Unsupervised comprehensive CT imaging clusters reveal distinct morphological phenotypes in asthma: insights from two prospective cohorts.

BACKGROUND

Asthma is a heterogeneous disease with diverse underlying structural abnormalities. Comprehensive evaluation of computed tomography (CT)-derived features may uncover novel morphological phenotypes with clinical relevance. This study aimed to identify and validate the morphological phenotypes of asthma using comprehensive inspiratory and expiratory CT features and assess their associations with clinical outcomes.

METHODS

We analyzed two independent prospective asthma cohorts (Kyoto asthma cohort, n = 223; PACTAS study, n = 94) and a cross-sectional healthy control cohort (n = 88). Quantitative CT measures—airway morphology, mucus plugs, parenchyma, extrapulmonary structures, and pulmonary arteries—were quantified. Unsupervised clustering using these parameters was performed separately in each asthma cohort. Associations with lung function, type 2 inflammation, symptom burden, and exacerbations were examined.

RESULTS

Four reproducible morphological phenotypes were identified. Cluster 1 (airway remodeling) was characterized by reduced airway lumen area (LA), and lower total airway count (TAC), along with higher mucus plug scores and elevated small airway dysfunction (SAD%) measured from inspiratory and expiratory lung density. Patients were older, had a longer asthma duration, and exhibited a lower forced expiratory volume in 1 s to forced vital capacity ratio (FEV₁/FVC). Cluster 2 (airway dilatation) showed enlarged LA, higher modified Reiff scores, and increased TAC. Patients tended to be younger, with a shorter disease duration and preserved lung function. Cluster 3 (metabolic abnormality) was characterized by reduced pectoralis muscle density and higher body mass index, and included predominantly female patients. This cluster was associated with a lower %FVC and greater symptom burden. Cluster 4 (parenchymal-dominant) consisted mainly of male smokers with emphysematous changes and higher SAD%, who exhibited lower FEV₁/FVC. Across both cohorts, Clusters 2–4 had exacerbation rates similar to or higher than in Cluster 1 in both unadjusted and adjusted analyses.

CONCLUSIONS

Comprehensive CT-based clustering revealed four reproducible morphological phenotypes of asthma—airway remodeling, airway dilatation, metabolic abnormality, and parenchymal-dominant—each associated with distinct structural and clinical features. These findings underscore the morphological heterogeneity of asthma and suggest that CT-based phenotyping may support more precise phenotype-specific treatment strategies.