Animal models provide unique in-vivo insight for pathophysiology understanding and pharmacological development. The automatic assessment of the pulmonary vasculature in animal models has not been widely demonstrated. Variable imaging quality and anatomy size and variability across species complicate the development of general methods and the automated analysis. Measures of blood volume using scale-space particles have been shown in humans to function as a potential biomarker for pulmonary vascular disease. Here, we will explore the utility of our quantitative pulmonary vascular platform in animal models.
Pulmonary vasculature was extracted using the scale-space particle approach that we have developed and used extensively in human epidemiological and clinical studies. Briefly, candidate vascular locations were detected using a deep neural network filter to initialize the scale-space particle algorithm. Scale-space particles were used to grow a topology-free cloud of equally distance points loosely coupled across space and scale arrange along the vascular tree centerline. Scale-space particles were tailored to accommodate the scale and density contrast specific to each animal model and imaging protocol. Vascular size for each particle was regressed using a deep learning approach based on a synthetic vessel model. Quantitative analysis was performed by assessing the distribution of vascular volume as a function of the vascular cross-section area.
CT images for a sheep and a mouse were acquired. The image resolution for the sheep and the mouse was 0.6 mm3 and 0.1 mm3, respectively. In both models, large and small intraparenchymal vessels were detected (Figure 1 - see full pdf). The total blood volume was 59.781 mL and 0.768 mL for the mouse and the sheep, respectively.
Quantitative pulmonary vascular imaging is feasible in large and small animal models using scale-space particles. This study demonstrates the versatility of the geometric approach inherent in the scale-space particle approach to be used in different species.