Computed Tomography Cluster-informed Subject-Specific Assessment of Airway Resistance and Particle Deposition in Post-COVID-19 Lungs

Patients who recovered from the severe acute respiratory syndrome coronavirus (SARS-CoV-2 or COVID-19) may have long-term symptoms, given the diagnosis long COVID or post-acute sequelae of COVID-19 (PASC). In this study, we employed a cluster-informed (or guided) strategy that utilized contrastive learning and K-means to identify post-COVID-19 clusters from computed tomography (CT) images, and applied a CT imaging-based subject-specific multi-scale whole-lung computational fluid and particle dynamics (CFPD) model to investigate fluid dynamics within clusters for assessment of disease risk or therapeutic response. 140 post-COVID-19 subjects and 105 healthy controls were analyzed. The average time between COVID-19 diagnosis and CT acquisition was 113 days. Two clusters were identified, characterized by small airways disease (cluster 1, C1) and fibrotic-like lung patterns (cluster 2, C2), respectively. C1 had increased lobar resistance during tidal breathing due to airway narrowing, while C2 exhibited decreased lobar resistance due to airway-associated interstitial lungs. Compared to healthy controls, the whole-lung deposition in C2 during tidal breathing was reduced by 2% (p < 0.05) for particles ranging from 0.01 to 10.0 μm.