Abstract
Background: HRCT is a cornerstone for the diagnosis and monitoring of ILD. Recently, radiomics has emerged as a promising research field, which automatically extracts mineable quantitative data from medical images, aiming to identify imaging biomarkers.
Aims: To evaluate 1) the potential of CT-based radiomics features for staging of experimental ILD and 2) the transferability of radiomics signatures to human ILD.
Methods: Radiomics analysis was performed on HRCT scans from bleomycin (BLM)-treated mice and NaCl controls at days 3 and 14 after intratracheal BLM instillation (n=5 mice/group), and on human HRCT scans (mild ILD=20, advanced ILD=46). In total, 154 radiomics features were calculated on segmented lungs using the in-house developed software Z-Rad. A univariate logistic model was built for each feature to discriminate between mild and advanced fibrosis.
Results: Radiomics features with good predictive performance (area under the receiver operating characteristic curve (AUC) >0.7 and p-value<0.05) were considered as candidate discriminators. Under this criterion, we identified 52/154 radiomics features, describing lung texture, that differentiated mild versus severe BLM-induced fibrosis. Notably, 51/52 (98%) of these staging predictors were successfully validated in ILD patients with an average AUC of 0.929 (SD=0.082).
Conclusions: This proof-of-concept study suggests that 1) the well-established mouse model of BLM-induced ILD is a valuable model system to test defined hypotheses in radiomics research and 2) that radiomics features show great potential as quantitative imaging biomarkers for staging of ILD.
Footnotes
Cite this article as: European Respiratory Journal 2019; 54: Suppl. 63, PA4806.
This is an ERS International Congress abstract. No full-text version is available. Further material to accompany this abstract may be available at www.ers-education.org (ERS member access only).
- Copyright ©the authors 2019