Informatics and Data Science

Identification of Radiological Progressive Pulmonary Fibrosis in Clinically Stable Patients Using e-Lung Quantitative CT Algorithm: Results from the REVISE PPF Study

Location: Concourse Foyer Kiosk

First Author(s)

TK

Tejaswini Kulkarni, MD

Assistant Professor
University of Alabama, Birmingham, USA, United States

Co-Author(s)

AP

Anna J. Podolanczuk, MD

Assistant Professor
Weill Cornell Medicine, New York, USA, United States
KB

Kristin Berger, MD

Instructor in Medicine
University of Alabama, Birmingham, USA, United States
YR

Yana Rabkova, MSc

Research Coordinator
Weill Cornell Medical College, New York, USA, United States
SR

Steven S Rothenberg, MD

Assistant Professor
Rocky Mountain Hospital for Children, Denver, USA, United States
CR

Christian Rennison-Jones, MSc

Medical Imaging Researcher and Life Sciences Project Manager
Brainomix, Oxford, UK, United Kingdom
GH

George Harston, DPhil

Chief Medical and Innovation Officer
Brainomix, Oxford, UK, United Kingdom
AD

Anand Devaraj, MD

Professor
Brainomix, Royal Brompton Hospital, and National Heart and Lung Institute, Imperial College London, UK, United Kingdom
AA

Ayodeji Adegunsoye, MD

Assistant Professor
University of Chicago, Chicago, USA, United States
PG

Peter M George, MBBS

Senior Medical Director
Brainomix, Royal Brompton Hospital, and National Heart and Lung Institute, Imperial College London, London, UK, United Kingdom

Purpose: Identification of progressive pulmonary fibrosis (PPF) relies on patients meeting criteria based on pulmonary function tests (PFTs) and CT. PFTs are subject to variability and the interpretation of CT scans is subjective. e-Lung (Brainomix) is an AI-driven CT platform that quantifies the extent of interstitial lung disease (ILD) on CT. We evaluated whether e-Lung could identify radiologic evidence of PPF in patients deemed clinically stable.

Methods/Materials: The Retrospective EValuation In the US of e-Lung in PPF (REVISE PPF) project is a multicenter retrospective observational study. Three US ILD centres (University of Alabama [UAB], University of Chicago, Weill Cornell Medicine) contributed clinical, lung function and serial imaging data from patients who were considered clinically stable, i.e. did not meet either ATS/ERS/JRS/ALAT guideline criteria or INBUILD trial criteria for PPF. e-Lung processed all available CT scans. e-Lung measures of radiologic progression were defined as: relative decline in e-Lung volume ≥10%, increase in weighted reticulovascular score (WRVS) ≥3%, or rise in total disease extent ≥1.5%.

Results: A total of 253 patients without PPF had serial CT scans evaluated (UAB n=76, Chicago n=122, Cornell n=55). The most frequent diagnosis was autoimmune ILD (UAB, 66%; Chicago, 46%; Cornell, 49%). Applying any one of the three e-Lung measures of radiologic progression, 48 patients (63%) from UAB, 118 patients (97%) from Chicago and 30 patients (55%) from Cornell had CT evidence of PPF. Whilst in the Chicago cohort, a rise of WRVS ≥3% was the most sensitive measure of progression based on e-Lung, in the UAB and Cornell cohorts, a rise of TDE ≥1.5% captured the most patients with progression.

Conclusions: e-Lung identified occult radiologic evidence of PPF in 55-97% of patients deemed clinically stable across three US ILD centers. These data suggest that applying e-Lung to routinely acquired CT scans could “upstage” patients from stable disease to PPF. Further work will establish associations with long-term outcomes in this group of patients.