This paper evaluates the relationship between changes in CT quantification of interstitial lung disease (ILD) using deep learning, specifically the AVIEW Lung Texture software, and their correlation with changes in forced vital capacity (FVC) and visual assessments of ILD progression. The study included 468 ILD patients, using deep learning-based texture analysis for segmenting ILD findings in CT images. Patients were grouped based on their absolute decline in predicted FVC and visual assessments of ILD progression by thoracic radiologists.
The study found significant increases in fibrosis and total ILD extent in patients with larger FVC declines. Those with ILD progression exhibited higher increases in these parameters. Notably, increases in fibrosis and total ILD extent were significant prognostic factors for survival, particularly when adjusted for FVC declines of ≥ 5% and ≥ 10%.
The research concludes that changes in ILD CT quantification correlate with FVC changes and visual ILD progression assessments, serving as independent prognostic factors in ILD patients. This emphasizes the potential role of deep learning-based CT quantification in diagnosing and prognosticating progressive fibrosing ILD.