Deep Learning-based Segmentation of CT Scans Predicts Disease Progression and Mortality in IPF

Am J Respir Crit Care Med. 2024 Mar 7. doi: 10.1164/rccm.202311-2185OC. Online ahead of print.


Rationale: Despite evidence demonstrating a prognostic role for CT scans in IPF, image-based biomarkers are not routinely used in clinical practice or trials.

Objectives: Develop automated imaging biomarkers using deep learning based segmentation of CT scans.

Methods: We developed segmentation processes for four anatomical biomarkers which were applied to a unique cohort of treatment-naive IPF patients enrolled in the PROFILE study and tested against a further UK cohort. The relationship between CT biomarkers, lung function, disease progression and mortality were assessed.

Measurements and main results: Data was analysed from 446 PROFILE patients. Median follow-up was 39.1 months (IQR 18.1-66.4) with cumulative incidence of death of 277 over 5 years (62.1%). Segmentation was successful on 97.8% of all scans, across multiple imaging vendors at slice thicknesses 0.5-5mm. Of 4 segmentations, lung volume showed strongest correlation with FVC (r=0.82, p<0.001). Lung, vascular and fibrosis volumes were consistently associated across cohorts with differential five-year survival, which persisted after adjustment for baseline GAP score. Lower lung volume (HR 0.98, CI 0.96-0.99, p=0.001), increased vascular volume (HR 1.30, CI 1.12-1.51, p=0.001) and increased fibrosis volume (HR 1.17, CI 1.12-1.22, p=<0.001) were associated with reduced two-year progression-free survival in the pooled PROFILE cohort. Longitudinally, decreasing lung volume (HR 3.41; 95% CI 1.36-8.54; p=0.009) and increasing fibrosis volume (HR 2.23; 95% CI 1.22-4.08; p=0.009) were associated with differential survival.

Conclusions: Automated models can rapidly segment IPF CT scans, providing prognostic near and long-term information, which could be used in routine clinical practice or as key trial endpoints. This article is open access and distributed under the terms of the Creative Commons Attribution 4.0 International License (

Keywords: IPF; Machine Learning.