The Extraction Methods Of Hepatic Vascular Centerline And Hierarchical Data

With the development of partial liver resection, especially living donor liver transplantation. Segmental anatomy of liver is becoming more and more important in operation. Centerline and hierarchical data as vascular structural description method, not only is the important basis of liver section, and describes the vascular anatomical position and series, also reflected the blood relationship, so accurate extraction liver vascular centerline and hierarchical data is a clinical diagnosis and analysis of an important part in the process.Due to the complexity of the liver vascular branch, and the segmentation result of liver blood vessels is not smooth, the traditional method is used to extract vessel centerline and hierarchical data extraction method, have errors branches, blood vessels, more bifurcation pixel redundancy and discrete defects centerline. In the process of extract the centerline and hierarchical data by pruning strategy to remove tiny branches, the introduction of a redundant pixels and judgment method to remove the centerline of the extra pixel can improve the robustness of the traditional method, finally to extract accurate vascular centerline and hierarchical data.Centerline extraction algorithm is based on Fast Marching Method(FMM). According to different FMM speed function, the calculation of different speed field. Firstly, algorithm set FMM velocity function as a often function, calculate each pixel in the blood vessels the shortest distance from the vascular boundary, get low speed field, and automatically select the point which has maximum distance value as the point source(Ps).Secondly, algorithm set the speed of FMM function is root function which the independent variable is the value of lower speed filed, calculate blood vessels each pixel point of the shortest distance from the source point(Ps), get medium speed field, take the whole block medium speed field through the breadth-first traversal method to obtain vascular endpoint. Finally algorithm set FMM speed function is a linear function which the independent variables is the low speed field value, calculate blood vessels each pixel the shortest distance to the source point to obtain high speed field, use gradient descent method in high speed field from blood vessel end points back to the source point, get vascular centerline.In the process of extracting hierarchical data, according to the geometrical characteristics of centerline points, algorithm marked bifurcation point firstly, then algrithm traverse the centerline from the main blood vessels point, and construct a dictionary tree.In ordert to express accurately the structure characteristics of blood vessels, algorithm use the method modified local point, eliminate the noise in the dictionary tree node and merge adjacent branch node. Lastly according to the feature, algorithm extract the accurate vascular classification data.Experiments show that, for most of the cases of data, algorithm can be quickly and effectively extract vascular centerline and hierarchical data.