Abstract Rationale Pulmonary vascular injury plays a key role in the development and progression of COPD. Perfusion abnormalities may reflect early vascular remodeling and gas exchange defects, but reliance on contrast-enhanced imaging limits population-level assessment. Non-contrast-based perfusion imaging biomarkers could enable earlier risk stratification of smokers at greater risk for COPD progression. Methods We developed a generative deep learning approach to estimate lung perfusion from non-contrast CT. Perfusion maps and corresponding virtual non-contrast images were generated from dual-energy CTs to train a generative AI model to estimate iodine maps from non-enhanced inspiratory scans. Pulmonary blood volume (PBV) images were obtained by normalizing values to the iodine concentration in the main pulmonary artery. For each subject, the lungs were divided into 130mm3 regions, and regional PBV values were averaged to compute the overall mean PBV and PBV heterogeneity measured as the coefficient of variation (CV) of regional PBVs. Disease progression was categorized as fast or slow by adjusted lung density (ALD) decline (1 mg/L/year), forced expiratory volume in 1 second (FEV1) decline (-35 mL/year), FEV1/forced vital capacity (FEV1/FVC) ratio decline (0.0), incident COPD decline (GOLD stage ≥1), or increasing gas trapping (0%) over five years. Multivariable logistic regression was used to analyze associations of mean PBV and PBV-CV tertiles with progression status adjusting for gender, age, race, smoking status, pack-years, body mass index (BMI), FEV1, FVC, emphysema percentage (pctEmph), small-vessel arterial volume (BV5), gas trapping, airway wall thickness (Pi10), and total lung capacity (TLC). Results Non-contrast CT scans from 2,461 pre-COPD participants in COPDGene (mean age: 57.8±8.3 years; 1,163 men, 1,298 women, 1,995 GOLD 0, 466 PRISM) were analyzed and grouped into tertiles. After multivariable adjustment, the lowest mean PBV tertile was associated with higher risk of rapid ALD decline (OR: 1.44, 95%CI: 1.08-1.92), FEV1 decline (OR: 1.38, 95%CI: 1.05-1.83), accelerated FEV1/FVC decline (OR: 1.56, 95%CI: 1.17-2.06), and increased gas trapping (OR: 2.53, 95%CI: 1.86-3.45), with a non-significant association for incident COPD (OR: 1.21, 95%CI: 0.80-1.84). The PBV-CV tertiles showed similar associations across outcomes except for FEV1 decline, where CV was not statistically significant (p 0.05), indicating that perfusion heterogeneity measures provide complimentary information relevant to most progression markers. Conclusions Low non-contrast PBV independently predicted multiple markers of disease progression, including accelerated functional decline and worsening gas trapping. Automated perfusion estimation from non-contrast CT may provide sensitive, scalable biomarkers of early vascular injury, improving risk stratification among smokers at risk for COPD progression. This abstract is funded by: NHLBI awards 5K25HL157601, 1R21HL177703, 1R01HL149877, U01 HL089897, and U01 HL089856, as well as NIH contract 75N92023D00011
Nardelli et al. (Fri,) studied this question.