Abstract
IPF is a progressive fibrotic lung disease. We previously used automated machine learning to measure airway/lung volume (1a) to correlate with short term disease progression (Roberts et al ERS 2021). Here we expand that work presenting analysis of 93 patients with data up to 7 years. Patients were split into cohorts with a multivariate approach of 5 variables: survival time, annualised changes in FVC, FEV1, DLCO, and alveolar volume. Individual plots of 2 variables were drawn e.g. FVC and DLCO (1b). Principal Component Analysis and unsupervised machine learning (K-means) of all 5 variables identified two distinct clusters (33 progressive and 60 stable). Application of Lung8 algorithm to serial CTs on the Qureight platform showed significant difference in lung volume decline between progressive and stable cohorts (-0.266 vs -0.133L/year p=0.00716) (1c,d). Introducing GAP score to our model resulted in marginal sensitivity and only 5% change in cohort classification. We benchmarked our multivariate approach against </>10% annualised FVC drop only, which identified 80 progressive patients and corresponded with Lung8 just outside significance (progressive -0.179 vs stable 0.089L/year p=0.0506). Multivariate cohorting in tandem with Lung8 constitutes a novel proxy for identifying phenotypes of IPF progression and may be superior to use of FVC alone.
Footnotes
Cite this article as Eur Respir J 2022; 60: Suppl. 66, 4502.
This article was presented at the 2022 ERS International Congress, in session “-”.
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 2022
