Vessel Image Segmentation
My research focuses on the image segmentation of Computed Tomography Angiography (CTA) images to extract important blood vessels such as coronary ateries, where atherosclerotic plaques often occur. The segmentation algorithm combines Bayesian theory and active contour models to effectively find the inner wall of the vessels and eliminate unimportant tissues. A 3D model of coronary arteries can be reconstructed based on the segmentation result. The model can serve as boundary conditions for the Computational Fluid Dynamics (CFD) simulation of the blood flow, or assist the localization of plqaues for the diagnosis of atherosclerosis. More research is being conducted to quantify the narrowing of the vessel lumen and identify soft plaques that do not have a high contrast to other tissues in CTA images. To verify the segmentation results, we plan to make vessel casts from ex-planted hearts and compare with the 3D model we obtain on the computer.
Segmentation of the left main (LM) coronary artery. (a)The original gray scaled image; (b) Classification result with labels; (c) Blood filled region marked with boundaries; (d) Segmented LM with marked boundaries.
3D surface rendering of left and right coronary arteries with the aorta. Red spots indicate calcium deposits in the vessel wall, and the narrowing of the vessel around the red spots indicate the existance of soft plaques.