Ragab Haissam, Westhaeusser Fabian, Ernst Anne, Yamamura Jin, Fuhlert Patrick, Zimmermann Marina, Sauerbeck Julia, Shenas Farzad, Özden Cansu, Weidmann Anna, Adam Gerhard, Bonn Stefan, Schramm Christoph
From the Department of Diagnostic and Interventional Radiology and Nuclear Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany (H.R., J.Y., J.S., F.S., C.Ö., A.W., G.A.); Institute of Medical Systems Biology, Center for Biomedical AI, Center for Molecular Neurobiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany (F.W., A.E., P.F., M.Z., S.B.); Department of Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany (M.Z., C.S.); Hamburg Center for Translational Immunology (HCTI), Hamburg, Germany (S.B., C.S.); and Martin Zeitz Center for Rare Diseases, University Medical Center Hamburg-Eppendorf, Hamburg, Germany (C.S.).
Radiol Artif Intell. 2023 Apr 19;5(3):e220160. doi: 10.1148/ryai.220160. eCollection 2023 May.
To develop, train, and validate a multiview deep convolutional neural network (DeePSC) for the automated diagnosis of primary sclerosing cholangitis (PSC) on two-dimensional MR cholangiopancreatography (MRCP) images.
This retrospective study included two-dimensional MRCP datasets of 342 patients (45 years ± 14 [SD]; 207 male patients) with confirmed diagnosis of PSC and 264 controls (51 years ± 16; 150 male patients). MRCP images were separated into 3-T ( = 361) and 1.5-T ( = 398) datasets, of which 39 samples each were randomly chosen as unseen test sets. Additionally, 37 MRCP images obtained with a 3-T MRI scanner from a different manufacturer were included for external testing. A multiview convolutional neural network was developed, specialized in simultaneously processing the seven images taken at different rotational angles per MRCP examination. The final model, DeePSC, derived its classification per patient from the instance expressing the highest confidence in an ensemble of 20 individually trained multiview convolutional neural networks. Predictive performance on both test sets was compared with that of four licensed radiologists using the Welch test.
DeePSC achieved an accuracy of 80.5% ± 1.3 (sensitivity, 80.0% ± 1.9; specificity, 81.1% ± 2.7) on the 3-T and 82.6% ± 3.0 (sensitivity, 83.6% ± 1.8; specificity, 80.0% ± 8.9) on the 1.5-T test set and scored even higher on the external test set (accuracy, 92.4% ± 1.1; sensitivity, 100.0% ± 0.0; specificity, 83.5% ± 2.4). DeePSC outperformed radiologists in average prediction accuracy by 5.5 ( = .34, 3 T) and 10.1 ( = .13, 1.5 T) percentage points.
Automated classification of PSC-compatible findings based on two-dimensional MRCP was achievable and demonstrated high accuracy on internal and external test sets. Neural Networks, Deep Learning, Liver Disease, MRI, Primary Sclerosing Cholangitis, MR Cholangiopancreatography © RSNA, 2023.
开发、训练并验证一种多视图深度卷积神经网络（DeePSC），用于在二维磁共振胰胆管造影（MRCP）图像上自动诊断原发性硬化性胆管炎（PSC）。
这项回顾性研究纳入了342例确诊为PSC的患者（45岁±14[标准差]；207例男性患者）和264例对照者（51岁±16；150例男性患者）的二维MRCP数据集。MRCP图像被分为3-T（n = 361）和1.5-T（n = 398）数据集，每个数据集中随机选取39个样本作为未见过的测试集。此外，还纳入了从不同制造商的3-T MRI扫描仪获得的37幅MRCP图像用于外部测试。开发了一种多视图卷积神经网络，专门用于同时处理每次MRCP检查在不同旋转角度拍摄的七幅图像。最终模型DeePSC根据20个单独训练的多视图卷积神经网络组成的集合中表达最高置信度的实例对每位患者进行分类。使用韦尔奇检验将两个测试集上的预测性能与四位有执照的放射科医生的预测性能进行比较。
DeePSC在3-T测试集上的准确率为80.5%±1.3（敏感性，80.0%±1.9；特异性，81.1%±2.7），在1.5-T测试集上的准确率为82.6%±3.0（敏感性，83.6%±1.8；特异性，80.0%±8.9），在外部测试集上得分更高（准确率，92.4%±1.1；敏感性，100.0%±0.0；特异性，83.5%±2.4）。DeePSC在平均预测准确率方面比放射科医生高出5.5个百分点（P = .34，3 T）和10.1个百分点（P = .13，1.5 T）。
基于二维MRCP对PSC相关表现进行自动分类是可行的，并且在内部和外部测试集上都显示出高准确率。神经网络、深度学习、肝脏疾病、MRI、原发性硬化性胆管炎、磁共振胰胆管造影 ©RSNA，2023。
