Risk factors and prognostic indicators for progressive fibrosing interstitial lung disease: a deep learning-based CT quantification approach

This study evaluated the utility of deep learning-based quantitative CT (QCT) in predicting progressive fibrosing interstitial lung disease (PF-ILD) and assessing prognosis in ILD patients. A total of 465 patients from Seoul National University Hospital were retrospectively analyzed. Using AVIEW Lung Texture software, key ILD features such as ground-glass opacity (GGO), reticular opacity (RO), and honeycombing were quantified. Baseline RO and fibrosis extent were significant predictors of PF-ILD, while RO, honeycombing, and fibrosis extent were associated with all-cause mortality. The Cox regression model that incorporated both baseline FVC and QCT outperformed the model using FVC alone in predicting mortality. One-year follow-up CT data further confirmed that increases in RO and fibrosis extent were associated with poor outcomes.