Research And Implementation Of Cerebrovascular Centerline Extraction Algorithm On 3D MRA Image

In recent years,with the rapid development of medical imaging technology and computer-aided diagnostic technology,Magnetic resonance angiography(MRA)are widely used in clinical diagnosis and treatment of cerebrovascular diseases.In the clinical diagnosis,the cerebrovascular centerline information of the image was obtained by the doctor plays a good role in the treatment of cerebrovascular disease.Therefore,the cerebrovascular centerline extraction technique of MRA image becomes one of the core issues in the medical image processing.Because the heart of coronary vessels and retinal vessels are usually researched by experts.And there are few studies on cerebrovascular,which are on two-dimensional cerebrovascular images.In addition,because of the three-dimensional cerebrovascular centerline of the study there are high complexity,the study needs to be explored by people.These issues which are the weak relative research of the two-dimensional cerebrovascular centerline and the lack of the three-dimensional cerebrovascular centerline are targeted.In this thesis,the cerebrovascular centerline extraction algorithm of MRA image is researched,including multi-scale filtering enhancement algorithm,ridge tracking algorithm and Unfold-Snake deformation model.In this paper,the traditional deformation profile model is improved by combining the advantages and disadvantages of the gradient vector flow field and the traditional deformation profile model with the complexity of 3D MRA cerebrovascular images.And the multi-scale filtering enhancement algorithm,the ridge trace algorithm and the Unfold-Snake deformation model are applied to the cerebrovascular centerline extraction.The main contents of this paper are as follows:(1)In this thesis,the gray features of cross-sectional cerebrovascular,Gaussian kernel scale space and Hessian matrix of vessels were studied and analyzed.And according to the analysis of cerebrovascular characteristics,a cerebrovascular response function is put forward by author.And the multi-scale filtering preprocessing enhancement algorithm is applied to the pretreatment of cerebrovascular centerline extraction.(2)Secondly,the ridge trace algorithm of cerebrovascular centerline is designed by this paper.And it is applied to the initial contour extraction of the target vessel in the cerebral angiography image.For the next step in the initial contour deformation,a lot of preparations are done.(3)Finally,the traditional activity contour model algorithm and the gradient vector flow field are studied and analyzed by this paper,and several traditional models are compared and analyzed,and an Unfold-Snake deformation model is designed and compared with the traditional model.And it is applied to the initial contour deformation of the cerebrovascular centerline.The results show that this algorithm has a good ability of cerebrovascular centerline extraction.