BACKGROUND: Early assessment of chronic obstructive pulmonary disease (COPD) and other respiratory disorders relies heavily on pulmonary function tests (PFTs). However, PFTs are effort-dependent and require strict quality control, limiting their application in certain clinical scenarios requiring broad accessibility or repeated assessment. Purpose: This study developed a multidimensional feature-based model derived from paired inspiratory-expiratory CT images for the noninvasive and quantitative prediction of key pulmonary function indices. Methods: A multidimensional feature set was constructed by integrating clinical variables, parametric response mapping (PRM) -derived lung density distribution features, registration-based ventilation heterogeneity features, and deformation-based local biomechanical response features. An elastic-net regression model was then developed to predict the ratio of forced expiratory volume in one second to forced vital capacity (FEV1/FVC) and the percent predicted forced expiratory volume in one second (FEV1%pred). Results: The model was evaluated on 120 subjects with paired inspiratory-expiratory CT scans and PFT data. The proposed multidimensional feature prediction model showed numerically better performance than models using clinical or PRM features alone for both prediction tasks. In the test set, the model achieved an R2 of 0. 572 with a mean absolute error (MAE) of 8. 38 for FEV1/FVC, and an R2 of 0. 472 with an MAE of 13. 65 for FEV1%pred. Interpretability analysis using SHapley Additive exPlanations (SHAP) revealed that among the newly proposed features, the proportion of voxels in the ventilation interval "0 < E-I < 50" (Vol0-50), as well as the higher-order statistics of the Jacobian determinant specifically fSADJSkew, fSADJKurt and EmphJSkew, were critical predictors of pulmonary function. Conclusion: This method may serve as a complementary tool to PFTs for the quantitative prediction of pulmonary function indices in opportunistic functional assessment and longitudinal follow-up among patients who have already undergone dual-phase CT. .
李玉赞 et al. (Wed,) studied this question.