Abstract Rational Individuals with chronic obstructive pulmonary disease (COPD) and emphysema have a higher risk of developing non-small cell lung cancer (NSCLC), even after accounting for smoking exposure. Chronic airway and parenchymal injury create a chronic pro-inflammatory microenvironment that facilitates carcinogenesis. Despite this, only about 10% of individuals with COPD develop lung cancer during their lifetime. In this study, we analyzed the COPDGene Phase 1 to identify pulmonary nodules and determine which clinical and imaging characteristics are associated with the presence of nodules 6 mm in diameter. Method A residual encoder Unet was trained for nodule detection and segmentation on publicly available datasets including COVID-19 CT Lung, Lung Image Database Consortium (LIDC), Medical Segmentation Decathlon Lung, and NSCLC-Radiomics. The model was trained for 1000 epochs, using a batch size of 2, and patch-size of (112,256,256). After training, the model was applied to CT scans from COPDGene Phase 1 dataset (n = 9419) to automatically extract pulmonary nodules. The largest AI detected nodule was recorded, and participants were stratified into three groups based on nodule diameter: 6 mm, 6 mm, and no nodules. Univariate analysis was performed non-parametric statistics. Results Among 9419 COPDGene Phase 1 participants, 1710 (18.15%) had no pulmonary nodules, 1455 (15.44%) had nodules 6 mm, and 6254 (65.39%) had nodules ≥ 6 mm, seen in table 1. Participants with nodules ≥ 6 mm were significantly older (median 60.5 years) and had a greater smoking exposure (40.5 pack-years) compared with those without nodules (55.8 years; 36.3 pack-years; p 0.001). Emphysema and gas trapping were more pronounced in participants with nodules ≥ 6 mm (median %emphysema 2.4; %gas-trapping 15.5) than in those without nodules (1.5 and 11.5, respectively; p 0.001). These individuals had higher mMRC dyspnea scores (≥2 in 42.33% vs. 37.26%) and elevated SGRQ total scores (median 23.1 vs. 18.5; p 0.001), exhibiting worse functional limitation. Increasing GOLD stage and BODE index severity were both associated with higher prevalence of nodules ≥ 6 mm (trend p 0.001). Conclusion This study represents the first characterization of AI detected pulmonary nodules in the COPDGene cohort. The prevalence of CT-detected nodules ≥6 mm was associated with older age, greater smoking exposure, and higher emphysema burden. These findings highlight potential risk factors for nodule development and may inform future risk stratification strategies for lung cancer. Funding NCI K00CA27471, and NHLBI grants 1R01HL149877, U01HL089897, U01HL089856, and NIH contract 75N92023D00011. This abstract is funded by: This work was supported by NCI K00CA27471, NHLBI grants 1R01HL149877, U01 HL089897, and U01 HL089856 and by NIH contract 75N92023D00011”
Masquelin et al. (Fri,) studied this question.