology for the diagnosis of H. pylori infection. The accuracy of LCI-CAD(computer-aided diagnosis using linked-color imaging)was 84.1% in uninfected individuals, 81.7% in currently infected patients, and 78.6% in post-eradication cases. The diagnostic accuracy based on the LCI-CAD data was superior to that based on the white light imaging CAD data in the uninfected, currently infected, and post-eradication cases. In addition, the diagnostic accuracy of LCI-CAD was comparable with those of experienced endoscopists. Furthermore, in this study, a CAD system was created using gastric X-ray double-contrast images. The X-ray CAD system showed a sensitivity of 86.7% and specificity of 91.7%. The use of AI for the diagnosis of H. pylori is associated with both a superior level of diagnostic accuracy and high diagnostic speed. In addition, it is possible to duplicate the program. Hence, we believe that the application of this method for the stratification of gastric cancer risk will provide diagnostic support for endoscopic and X-ray screening programs in patients with early gastric cancer.
Copyright © 2021, Igaku-Shoin Ltd. All rights reserved.