Information Technology Reference
In-Depth Information
5.5 Conclusions and Future Work
Accurate segmentation of vascular structures is an essential and fundamental step
for various clinical tasks, such as stenosis grading and surgical planning. In this
chapter, we presented a computer vision system which contributes towards the
development of the proposed framework for segmentation of coronary arteries in
3D CTA images. In particular, a novel two-step algorithm was proposed to effi-
ciently segment coronary arteries from CT images by making use of both global
and local intensity statistics. The global energy was designed as a normalised CDF
based on the histogram of the input image data, which adapts its contribution to the
overall active contour energy by considering the spatially varying properties of the
artery. The kissing vessels were identified and tracked throughout the axial slices
in the second stage of the segmentation based on a multi-region competition
algorithm. Experimental results show that the proposed approach is able to cor-
rectly segment the major branches of the arterial tree, with an average voxel-wise
distance of 0.73 voxels to the manually delineated ground truth. Furthermore, in
the presence of kissing vessel artefacts, the overall performance of the segmen-
tation can be significantly improved by the slice-by-slice correction scheme. The
FP rate from these cross-sections containing kissing vessels was reduced from 294
to 22.8 %.
In terms of future research, we intend to introduce spatial/intensity correlations
between frames to the correction process, since the information extracted from
previous frames can be used to assist in segmentation of coronary arteries in the
subsequent frame. Based on the 3D surface reconstructed from the segmentation,
the geometric parameters along each of the main branches of the coronary arteries
can be precisely computed, which are very useful for diagnostic purposes, such as
stenosis detection and grading.
Acknowledgments The authors would like to acknowledge the support of City University,
which enabled this research through the award of a University Research Studentship and Dr
Gerry Carr-White and Rebecca Preston at St Thomas and Guys Hospitals for their invaluable
advice and the provision of the CTA datasets.
References
1. WHO. Coronary Heart Disease (2010)
2. Netter, F.H.: Atlas of Human Anatomy, 2nd edn. Rittenhouse Book Distributors Inc, King of
Prussia (1997)
3. Abe, T., Hirohata, M., Tanaka, N., et al.: Clinical benefits of totational 3D angiography in
endovascular treatment of ruptured cerbral aneurysm. Am. J. Neuroradiol. 23, 686-688 (2002)
4. Tillich, M., Hill, B.B., Paik, D.S., et al.: Prediction of aorotoiliac stent graft length:
comparison of measurement methods. Radiology 220(2), 475-483 (2001)
5. Philips. Brilliance iCT scanner.
http://www.healthcare.philips.com/us/products/ct/products/
ct_brilliance_ict/index.wpd
(2011)
Search WWH ::
Custom Search