A semiautomatic approach for segmentation of carotid vasculature from patients’ CTA images

Segmentation of vasculature specific to the patients’ carotid vasculature is a complicated and challenging task because of its complex geometrical structure and interconnections. Accurate or approximate digital phantoms of the vasculature are extremely useful in quick analysis of the vascular geometry and the modelling of blood flow in the cerebrovasculature. All these analyses lead to effective diagnosis and detection/localization of the diseased arterial segment in the cerebrovasculature. In this work, we have proposed a semiautomatic geodesic path propagation algorithm based on fuzzy distance transform to generate digital cerebrovascular phantoms from the patients’ CT angiogram (CTA) images. We have also custom-developed a 2-D/3-D user interface for accurate placement of user-specified seeds on the input images. The proposed method effectively separates the artery/vein regions from the soft bones in the overlapping intensity regions using minimal human interaction. Qualitative results along with 3-D rendition of the segmented cerebrovasculature on eight patients’ CTA images are presented here.