Morphological Features of Murine Pulmonary Arterial Networks for Diagnosing Vascular Disease

Abstract

Hypoxia-induced pulmonary hypertension results from prolonged time spent in a low-oxygen environment. The induced pulmonary hypertension leads to vascular remodeling, including vessel dilation and decreased compliance. This study aims to identify changes in vascular morphology in mice with pulmonary hypertension induced by hypoxia. To do so, we examine the pulmonary arterial network morphology using methods from geometry, topological data analysis, and statistical shape analysis. Tools from these fields are applied to arterial networks excised from micro-CT images of thirteen mice at four pressures. Networks from the hypertensive mice were normalized via a radius pruning algorithm, and analysis was conducted on the entire arterial network, as well as on a per-lobe basis. The main objective of this study was to present an adaptable pipeline for extracting features from micro-CT images to study how vascular disease affects arterial network morphology. Using this pipeline, we showed that the hypertensive mice have increased radius ratios and decreased length-to-radius ratios. Persistent homology analysis revealed that hypertension leads to lower branch-to-branch proximity and lower directional complexity in both the posterior and anterior directions. Persistence curves, particularly Betti curves, revealed lobe-specific differences in the left and inferior lobes. The first two shape principal component scores were also able to discriminate between control and hypertension at lower pressures. Importance metrics for each feature are computed using a boosted decision tree, and then an SVM classification model is fit with the most informative predictors for hypertensive classification.