Computer based analysis of diffuse fibrosing lung diseases

Abstract

Computer based CT analysis of interstitial lung disease (ILD) may finally have come of age with sophisticated new software algorithms that can analyse CTs of patients with idiopathic pulmonary fibrosis (IPF) with similar, or superior, precision to visual scoring by radiologists. However, the latest iterations of such tools have not been evaluated in large populations or in fibrosing lung diseases (FLD) other than IPF.

The investigations in this thesis aimed to identify the role of a computer-based tool CALIPER, compared to visual CT scoring and pulmonary function indices, in predicting prognosis at baseline in a large IPF cohort and across a range of FLDs. A second line of enquiry was exploration of the utility of serial CT as a prognosticator in IPF patients. Visual scores for these investigations were provided by four experienced sub-specialty thoracic radiologists.

Baseline CT analysis in IPF demonstrated that CALIPER parameters, particularly the percentage of the lung occupied by vessels (parenchymal vascular percentage - PVP), were strong predictors of pulmonary function indices as well as mortality. CALIPER variables were the strongest predictors of mortality in patients with hypersensitivity pneumonitis and connective tissue disease-related-ILD and were one of the strongest predictors of mortality in an “all-comers” cohort of FLD.

Serial evaluation of CT scans demonstrated that change in CALIPER variables, specifically change in PVP, was a stronger and more sensitive predictor of survival than the best available measure of disease worsening in IPF, namely change in forced vital capacity (FVC). It is concluded that CALIPER is a viable complementary tool in the baseline and serial evaluation of IPF and the baseline analysis of the majority of FLDs. Change in PVP, in particular, may represent a new index in the clinical evaluation of disease progression in IPF and could represent a co-endpoint alongside FVC change in clinical trials.