基于深度神经网络的专用CT肺动脉造影自动肺栓塞检测准确性研究
本研究回顾性纳入2022年9月至2023年1月因疑似肺栓塞行CT肺动脉造影的1000例患者,最终分析903例(肺栓塞患病率12%),评估基于深度神经网络的原型算法自动检测肺栓塞的性能。模型敏感度84.6%、特异度95.1%、阳性预测值70.5%、阴性预测值97.8%、总准确率93.8%。假阳性常见于肺部肿块、肺炎和对比剂流动伪影,假阴性常见于慢性肺栓塞和亚段肺栓塞。
要点
- 模型检测肺栓塞敏感度84.6%、特异度95.1%、阴性预测值97.8%,总准确率93.8%。
- 假阳性主要来自肺部肿块、肺炎和对比剂流动伪影。
- 假阴性主要来自慢性肺栓塞和亚段肺栓塞。
核心结论
- 基于深度神经网络的原型算法对肺栓塞检测表现出优异性能,提示其可能有助于支持放射科医生临床阅片和检查优先级排序。
研究设计与样本
- 回顾性研究,纳入2022年9月至2023年1月因疑似肺栓塞行CT肺动脉造影的患者1000例(58.8%为女性),使用三种CT系统(SOMATOM Force、go.Top和Definition AS;Siemens Healthineers,Forchheim,Germany),以临床报告为参考标准,最终分析903例。
诊断标准与关键数值
- 深度神经网络模型检测肺栓塞的敏感度84.6%、特异度95.1%、阳性预测值70.5%、阴性预测值97.8%、总准确率93.8%。
临床应用与阅片要点
- 作者认为该算法可能有助于支持放射科医生临床阅片和检查优先级排序;阅片时需注意肺部肿块、肺炎、对比剂流动伪影可致假阳性,慢性肺栓塞和亚段肺栓塞可致假阴性。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
PURPOSE: To assess the performance of a Deep Neural Network (DNN)-based prototype algorithm for automated PE detection on CTPA scans.
METHODS: Patients who had previously undergone CTPA with three different systems (SOMATOM Force, go.Top, and Definition AS; Siemens Healthineers, Forchheim, Germany) because of suspected PE from September 2022 to January 2023 were retrospectively enrolled in this study (n = 1,000, 58.8 % women). For detailed evaluation, all PE were divided into three location-based subgroups: central arteries, lobar branches, and peripheral regions. Clinical reports served as ground truth. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were determined to evaluate the performance of DNN-based PE detection.
RESULTS: Cases were excluded due to incomplete data (n = 32), inconclusive report (n = 17), insufficient contrast detected in the pulmonary trunk (n = 40), or failure of the preprocessing algorithms (n = 8). Therefore, the final cohort included 903 cases with a PE prevalence of 12 % (n = 110). The model achieved a sensitivity, specificity, PPV, and NPV of 84.6, 95.1, 70.5, and 97.8 %, respectively, and delivered an overall accuracy of 93.8 %. Among the false positive cases (n = 39), common sources of error included lung masses, pneumonia, and contrast flow artifacts. Common sources of false negatives (n = 17) included chronic and subsegmental PEs.
CONCLUSION: The proposed DNN-based algorithm provides excellent performance for the detection of PE, suggesting its potential utility to support radiologists in clinical reading and exam prioritization.
引用
- Zsarnoczay E, Rapaka S, Schoepf UJ, et al. Accuracy of a deep neural network for automated pulmonary embolism detection on dedicated CT pulmonary angiograms. Eur J Radiol. 2025. PMID: 40187199. doi:10.1016/j.ejrad.2025.112077
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。