Fig. 2: Independent testing of the CXR-Lung-Risk model in the PLCO testing dataset and in NLST to estimate lung disease mortality.
From: Deep learning to estimate lung disease mortality from chest radiographs
The CXR-Lung-Risk model was independently tested in (a) the PLCO testing set (n = 10155 independent individuals) and in (b) NLST (n = 5414 independent individuals). Kaplan-Meier survival analysis shows a graded association between CXR-Lung-Risk groups and lung disease mortality. Pairwise comparison of survival curves was performed using two-sided Log-Rank tests. P-values are adjusted for multiple comparisons using the Bonferroni-Holm method. Forest plots show univariable and multivariable-adjusted hazard ratios (box) with 95% confidence intervals (error bars) for the different CXR-Lung-Risk groups. Multivariable models are adjusted for: chronological age, sex, race, smoking status, pack years, body mass index, prevalent diabetes mellitus, hypertension, history of stroke, myocardial infarction, cancer and 9 chest x-ray findings as described in the methods. Source data are provided as a Source Data file. ***p values <2*10−16. CXR chest radiograph, PLCO Prostate, Lung, Colorectal, Ovarian Cancer Screening Trial, NLST National Lung Screening Trial; y years.
