Deep learning analysis of serial CT scans correlates with changes in how patients feel, function and survive in fibrotic interstitial lung disease

Résumé fourni par la source

Background: Monitoring disease progression in fibrotic ILD is commonly performed using pulmonary function tests (PFTs). Deep learning image analysis of CT scans may reveal additional disease progression and prognostic information vital to personalising patient care. Aims: We aimed to prospectively assess the role of serial CT scans in the monitoring of fibrotic ILD. Methods: Patients with fibrotic ILD (idiopathic pulmonary fibrosis (IPF) and non-IPF) were recruited to a prospective observational study. Participants underwent deep learning image analysis on the Qureight platform alongside PFTs and patient reported outcome measures (PROMS) (K-BILD and EQ-5D-3L) at 0, 6 and 12 months. Pearson correlation (SPSS) examined correlation between variables and Cox regression analysis used for transplant-free survival (TFS). Results: 18 participants with IPF and 18 participants with non-IPF fibrosis were recruited (mean age 71.81 +/- SD 7.1, 25 male, 69.4%). 30/36 completed 12 months follow up (5 deaths, 1 drop out). At baseline, Airway Volume% (Air8) correlated with EQ-5D-3L (-0.41, p=0.005), K-BILD -0.38,p=0.025), FVC% (-0.56, p<0.001) and TLCO% (-0.46, p=0.008). 12 month change in Air8% correlated with change in FVC (-0.41 p=0.03) and K-BILD (-0.43, p=0.03). Change in FVC did not correlate with changes in K-BILD (0.72, p=0.74). Baseline Fibrosis volume% (Fibr8) was associated with 12-month TFS independent of age and gender (HR 1.13, p=0.021). Conclusions: Deep learning image analysis in a prospective, observational non-interventional correlates with PFTs and PROMs. These methods may provide an additional monitoring option in patients with fibrotic ILD.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.