Joseph Jacob1, B J Bartholmai2, S Rajagopalan3, R Karwoski3, S M Mak4, W Mok4, G Della Casa5, K Sugino6, S L F Walsh7, A U Wells8, D M Hansell4. 1. Department of Radiology, Royal Brompton Hospital, Royal Brompton and Harefield NHS Foundation Trust, London, UK. akelajacob@gmail.com. 2. Division of Radiology, Mayo Clinic Rochester, Rochester, MN, USA. 3. Biomedical Imaging Resource, Mayo Clinic Rochester, Rochester, MN, USA. 4. Department of Radiology, Royal Brompton Hospital, Royal Brompton and Harefield NHS Foundation Trust, London, UK. 5. Universita degli Studi di Modena e Reggio Emilia, Modena and Reggio Emilia, Modena, Emilia-Romagna, Italy. 6. Toho University Omori Medical Centre, Tokyo, Japan. 7. Kings College Hospital, London, UK. 8. Interstitial Lung Disease Unit, Royal Brompton Hospital, Royal Brompton and Harefield NHS Foundation Trust, London, UK.
Abstract
BACKGROUND: Hypersensitivity pneumonitis (HP) has a variable clinical course. Modelling of quantitative CALIPER-derived CT data can identify distinct disease phenotypes. Mortality prediction using CALIPER analysis was compared to the interstitial lung disease gender, age, physiology (ILD-GAP) outcome model. METHODS: CALIPER CT analysis of parenchymal patterns in 98 consecutive HP patients was compared to visual CT scoring by two radiologists. Functional indices including forced vital capacity (FVC) and diffusion capacity for carbon monoxide (DLco) in univariate and multivariate Cox mortality models. Automated stratification of CALIPER scores was evaluated against outcome models. RESULTS: Univariate predictors of mortality included visual and CALIPER CT fibrotic patterns, and all functional indices. Multivariate analyses identified only two independent predictors of mortality: CALIPER reticular pattern (p = 0.001) and DLco (p < 0.0001). Automated stratification distinguished three distinct HP groups (log-rank test p < 0.0001). Substitution of automated stratified groups for FVC and DLco in the ILD-GAP model demonstrated no loss of model strength (C-Index = 0.73 for both models). Model strength improved when automated stratified groups were combined with the ILD-GAP model (C-Index = 0.77). CONCLUSIONS: CALIPER-derived variables are the strongest CT predictors of mortality in HP. Automated CT stratification is equivalent to functional indices in the ILD-GAP model for predicting outcome in HP. KEY POINTS: • Computer CT analysis better predicts mortality than visual CT analysis in HP. • Quantitative CT analysis is equivalent to functional indices for prognostication in HP. • Prognostication using the ILD-GAP model improves when combined with quantitative CT analysis.
BACKGROUND:Hypersensitivitypneumonitis (HP) has a variable clinical course. Modelling of quantitative CALIPER-derived CT data can identify distinct disease phenotypes. Mortality prediction using CALIPER analysis was compared to the interstitial lung disease gender, age, physiology (ILD-GAP) outcome model. METHODS: CALIPER CT analysis of parenchymal patterns in 98 consecutive HP patients was compared to visual CT scoring by two radiologists. Functional indices including forced vital capacity (FVC) and diffusion capacity for carbon monoxide (DLco) in univariate and multivariate Cox mortality models. Automated stratification of CALIPER scores was evaluated against outcome models. RESULTS: Univariate predictors of mortality included visual and CALIPER CT fibrotic patterns, and all functional indices. Multivariate analyses identified only two independent predictors of mortality: CALIPER reticular pattern (p = 0.001) and DLco (p < 0.0001). Automated stratification distinguished three distinct HP groups (log-rank test p < 0.0001). Substitution of automated stratified groups for FVC and DLco in the ILD-GAP model demonstrated no loss of model strength (C-Index = 0.73 for both models). Model strength improved when automated stratified groups were combined with the ILD-GAP model (C-Index = 0.77). CONCLUSIONS: CALIPER-derived variables are the strongest CT predictors of mortality in HP. Automated CT stratification is equivalent to functional indices in the ILD-GAP model for predicting outcome in HP. KEY POINTS: • Computer CT analysis better predicts mortality than visual CT analysis in HP. • Quantitative CT analysis is equivalent to functional indices for prognostication in HP. • Prognostication using the ILD-GAP model improves when combined with quantitative CT analysis.
Authors: Sushravya Raghunath; Srinivasan Rajagopalan; Ronald A Karwoski; Brian J Bartholmai; Richard A Robb Journal: Med Image Comput Comput Assist Interv Date: 2011
Authors: Kevin R Flaherty; Jeanette A Mumford; Susan Murray; Ella A Kazerooni; Barry H Gross; Thomas V Colby; William D Travis; Andrew Flint; Galen B Toews; Joseph P Lynch; Fernando J Martinez Journal: Am J Respir Crit Care Med Date: 2003-05-28 Impact factor: 21.405
Authors: Peter J Castaldi; Raúl San José Estépar; Carlos S Mendoza; Craig P Hersh; Nan Laird; James D Crapo; David A Lynch; Edwin K Silverman; George R Washko Journal: Am J Respir Crit Care Med Date: 2013-11-01 Impact factor: 21.405
Authors: Joseph Jacob; Brian J Bartholmai; Srinivasan Rajagopalan; Maria Kokosi; Arjun Nair; Ronald Karwoski; Sushravya M Raghunath; Simon L F Walsh; Athol U Wells; David M Hansell Journal: J Thorac Imaging Date: 2016-09 Impact factor: 3.000
Authors: Joseph Jacob; Brian J Bartholmai; Ryoko Egashira; Anne Laure Brun; Srinivasan Rajagopalan; Ronald Karwoski; Maria Kokosi; David M Hansell; Athol U Wells Journal: BMC Pulm Med Date: 2017-05-04 Impact factor: 3.317
Authors: Toby M Maher; Kevin K Brown; Michael Kreuter; Anand Devaraj; Simon L F Walsh; Lisa H Lancaster; Elizabeth A Belloli; Maria Padilla; Juergen Behr; Rainer-Georg Goeldner; Kay Tetzlaff; Rozsa Schlenker-Herceg; Kevin R Flaherty Journal: Eur Respir J Date: 2022-02-03 Impact factor: 16.671