From the Department of Radiology (J.G.N., H.K., J.M.G., C.M.P.) and Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine (C.H.L.), Seoul National University Hospital, Seoul, Republic of Korea; Seoul National University College of Medicine, 101 Daehak-ro, Jongno-gu, Seoul 03080, Republic of Korea (J.G.N., H.K., C.R., J.M.G., C.H.L., C.M.P.); Division of Pulmonary Medicine, Department of Internal Medicine, Veteran Health Service Medical Center, Seoul, Republic of Korea (H.R.K.); Department of Radiology (S.M.L.), Research Institute of Radiology (S.M.L.), Department of Pulmonary and Critical Care Medicine (Y.M.O.), and Clinical Research Center for Chronic Obstructive Airway Diseases (Y.M.O.), Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea; Institute of Radiation Medicine (J.M.G., C.M.P.) and Institute of Medical and Biological Engineering (C.M.P.), Seoul National University Medical Research Center, Seoul, Republic of Korea.
Radiology. 2022 Oct;305(1):199-208. doi: 10.1148/radiol.212071. Epub 2022 Jun 7.
Background Preexisting indexes for predicting the prognosis of chronic obstructive pulmonary disease (COPD) do not use radiologic information and are impractical because they involve complex history assessments or exercise tests. Purpose To develop and to validate a deep learning-based survival prediction model in patients with COPD (DLSP) using chest radiographs, in addition to other clinical factors. Materials and Methods In this retrospective study, data from patients with COPD who underwent postbronchodilator spirometry and chest radiography from 2011-2015 were collected and split into training ( = 3475), validation ( = 435), and internal test ( = 315) data sets. The algorithm for predicting survival from chest radiographs was trained (hereafter, DLSP), and then age, body mass index, and forced expiratory volume in 1 second (FEV) were integrated within the model (hereafter, DLSP). For external test, three independent cohorts were collected ( = 394, 416, and 337). The discrimination performance of DLSP was evaluated by using time-dependent area under the receiver operating characteristic curves (TD AUCs) at 5-year survival. Goodness of fit was assessed by using the Hosmer-Lemeshow test. Using one external test data set, DLSP was compared with four COPD-specific clinical indexes: BODE, ADO, COPD Assessment Test (CAT), and St George's Respiratory Questionnaire (SGRQ). Results DLSP had a higher performance at predicting 5-year survival than FEV in two of the three external test cohorts (TD AUC: 0.73 vs 0.63 [ = .004]; 0.67 vs 0.60 [ = .01]; 0.76 vs 0.77 [ = .91]). DLSP demonstrated good calibration in all cohorts. The DLSP model showed no differences in TD AUC compared with BODE (0.87 vs 0.80; = .34), ADO (0.86 vs 0.89; = .51), and SGRQ (0.86 vs 0.70; = .09), and showed higher TD AUC than CAT (0.93 vs 0.55; < .001). Conclusion A deep learning model using chest radiographs was capable of predicting survival in patients with chronic obstructive pulmonary disease. © RSNA, 2022
背景 预测慢性阻塞性肺疾病（COPD）预后的现有指标不使用放射学信息，而且不切实际，因为它们涉及复杂的病史评估或运动测试。
目的 利用胸部 X 线片和其他临床因素，开发并验证一种用于 COPD 患者的基于深度学习的生存预测模型（DLSP）。
材料与方法 在这项回顾性研究中，收集了 2011 年至 2015 年间接受支气管扩张剂后肺量测定和胸部 X 线检查的 COPD 患者的数据，并将其分为训练集（=3475）、验证集（=435）和内部测试集（=315）。训练用于预测生存的基于放射照片的算法（以下简称 DLSP），然后在模型中整合年龄、体重指数和 1 秒用力呼气量（FEV）（以下简称 DLSP）。为了外部测试，收集了三个独立的队列（=394、416 和 337）。通过使用 5 年生存时间依赖性接收器工作特征曲线下面积（TD AUC）评估 DLSP 的鉴别性能。通过 Hosmer-Lemeshow 检验评估拟合优度。使用一个外部测试数据集，将 DLSP 与四种 COPD 特异性临床指标进行比较：BODE、ADO、COPD 评估测试（CAT）和圣乔治呼吸问卷（SGRQ）。
结果 在两个外部测试队列中的三个队列中，DLSP 在预测 5 年生存率方面的表现均优于 FEV（TD AUC：0.73 比 0.63 [=0.004]；0.67 比 0.60 [=0.01]；0.76 比 0.77 [=0.91]）。DLSP 在所有队列中均具有良好的校准度。DLSP 模型与 BODE（0.87 比 0.80；=0.34）、ADO（0.86 比 0.89；=0.51）和 SGRQ（0.86 比 0.70；=0.09）相比，TD AUC 没有差异，与 CAT（0.93 比 0.55；<0.001）相比，TD AUC 更高。
结论 使用胸部 X 射线的深度学习模型能够预测慢性阻塞性肺疾病患者的生存率。
