Cigarette smoke is a major global health concern due to its detrimental effects on the respiratory system. The respiratory tract is directly exposed to harmful particles and chemicals present in tobacco smoke. We hypothesize that lung resilience depends on the airway submucosal gland (SMG), an ectodermal appendage lining the airways, renowned for its unique properties in mucus secretion, antimicrobial activity and a potential niche for proximal airway stem cells. The objective of this study is to assess the relationship between SMG size and structure and cigarette smoke exposure, using an artificial intelligence (AI)-based approach. The analysis is based on hematoxylin and eosin (H& E)-stained sections from 278 non-diseased bronchial specimens, sampled from non-diseased regions and obtained from the “CaProMat” cohort, which includes lung cancer patients with characterized occupational and tobacco smoke exposure.

We developed a supervised AI-based approach to automatically segment SMG ducts and acini, airway epithelium, bronchial smooth muscle and cartilage structures from H& E-stained slides. The model was trained on a dataset of 29 annotated images. Performance evaluation showed promising results, with all target structures achieving mean Dice similarity coefficients exceeding 0.8.

AI-powered analysis of H& E-stained sections is a promising tool for identifying tissue-relevant features and exploring the relationship between spatial tissue patterns and particulate exposure. Our preliminary findings show that SMG area is larger in former smokers compared to non-smokers, with a further increase observed in active smokers. These results suggest that SMGs undergo adaptive changes in response to particulate exposure, which may be partially reversible after exposure cessation.