Department of Gastroenterology, São João University Hospital, Porto, Portugal; WGO Gastroenterology and Hepatology Training Center, Porto, Portugal; Department of Gastroenterology, Faculty of Medicine, University of Porto, Porto, Portugal.
Department of Gastroenterology, São João University Hospital, Porto, Portugal; WGO Gastroenterology and Hepatology Training Center, Porto, Portugal.
Gastrointest Endosc. 2022 Feb;95(2):339-348. doi: 10.1016/j.gie.2021.08.027. Epub 2021 Sep 8.
The diagnosis and characterization of biliary strictures (BSs) is challenging. The introduction of digital single-operator cholangioscopy (DSOC) that allows direct visual inspection of the lesion and targeted biopsy sampling significantly improved the diagnostic yield in patients with indeterminate BSs. However, the diagnostic efficiency of DSOC remains suboptimal. Convolutional neural networks (CNNs) have shown great potential for the interpretation of medical images. We aimed to develop a CNN-based system for automatic detection of malignant BSs in DSOC images.
We developed, trained, and validated a CNN-based on DSOC images. Each frame was labeled as a normal/benign finding or as a malignant lesion if histopathologic evidence of biliary malignancy was available. The entire dataset was split for 5-fold cross-validation. In addition, the image dataset was split for constitution of training and validation datasets. The performance of the CNN was measured by calculating the area under the receiving operating characteristic curve (AUC), sensitivity, specificity, and positive and negative predictive values.
A total of 11,855 images from 85 patients were included (9695 malignant strictures and 2160 benign findings). The model had an overall accuracy of 94.9%, sensitivity of 94.7%, specificity of 92.1%, and AUC of .988 in cross-validation analysis. The image processing speed of the CNN was 7 ms per frame.
The developed deep learning algorithm accurately detected and differentiated malignant strictures from benign biliary conditions. The introduction of artificial intelligence algorithms to DSOC systems may significantly increase its diagnostic yield for malignant strictures.
胆管狭窄（BS）的诊断和特征描述具有挑战性。数字单操作员胆管镜（DSOC）的引入允许对病变进行直接目视检查和靶向活检采样，这显著提高了不确定 BS 患者的诊断产量。然而，DSOC 的诊断效率仍然不理想。卷积神经网络（CNN）在医学图像解释方面显示出巨大的潜力。我们旨在开发一种基于 CNN 的系统，用于自动检测 DSOC 图像中的恶性 BS。
我们开发、训练和验证了一种基于 DSOC 图像的 CNN。如果组织病理学证据表明存在胆管恶性肿瘤，则将每一帧标记为正常/良性发现，或标记为恶性病变。整个数据集被分为 5 折交叉验证。此外，图像数据集被分为训练和验证数据集。通过计算接收者操作特征曲线下的面积（AUC）、敏感性、特异性、阳性和阴性预测值来衡量 CNN 的性能。
共纳入 85 名患者的 11855 张图像（9695 例恶性狭窄和 2160 例良性发现）。在交叉验证分析中，该模型的总体准确率为 94.9%，敏感性为 94.7%，特异性为 92.1%，AUC 为.988。CNN 的图像处理速度为每帧 7 毫秒。
开发的深度学习算法准确地检测和区分了恶性狭窄和良性胆管病变。人工智能算法引入 DSOC 系统可能会显著提高其对恶性狭窄的诊断产量。
