Introduction:
Delays identifying progressive pulmonary fibrosis (PPF) worsen patient outcomes, highlighting the need for tools to define progression earlier.
Aims:
To evaluate whether an AI imaging software that quantifies CT changes consistent with interstitial lung disease (ILD) would have supported the earlier identification of PPF.
Methods:
The Retrospective EValuation in the US of e-Lung in PPF (REVISE PPF) project is a multicentre observational study. University of Chicago [Chicago], Weill Cornell Medical Center [Cornell], and University of Alabama [UAB] contributed data. Serial HRCT scans were processed with an FDA-cleared AI software (e-Lung, Brainomix, Oxford, UK). Independently derived e-Lung biomarker thresholds of disease progression were assessed: ≥10% decline in e-Lung volume, ≥3% increase in weighted reticulovascular score, and ≥1.5% rise in total disease extent. The earliest radiologic progression detected by e-Lung was compared to the timing of guideline-defined clinical PPF diagnosis.
Results:
265 patients with PPF were evaluated (Chicago n=150, Cornell n=71, UAB n=44) with autoimmune ILD (45%) and hypersensitivity pneumonitis (32%) the commonest diagnoses. e-Lung identified progression earlier than standard-of-care clinical diagnoses in 73 (49%) patients at Chicago, 55 (77%) at Cornell, and 36 (82%) at UAB, with a mean (SD) 212 (264), 626 (761), and 322 (379) days, and median lead time of 114, 412 and 169 days respectively.
Conclusions:
In this retrospective, multi-institutional cohort, use of e-Lung could have accelerated the diagnosis of PPF for between 49% and 82% of patients up to an average of 21 months earlier than current clinical pathways.
