An Automated System for 3D Segmentation of CT Angiograms - Visual Computing: Scientific Visualization and Imaging Systems - page 94

Chapter 5
An Automated System for 3D
Segmentation of CT Angiograms

Y. Wang and P. Liatsis

Abstract This chapter presents a novel automated two-step algorithm for
segmentation of the entire arterial tree in 3D contrast-enhanced Computed
Tomography Angiography (CTA) datasets. In the first stage of the proposed
algorithm, the main branches of the coronary arteries are extracted from the
volume datasets based on a generalised active contour model by utilising both
local and global intensity features. The use of local regional information allows for
accommodating uneven brightness distribution across the image. The global
energy term, derived from the histogram distribution of the input images, is used to
deform the contour towards to desired boundaries without being trapped in local
stationary points. Possible outliers, such as kissing vessel artefacts, are removed in
the following stage by the proposed slice-by-slice correction algorithm. Experi-
mental results on real clinical datasets have shown that our method is able to
extract the major branches of the coronaries with an average distance of 0.7 voxels
to the manually defined reference data. Furthermore, in the presence of kissing
vessel artefacts, the outer surface of the coronary tree extracted by the proposed
system is smooth and contains less erroneous segmentation as compared to the
initial segmentation.

Keywords Computed tomography angiography 3D segmentation Coronary
arteries Active contour models