Centerline Extraction Methods Under Complex Background

Centerline extraction is an important step for both depicting vascular topological structureand quantifying vascular structures in analytical approaches. Experts have put lots of effort forfinding out fast and accurate centerline extraction methods of vascular structures in medicalimages. However, there are still many obstacles in finding out one proper centerline extractionmethod for all purposes. The background of the vascular structures are quite complex due toimage noises and lower intensity contrast. And also the vascular structures themselves havevariations in radius, curvature and bifurcations, which make centerline extraction a tough task.In this paper, we focus on vessel centerline extraction method under complex background,and improved methods have been proposed for better centerline results.Firstly, we analyze the complex background of vascular images, and carried outpre-processing steps, like reducing background and enhancing vascular structures, which will leadto better centerline extraction results later.Secondly, after carrying out a wide research on vessel centerline extraction methods, weimproved traditional methods for dedicate purpose. In chapter3, we present a vessel tracking andthinning method to get the entire topology of vessel image, and the results show that low contrastbifurcations of centerlines can be found. In chapter4, we modified the speed function of minimumpath centerline extraction method, which constrains the propagation area close to vessel structuresand gets accurate centerline. In chapter5, we proposed a centerline-based method for cystic arterysegmentation. Centerlines are extracted by tubular feature and GVF feature. Results show that thetiny bifurcations can be detected for cystic artery segmentation.