From the Department of Radiology and Research Institute of Radiology (H.J.P., S.Y.K., S.H.P., J.H.B., H.J.K.) and Department of Bioengineering, Asan Medical Institute of Convergence Science and Technology (K.S., S.G.K., N.K.), Asan Medical Center, University of Ulsan College of Medicine, 88 Olympic-ro 43-gil, Songpa-gu, Seoul 05505, Republic of Korea; and Department of Radiology, Kyung Hee University Hospital, Seoul, Republic of Korea (M.W.Y.).
Radiology. 2023 Jan;306(1):140-149. doi: 10.1148/radiol.220171. Epub 2022 Aug 23.
Background Deep learning (DL) may facilitate the diagnosis of various pancreatic lesions at imaging. Purpose To develop and validate a DL-based approach for automatic identification of patients with various solid and cystic pancreatic neoplasms at abdominal CT and compare its diagnostic performance with that of radiologists. Materials and Methods In this retrospective study, a three-dimensional nnU-Net-based DL model was trained using the CT data of patients who underwent resection for pancreatic lesions between January 2014 and March 2015 and a subset of patients without pancreatic abnormality who underwent CT in 2014. Performance of the DL-based approach to identify patients with pancreatic lesions was evaluated in a temporally independent cohort (test set 1) and a temporally and spatially independent cohort (test set 2) and was compared with that of two board-certified radiologists. Performance was assessed using receiver operating characteristic analysis. Results The study included 852 patients in the training set (median age, 60 years [range, 19-85 years]; 462 men), 603 patients in test set 1 (median age, 58 years [range, 18-82 years]; 376 men), and 589 patients in test set 2 (median age, 63 years [range, 18-99 years]; 343 men). In test set 1, the DL-based approach had an area under the receiver operating characteristic curve (AUC) of 0.91 (95% CI: 0.89, 0.94) and showed slightly worse performance in test set 2 (AUC, 0.87 [95% CI: 0.84, 0.89]). The DL-based approach showed high sensitivity in identifying patients with solid lesions of any size (98%-100%) or cystic lesions measuring 1.0 cm or larger (92%-93%), which was comparable with the radiologists (95%-100% for solid lesions [ = .51 to > .99]; 93%-98% for cystic lesions ≥1.0 cm [ = .38 to > .99]). Conclusion The deep learning-based approach demonstrated high performance in identifying patients with various solid and cystic pancreatic lesions at CT. © RSNA, 2022
背景 深度学习（DL）可辅助对各种胰腺病变进行影像学诊断。目的 旨在开发和验证一种基于深度学习的方法，用于自动识别腹部 CT 检查中各种胰腺实性和囊性肿瘤患者，并与放射科医师的诊断表现进行比较。
材料与方法 本回顾性研究利用 2014 年 1 月至 2015 年 3 月间因胰腺病变接受切除术患者的 CT 数据和 2014 年接受 CT 检查且无胰腺异常患者的部分 CT 数据对一个三维 nnU-Net 深度学习模型进行了训练。在一个时间独立的队列（测试集 1）和一个时间和空间独立的队列（测试集 2）中评估了该基于深度学习的方法对识别胰腺病变患者的表现，并与两名经过董事会认证的放射科医师的表现进行了比较。采用受试者工作特征曲线分析评估了表现。
结果 研究纳入了训练集中的 852 例患者（中位年龄，60 岁[范围，19-85 岁]；462 例男性）、测试集 1 中的 603 例患者（中位年龄，58 岁[范围，18-82 岁]；376 例男性）和测试集 2 中的 589 例患者（中位年龄，63 岁[范围，18-99 岁]；343 例男性）。在测试集 1 中，基于深度学习的方法的受试者工作特征曲线下面积（AUC）为 0.91（95%CI：0.89，0.94），在测试集 2 中的表现略差（AUC：0.87[95%CI：0.84，0.89]）。基于深度学习的方法在识别任何大小的实性病变（98%-100%）或直径≥1.0 cm 的囊性病变（92%-93%）患者方面具有较高的敏感度，与放射科医师的表现相当（实性病变 95%-100%[ =.51 至 >.99]；直径≥1.0 cm 的囊性病变 93%-98%[ =.38 至 >.99]）。
结论 基于深度学习的方法在 CT 检查中识别各种胰腺实性和囊性病变患者方面表现出较高的性能。
© 2022 RSNA
