Performance Evaluation and Validation of e-Lung in non-IPF fibrotic ILD

<bold>Background:</bold> Brainomix 360 e-Lung, CT-based, AI powered biomarkers are associated with outcomes in IPF. Their performance in non-IPF fibrotic ILD (fILD) requires further evaluation. <bold>Aim:</bold> To evaluate e-Lung in non-IPF fILD and validate prognostic thresholds in independent cohorts. <bold>Methods:</bold> In the test cohort, consecutive patients with non-IPF fILD from the Open Source Imaging Consortium were included if they had a CT scan within 12 weeks of baseline lung function. The e-Lung weighted reticulovascular score (WRVS) biomarker was tested for associations with mortality and FVC decline with an optimal WRVS threshold defined. Analyses were validated in a cohort of patients prospectively recruited to the CoWorker fILD registry. Regression analyses, hazard ratios (HR) & 95% confidence intervals are reported. <bold>Results:</bold> The test cohort comprised 289 patients, mean age 65, f=135(45%). Adjusting for age, sex and FVC, WRVS was associated with mortality (HR 1.10,1.05-1.14, p<0.001) and trend to 10% FVC decline (HR 1.09, 0.99-1.22,p=0.089). The optimal WRVS threshold was found to be 10% and adjusting for the same covariates, was associated with mortality (HR 2.24, 1.35-3.72, p=0.002) and FVC decline (HR 3.68, 1.50-9.62, p=0.006). The validation cohort comprised 347 patients, mean age 67, f=155(45%). Adjusted WRVS was associated with mortality (HR 1.15, 1.10-1.19, p<0.001) and 10% FVC decline (HR 1.07, 1.02-1.11, p=0.002). The pre-defined 10% WRVS threshold was associated with mortality (HR 2.54, 1.70-3.81, p<0.001) and FVC decline (HR 1.60, 1.07-2.37, p=0.02). <bold>Conclusion:</bold> This 2-cohort validation study of patients with non-IPF fILD demonstrates that e-Lung can prognosticate from a baseline CT scan with potential clinical applicability.