Performance comparison of e-Lung with densitometry in the CT evaluation of patients with IPF

<bold>Introduction:</bold> e-Lung (Brainomix, Oxford) is an AI based imaging module used to assess CT scans from patients with interstitial lung disease. An e-lung derived biomarker, the reticulovascular score (RVS) quantifies reticular and vascular structures within the lungs with the weighted RVS (WRVS) assigning increased importance to the periphery. An alternative approach is to quantify densitometry with high attenuation areas (HAA) defined as % lung between -250 and -600 Hounsfield Units. <bold>Aims:</bold> To compare the performance of e-Lung derived RVS and WRVS biomarkers with HAA% in patients with idiopathic pulmonary fibrosis (IPF)(Figure 1) <fig><object-id>erj;64/suppl_68/OA4556/F1</object-id><object-id>F1</object-id><object-id>F1</object-id><graphic></graphic></fig> <bold>Methods:</bold> Consecutive patients with IPF from the Open Source Imaging Consortium were studied. Relationships between median baseline imaging biomarkers and survival were evaluated using Cox regression with hazard ratios (HR) and 95% confidence intervals (CI) presented. <bold>Results:</bold> Of 344 patients included, 290 (84%) were male, mean age 69, mean forced vital capacity 2.8L (81%). Baseline WRVS% was most strongly associated with transplant free survival with a HR of 4.55 (3.05-6.78), p<0.001. Baseline RVS% had a HR of 3.50 (2.21-5.56), p<0.001. An analysis of the association between HAA% and transplant free survival showed a HR of 2.60 (1.27–5.32), p<0.001. <bold>Conclusions:</bold> In a diverse cohort of IPF patients, the prognostic performance of e-Lung WRVS and RVS biomarkers from a baseline CT scan outperformed HAA.