PURPOSE: The variability in physiologic testing for evaluation of interstitial lung disease (ILD) and lack of objective imaging gold standard make accurate prognostication and disease management difficult. New image analysis tools facilitate characterization and quantitative analysis of disease through CT imaging. The Lung Tissue Research Consortium (LTRC) provides a unique repository containing in-depth clinical and volumetric high-resolution CT (HRCT) data which can be used to validate quantitative techniques. We hypothesize a relationship between physiological parameters and automated quantitative measures of lung parenchyma obtained from HRCT.
METHODS: Characterization of pulmonary HRCT was performed using CALIPER (Computer-Aided Lung Informatics for Pathology Evaluation and Rating), which included automated extraction of the lungs from standardized HRCT scans. Normal, ground glass, reticular densities, and honeycombing were quantified by supervised classification of these volumetric CT scans based on histogram signatures and training data obtained by consensus assessment of radiologists. Correlation of CT classification with physiologic parameters was performed using Spearman's correlation.
RESULTS: The patient population included 119 subjects with proven diagnosis of ILD. Significant correlations are noted between multiple physiologic parameters and quantitative tissue characterization. Specifically, percent involvement of reticular densities correlated with significant changes in 6 minute walk total distance (r=-0.32; p<0.001; n=108), FVC pre-bronchodilator (r=-0.63; p<0.001; n=119), DLCO (r=-0.65; p<0.001; n=101) and TLC (r=-0.44; p<0.001; n=98). Similarly, significant correlation between lung classified as normal and physiologic tests was demonstrated, including 6-minute walk test (r=0.32; p<0.001; n=108), FVC pre-bronchodilator (r=0.66; p<0.001; n=119), DLCO (r=0.59; p<0.001; n=101) and TLC (r=0.56; p<0.001; n=98). Lower correlation between detected ground glass opacity and physiologic measures is possibly due to either less physiologic impact of these abnormalities or misclassification of that feature of disease.
CONCLUSIONS: Classification and quantitative analysis of diffuse pulmonary disease on CT correlates with known physiologic indicators of ILD severity.
CLINICAL IMPLICATIONS: An automated tool like CALIPER has the potential to provide a reproducible standard for evaluation of disease progression and/or response to therapy over time.
DISCLOSURE: The following authors have nothing to disclose: Sushravya Raghunath, Teng Moua, Colin Segovis, Fabien Maldonado, Jay Ryu, Paul Decker, Srinivasan Rajagopalan, Ronald Karwoski, Richard Robb, Brian Bartholmai
No Product/Research Disclosure Information