Coronary artery segmentation via Hessian filter and curve-skeleton extraction

Abstract

Precise coronary artery segmentation is a prerequisite for quantitatively assessing the severity of coronary artery stenosis. Extracting the centre line of the 3D volumetric coronary artery tree, also named as 3D skeletonization, plays an important role in identify the variations of cross-sectional profile. Typically there are three skeletonization methods, viz. distance transformation, Voronoi method and topological thinning method. All these three skeletonization methods were applied in this study to extract the curve-skeleton of coronary arteries, after segmenting the coronary artery tree with Hessin filter. Among them, topological thinning method is recommended, as it produces reliable and accurate curve-skeleton for vessels with varying size. This will facilitate quantitative assessment of the severity of coronary artery stenosis, help clinical diagnosis and treatment planning of coronary artery disease.

Quantitative analysis of coronary artery from computed tomography angiography images

“…Van's methods were applied on four different cardiac CTA datasets in DICOM format (Figure 4.7). The Hessian filter method proposed in [107] was used to segment the complete 3D digital coronary artery trees. Table 4.…”

“…The approach has been applied to CTA datasets and it demonstrates very strong vessel enhancement and background noise suppression, which outperforms previous vessel filtering techniques. However, due to For complete coronary artery tree segmentation, the present approach is based on Hessian filter and connected components [107]. For different CTA datasets, all parameters and thresholds need to be tuned to fit the data, and the computational cost is high.…”

“…Finally, classical Marching Cubes algorithm [103] is used to reconstruct the 3D surface mesh of the vessel. More detailed information about the segmentation approach can be found in our previous work [107]. In this work, post-processing steps such as filling hole and surface smoothing are also performed to accurately determine the centerline.…”

“…In the literature, several centerline extraction approaches for CTA images have been developed [67,107,109,110], most of which can be divided into three different categories. In the first category, the approach is based on topological thinning methods [81].…”

“…The final centerline can be formed by connecting the set of discrete skeleton points with the minimum spanning tree algorithm. More detailed information about centerline extraction can be found in[107]. Both techniques are applied over the CTA dataset to extract the centerline.…”