Disease Staging and Prognosis in Smokers Using Deep Learning in Chest Computed Tomography
Disease Staging and Prognosis in Smokers Using Deep Learning in Chest Computed Tomography
复制标题
胸部CT深度学习对吸烟者疾病分期和预后的影响
DOI：
10.1164/rccm.201705-0860oc
复制
发表时间：
2018-01-15
影响因子：
24.7
通讯作者：
Washko, George R.
中科院分区：
文献类型：
--
作者：
Gonzalez, German;Ash, Samuel Y.;Washko, George R.
基本原理：深度学习是一种强大的工具，可以改善结果预测。目的：确定深度学习，特别是卷积神经网络（CNN）分析，是否可以检测和分期慢性阻塞性肺疾病（COPD），并预测吸烟者的急性呼吸道疾病（ARD）事件和死亡率。方法：CNN使用来自7，983名COPDGene参与者的计算机断层扫描进行训练，并使用1，000名非重叠COPDGene参与者和1，672名ECLIPSE参与者进行评估。Logistic回归（C统计量和Hosmer-Lemeshow检验）用于评估COPD诊断和ARD预测。考克斯回归（C指数和Greenwood-Nam-D 'Agnostino检验）用于评估死亡率。共有51.1%的COPDGene参与者准确分期，74.95%在一个阶段内。在ECLIPSE中，29.4%被准确分期，74.6%在一期内。在COPDGene和ECLIPSE中，ARD事件的C统计量分别为0.64和0.55，Hosmer-Lemeshow P值分别为0.502和0.380，表明没有证据表明校准不良。在COPDGene和ECLIPSE中，CNN以公平的区分预测死亡率（C指数分别为0.72和0.60），且无校准不良证据（Greenwood-Nam-D 'Agnostino P值分别为0.307和0.331）。很深的-仅使用计算机断层扫描成像数据的学习方法可以识别患有COPD的吸烟者，并预测谁最有可能患有ARD事件和死亡率最高的事件。在人群水平上，CNN分析可能是风险评估的有力工具。
Rationale: Deep learning is a powerful tool that may allow for improved outcome prediction.Objectives: To determine if deep learning, specifically convolutional neural network (CNN) analysis, could detect and stage chronic obstructive pulmonary disease (COPD) and predict acute respiratory disease (ARD) events and mortality in smokers.Methods: A CNN was trained using computed tomography scans from 7,983 COPDGene participants and evaluated using 1,000 nonoverlapping COPDGene participants and 1,672 ECLIPSE participants. Logistic regression (C statistic and the Hosmer-Lemeshow test) was used to assess COPD diagnosis and ARD prediction. Cox regression (C index and the Greenwood-Nam-D'Agnostino test) was used to assess mortality.Measurements and Main Results: In COPDGene, the C statistic for the detection of COPD was 0.856. A total of 51.1% of participants in COPDGene were accurately staged and 74.95% were within one stage. In ECLIPSE, 29.4% were accurately staged and 74.6% were within one stage. In COPDGene and ECLIPSE, the C statistics for ARD events were 0.64 and 0.55, respectively, and the Hosmer-Lemeshow P values were 0.502 and 0.380, respectively, suggesting no evidence of poor calibration. In COPDGene and ECLIPSE, CNN predicted mortality with fair discrimination (C indices, 0.72 and 0.60, respectively), and without evidence of poor calibration (Greenwood-Nam-D'Agnostino P values, 0.307 and 0.331, respectively).Conclusions: A deep-learning approach that uses only computed tomography imaging data can identify those smokers who have COPD and predict who are most likely to have ARD events and those with the highest mortality. At a population level CNN analysis may be a powerful tool for risk assessment.