Objectives: To evaluate automated texture-based segmentation of dual-energy CT (DECT) images in diffuse interstitial lung disease (DILD) patients and prognostic stratification by overlapping morphologic and perfusion information of total lung.
Methods: Suspected DILD patients scheduled for surgical biopsy were prospectively included. Texture patterns included ground-glass opacity (GGO), reticulation and consolidation. Pattern- and perfusion-based CT measurements were assessed to extract quantitative parameters. Accuracy of texture-based segmentation was analysed. Correlations between CT measurements and pulmonary function test or 6-minute walk test (6MWT) were calculated. Parameters of idiopathic pulmonary fibrosis/usual interstitial pneumonia (IPF/UIP) and non-IPF/UIP were compared. Survival analysis was performed.
Results: Overall accuracy was 90.47% for whole lung segmentation. Correlations between mean iodine values of total lung, 50-97.5th (%) attenuation and forced vital capacity or 6MWT were significant. Volume of GGO, reticulation and consolidation had significant correlation with DLco or SpO2 on 6MWT. Significant differences were noted between IPF/UIP and non-IPF/UIP in 6MWT distance, mean iodine value of total lung, 25-75th (%) attenuation and entropy. IPF/UIP diagnosis, GGO ratio, DILD extent, 25-75th (%) attenuation and SpO2 on 6MWT showed significant correlations with survival.
Conclusion: DECT combined with pattern analysis is useful for analysing DILD and predicting survival by provision of morphology and enhancement.
Key points: • Dual-energy CT (DECT) produces morphologic and parenchymal enhancement information. • Automated lung segmentation enables analysis of disease extent and severity. • This prospective study showed value of DECT in DILD patients. • Parameters on DECT enable characterization and survival prediction of DILD.
Keywords: DECT; DILD; IPF/UIP; Perfusion- or pattern-based CT quantification parameters; Survival.