基于深度学习的 CT 扫描胰腺癌症检测：一项全国范围内的基于人群的研究。
Pancreatic Cancer Detection on CT Scans with Deep Learning: A Nationwide Population-based Study.
机构信息
From the Department of Medical Imaging (P.T.C., K.L.L.) and Division of Gastroenterology and Hepatology, Department of Internal Medicine (M.S.W., W.C.L.), National Taiwan University Hospital, National Taiwan University College of Medicine, Taipei, Taiwan; Institute of Applied Mathematical Sciences (T.W., D.C., W.W.) and Departments of Computer Science and Information Engineering (P.W.) and Internal Medicine, College of Medicine (M.S.W., W.C.L.), National Taiwan University, No. 1, Section 4, Roosevelt Rd, Taipei 10617, Taiwan; Department of Medical Imaging, National Taiwan University Cancer Center, Taipei, Taiwan (K.L.L.); NVIDIA, Bethesda, Md (H.R.R.); and National Health Insurance Administration, Ministry of Health and Welfare, Taipei, Taiwan (P.C.L.).
出版信息
Radiology. 2023 Jan;306(1):172-182. doi: 10.1148/radiol.220152. Epub 2022 Sep 13.
Background Approximately 40% of pancreatic tumors smaller than 2 cm are missed at abdominal CT. Purpose To develop and to validate a deep learning (DL)-based tool able to detect pancreatic cancer at CT. Materials and Methods Retrospectively collected contrast-enhanced CT studies in patients diagnosed with pancreatic cancer between January 2006 and July 2018 were compared with CT studies of individuals with a normal pancreas (control group) obtained between January 2004 and December 2019. An end-to-end tool comprising a segmentation convolutional neural network (CNN) and a classifier ensembling five CNNs was developed and validated in the internal test set and a nationwide real-world validation set. The sensitivities of the computer-aided detection (CAD) tool and radiologist interpretation were compared using the McNemar test. Results A total of 546 patients with pancreatic cancer (mean age, 65 years ± 12 [SD], 297 men) and 733 control subjects were randomly divided into training, validation, and test sets. In the internal test set, the DL tool achieved 89.9% (98 of 109; 95% CI: 82.7, 94.9) sensitivity and 95.9% (141 of 147; 95% CI: 91.3, 98.5) specificity (area under the receiver operating characteristic curve [AUC], 0.96; 95% CI: 0.94, 0.99), without a significant difference ( = .11) in sensitivity compared with the original radiologist report (96.1% [98 of 102]; 95% CI: 90.3, 98.9). In a test set of 1473 real-world CT studies (669 malignant, 804 control) from institutions throughout Taiwan, the DL tool distinguished between CT malignant and control studies with 89.7% (600 of 669; 95% CI: 87.1, 91.9) sensitivity and 92.8% specificity (746 of 804; 95% CI: 90.8, 94.5) (AUC, 0.95; 95% CI: 0.94, 0.96), with 74.7% (68 of 91; 95% CI: 64.5, 83.3) sensitivity for malignancies smaller than 2 cm. Conclusion The deep learning-based tool enabled accurate detection of pancreatic cancer on CT scans, with reasonable sensitivity for tumors smaller than 2 cm. © RSNA, 2022 See also the editorial by Aisen and Rodrigues in this issue.
背景
在腹部 CT 检查中，约有 40%的小于 2cm 的胰腺肿瘤漏诊。目的：开发并验证一种基于深度学习（DL）的工具，使其能够在 CT 上检测胰腺癌。材料与方法：回顾性收集 2006 年 1 月至 2018 年 7 月期间诊断为胰腺癌的患者的增强 CT 研究，并与 2004 年 1 月至 2019 年 12 月期间获得的正常胰腺（对照组）的 CT 研究进行比较。开发了一个端到端的工具，包括一个分割卷积神经网络（CNN）和一个集成五个 CNN 的分类器，并在内部测试集和全国范围内的真实世界验证集中进行了验证。使用 McNemar 检验比较计算机辅助检测（CAD）工具和放射科医生解读的敏感性。结果：共纳入 546 例胰腺癌患者（平均年龄 65 岁±12[SD]，297 例男性）和 733 例对照组患者，随机分为训练集、验证集和测试集。在内部测试集中，DL 工具的敏感性为 89.9%（109 例中的 98 例；95%CI：82.7%，94.9%），特异性为 95.9%（147 例中的 141 例；95%CI：91.3%，98.5%）（受试者工作特征曲线下面积[AUC]，0.96；95%CI：0.94，0.99），与原始放射科医生报告（敏感性为 96.1%[102 例中的 98 例；95%CI：90.3%，98.9%]）相比，敏感性差异无统计学意义（=0.11）。在来自台湾各机构的 1473 例真实世界 CT 研究（669 例恶性，804 例对照）的测试集中，DL 工具区分 CT 恶性和对照研究的敏感性为 89.7%（669 例中的 600 例；95%CI：87.1%，91.9%），特异性为 92.8%（804 例中的 746 例；95%CI：90.8%，94.5%）（AUC，0.95；95%CI：0.94，0.96），对于小于 2cm 的恶性肿瘤，敏感性为 74.7%（91 例中的 68 例；95%CI：64.5%，83.3%）。结论：基于深度学习的工具能够在 CT 扫描中准确检测胰腺癌，对小于 2cm 的肿瘤具有合理的敏感性。©RSNA，2022 另请参见本期 Aisen 和 Rodrigues 的社论。
Zotero 插件