Department of Gastroenterology, Graduate School of Medicine, University of Tokyo, Tokyo, Japan.
Division of Next-Generation Endoscopic Computer Vision, Graduate School of Medicine, University of Tokyo, Tokyo, Japan.
J Gastroenterol Hepatol. 2024 Jan;39(1):157-164. doi: 10.1111/jgh.16369. Epub 2023 Oct 13.
Convolutional neural network (CNN) systems that automatically detect abnormalities from small-bowel capsule endoscopy (SBCE) images are still experimental, and no studies have directly compared the clinical usefulness of different systems. We compared endoscopist readings using an existing and a novel CNN system in a real-world SBCE setting.
Thirty-six complete SBCE videos, including 43 abnormal lesions (18 mucosal breaks, 8 angioectasia, and 17 protruding lesions), were retrospectively prepared. Three reading processes were compared: (A) endoscopist readings without CNN screening, (B) endoscopist readings after an existing CNN screening, and (C) endoscopist readings after a novel CNN screening.
The mean number of small-bowel images was 14 747 per patient. Among these images, existing and novel CNN systems automatically captured 24.3% and 9.4% of the images, respectively. In this process, both systems extracted all 43 abnormal lesions. Next, we focused on the clinical usefulness. The detection rates of abnormalities by trainee endoscopists were not significantly different across the three processes: A, 77%; B, 67%; and C, 79%. The mean reading time of the trainees was the shortest during process C (10.1 min per patient), followed by processes B (23.1 min per patient) and A (33.6 min per patient). The mean psychological stress score while reading videos (scale, 1-5) was the lowest in process C (1.8) but was not significantly different between processes B (2.8) and A (3.2).
Our novel CNN system significantly reduced endoscopist reading time and psychological stress while maintaining the detectability of abnormalities. CNN performance directly affects clinical utility and should be carefully assessed.
自动从小肠胶囊内镜 (SBCE) 图像中检测异常的卷积神经网络 (CNN) 系统仍处于试验阶段，尚无研究直接比较不同系统的临床实用性。我们在真实的 SBCE 环境中比较了使用现有和新型 CNN 系统的内镜医师的阅读结果。
回顾性准备 36 段完整的 SBCE 视频，包括 43 个异常病变（18 个黏膜破裂、8 个血管扩张和 17 个突出性病变）。比较了三种阅读过程：（A）没有 CNN 筛查的内镜医师阅读，（B）现有 CNN 筛查后的内镜医师阅读，和（C）新型 CNN 筛查后的内镜医师阅读。
每位患者的小肠图像平均数量为 14747 张。在这些图像中，现有的和新型 CNN 系统分别自动捕获了 24.3%和 9.4%的图像。在这个过程中，两个系统都提取了所有 43 个异常病变。接下来，我们重点关注临床实用性。新手内镜医师在这三个过程中对异常病变的检出率没有显著差异：A 过程为 77%，B 过程为 67%，C 过程为 79%。新手内镜医师的阅读时间在 C 过程中最短（每位患者 10.1 分钟），其次是 B 过程（每位患者 23.1 分钟）和 A 过程（每位患者 33.6 分钟）。阅读视频时的平均心理压力评分（1-5 分）在 C 过程中最低（1.8），但在 B 过程（2.8）和 A 过程（3.2）中无显著差异。
我们的新型 CNN 系统显著减少了内镜医师的阅读时间和心理压力，同时保持了异常病变的检出率。CNN 的性能直接影响临床实用性，应仔细评估。
