Quantitative chest CT assessment of pulmonary alveolar proteinosis with deep learning: a real-world longitudinal...

High-resolution computed tomography (HRCT) plays an important role in accessing the severity of pulmonary alveolar proteinosis (PAP). Visual evaluation of changes between two HRCT scans is subjective. This study was conducted to quantitatively evaluate lung burden changes in patients with PAP using HRCT-based automated deep-learning method following 12 months of statin therapy.

In this prospective real-world observational study, patients with PAP who underwent chest HRCT were evaluated from November 28, 2018, to April 12, 2021. Oral statin administration was initiated as therapy for these PAP patients with 12 months of follow-up. HRCT-derived lung ground-glass opacification percentage of the whole lung and 5 lobes and the percentage of different densities of ground glass were automatically quantified with deep-learning software. Longitudinal changes of the HRCT quantitative parameter were also compared.

The study enrolled 50 patients with PAP, including 25 mild-moderate PAP cases and 25 severe PAP cases. The percentage of lung ground-glass opacification of the whole lung and 5 lobes and the percentage of different densities of ground glass were significantly different among the 2 different clinical types at baseline (all P values <0.05). Overall, the percentage of whole-lung ground-glass opacification significantly decreased between the baseline HRCT and the HRCT results after 12 months of follow-up (P=0.023; 95% CI: 1.384-18.684). Changes in the total opacification of the whole lung were positively correlated with changes in partial pressure of arterial oxygen (PaO₂; r=0.716; P<0.001) and percentage of predicted diffusion capacity for carbon monoxide (DLCO%pred; r=0.664; P<0.001).

A quantitative image parameter automatically generated by a deep-learning tool from chest HRCT scans may be used to evaluate the severity of PAP and may help to evaluate and quantify the response to statin therapy.