Evaluation of classifications of the monopodial bronchopulmonary vasculature using clustering methods

Mammalian pulmonary arteries divide multiple times before reaching the vast capillary network of the alveoli. Morphological analyses of the arterial branches can be challenging because more proximal branches are likely biologically distinct from more peripheral parts. Thus, it is useful to group the arterial branches into groups of coherent biology. While the generational approach of dichotomous branching is straightforward, the grouping of arterial branches in the asymmetrically branching monopodial lung is less clear. Several established classification methods return highly dissimilar groupings when employed on the same organ. Here, we established a workflow allowing the quantification of grouping results for the monopodial lung and tested various methods to group the branches of the arterial tree into coherent groups. A mouse lung was imaged by synchrotron x-ray microcomputed tomography, and the arteries were digitally segmented. The arterial tree was divided into its individual segments, morphological properties were assessed from corresponding light microscopic scans, and different grouping methods were employed, such as (fractal) generation or (Strahler) order. The results were ranked by the morphological similarity within and dissimilarity between the resulting groups. Additionally, a method from the mathematical field of cluster analysis was employed for creating a reference classification. In conclusion, there were significant differences in method performance. The Strahler order was significantly superior to the generation system commonly used to classify human lung structure. Furthermore, a clustering approach indicated more precise ways to classify the monopodial lung vasculature exist.