3D MRA segmentation using the vesselness filter

Abstract

Blood vessels in Magnetic Resonance Angiography (MRA) image plays an important role in medical diagnosis of divers diseases. Cerebrovascular accident (CVA) is the main cause of death. The three dimensional segmentation of MRA images is helpful for the detection of the CVA in early stage. Due to the low contrast of thin vessels, loud noise and the complex structure of vessels, it is difficult to extract vessels from MRA images precisely. In this paper, we present a new method of segmentation. The proposed algorithm contains two major steps: Firstly, the Contrast Limited Adaptative Histogram Equalization (CLAHE) method is applied to enhance the image. Then, the vesselness filter is used to extract the blood vessels. Our method was tested and evaluated on 3D MRA database. It demonstrates the ability to extract the most of the vascular structures successfully. The accuracy of the proposed method reaches more than 95% which was higher than the recent methods.