From the Department of Radiology and Research Institute of Radiology (J.C., H.J.H., J.B.S., S.M.L., K.J., R.P., J.K., N.K.), Department of Convergence Medicine, Biomedical Engineering Research Center (J. Yun), and Department of Clinical Epidemiology and Biostatistics (M.J.K.), University of Ulsan College of Medicine, Asan Medical Center, 86 Asanbyeongwon-Gil, Songpa-Gu, Seoul 138-735, Korea; Department of Radiology, Soonchunhyang University Seoul Hospital, Soonchunhyang University College of Medicine, Seoul, Korea (J.J.); Department of Internal Medicine, Ajou University School of Medicine, Suwon, Korea (Y.L.); Department of Internal Medicine, Seoul St. Mary's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Korea (H.J.); and Coreline Soft, Seoul, Korea (J. Yi, D.Y., B.K.).
Radiology. 2022 Jan;302(1):187-197. doi: 10.1148/radiol.2021204164. Epub 2021 Oct 12.
Background Evaluation of interstitial lung disease (ILD) at CT is a challenging task that requires experience and is subject to substantial interreader variability. Purpose To investigate whether a proposed content-based image retrieval (CBIR) of similar chest CT images by using deep learning can aid in the diagnosis of ILD by readers with different levels of experience. Materials and Methods This retrospective study included patients with confirmed ILD after multidisciplinary discussion and available CT images identified between January 2000 and December 2015. Database was composed of four disease classes: usual interstitial pneumonia (UIP), nonspecific interstitial pneumonia (NSIP), cryptogenic organizing pneumonia, and chronic hypersensitivity pneumonitis. Eighty patients were selected as queries from the database. The proposed CBIR retrieved the top three similar CT images with diagnosis from the database by comparing the extent and distribution of different regional disease patterns quantified by a deep learning algorithm. Eight readers with varying experience interpreted the query CT images and provided their most probable diagnosis in two reading sessions 2 weeks apart, before and after applying CBIR. Diagnostic accuracy was analyzed by using McNemar test and generalized estimating equation, and interreader agreement was analyzed by using Fleiss κ. Results A total of 288 patients were included (mean age, 58 years ± 11 [standard deviation]; 145 women). After applying CBIR, the overall diagnostic accuracy improved in all readers (before CBIR, 46.1% [95% CI: 37.1, 55.3]; after CBIR, 60.9% [95% CI: 51.8, 69.3]; < .001). In terms of disease category, the diagnostic accuracy improved after applying CBIR in UIP (before vs after CBIR, 52.4% vs 72.8%, respectively; < .001) and NSIP cases (before vs after CBIR, 42.9% vs 61.6%, respectively; < .001). Interreader agreement improved after CBIR (before vs after CBIR Fleiss κ, 0.32 vs 0.47, respectively; = .005). Conclusion The proposed content-based image retrieval system for chest CT images with deep learning improved the diagnostic accuracy of interstitial lung disease and interreader agreement in readers with different levels of experience. © RSNA, 2021 See also the editorial by Wielpütz in this issue.
背景 对 CT 下的间质性肺病（ILD）进行评估是一项具有挑战性的任务，需要经验，并且存在很大的读者间差异。目的 本研究旨在探究通过深度学习对相似胸部 CT 图像进行基于内容的图像检索（CBIR）是否有助于不同经验水平的读者进行 ILD 诊断。材料与方法 本回顾性研究纳入了经多学科讨论确诊为 ILD 且有 CT 图像的患者，这些 CT 图像采集于 2000 年 1 月至 2015 年 12 月之间。数据库由四类疾病组成：寻常型间质性肺炎（UIP）、非特异性间质性肺炎（NSIP）、特发性机化性肺炎和慢性过敏性肺炎。从数据库中选择 80 名患者作为查询对象。通过深度学习算法对不同区域性疾病模式的程度和分布进行量化，提出的 CBIR 检索数据库中诊断结果最相似的前三幅 CT 图像。8 名不同经验水平的读者在两次阅读会话（间隔 2 周）中对查询 CT 图像进行解读，一次在应用 CBIR 之前，一次在之后。采用 McNemar 检验和广义估计方程分析诊断准确性，采用 Fleiss κ 分析读者间一致性。结果 共纳入 288 例患者（平均年龄，58 岁±11[标准差]；145 例女性）。应用 CBIR 后，所有读者的整体诊断准确性均提高（应用 CBIR 前，46.1%[95%CI：37.1,55.3%]；应用 CBIR 后，60.9%[95%CI：51.8,69.3%]；<.001）。就疾病类别而言，UIP（应用 CBIR 前 vs 后，分别为 52.4% vs 72.8%；<.001）和 NSIP 病例（应用 CBIR 前 vs 后，分别为 42.9% vs 61.6%；<.001）的诊断准确性提高。应用 CBIR 后，读者间一致性提高（应用 CBIR 前 vs 后 Fleiss κ 值，分别为 0.32 和 0.47；=.005）。结论 基于深度学习的胸部 CT 图像内容检索系统提高了不同经验水平读者对间质性肺病的诊断准确性和读者间一致性。
