Distinguishing pulmonary arterial and venous (A/V) trees via in vivo imaging is a critical first step in the
quantification of vascular geometry for purposes of determining, for instance, pulmonary hypertension, detection of
pulmonary emboli and more. A multi-scale topo-morphologic opening algorithm has recently been introduced by us
separating A/V trees in pulmonary multiple-detector X-ray computed tomography (MDCT) images without contrast.
The method starts with two sets of seeds - one for each of A/V trees and combines fuzzy distance transform, fuzzy
connectivity, and morphologic reconstruction leading to multi-scale opening of two mutually fused structures while
preserving their continuity. The method locally determines the optimum morphological scale separating the two
structures. Here, a validation study is reported examining accuracy of the method using mathematically generated
phantoms with different levels of fuzziness, overlap, scale, resolution, noise, and geometric coupling and MDCT
images of pulmonary vessel casting of pigs. After exsanguinating the animal, a vessel cast was generated using
rapid-hardening methyl methacrylate compound with additional contrast by 10cc of Ethiodol in the arterial side
which was scanned in a MDCT scanner at 0.5mm slice thickness and 0.47mm in plane resolution. True
segmentations of A/V trees were computed from these images by thresholding. Subsequently, effects of
distinguishing A/V contrasts were eliminated and resulting images were used for A/V separation by our method.
Experimental results show that 92% - 98% accuracy is achieved using only one seed for each object in phantoms
while 94.4% accuracy is achieved in MDCT cast images using ten seeds for each of A/V trees.