Design And Implementation Of MRA Brain Vessel Segmentation

With high morbidity and mortality rates,the cerebrovascular disease have a great threat to the health of human beings.If we early detect and treat the cerebrovascular disease,we can curb the development of it.Accurate cerebrovascular segmentation is the basis of auxiliary diagnosis treatment of cerebrovascular disease.Therefore,cerebrovascular segmentation is of great significance to the auxiliary diagnosis of cerebrovascular diseases and health of human beings.Cerebrovascular segmentation not only can greatly reduce the doctor read workload,but also can provide doctors with more richer and more accurate information about the disease.Hence,cerebrovascular segmentation study has always been an enduring hot issue.In view of the cerebrovascular segmentation,this paper mainly completed the research of the following three aspects.1.Through the study of the based algorithms of medical image segmentation algorithms,such as threshold segmentation method(iterative adaptive threshold and Otsu adaptive threshold method),region growing segmentation method,level set segmentation method,we have a better grasp of the basic image segmentation knowledge and lay the foundation for proposing model and improving model in the following.2.Design and implementation of multi-scale enhancement algorithm based on the improved Hessian.Traditional Hessian multi-scale enhancement algorithm not only enhance brain blood vessels,but also enhance the nasal soft tissues(the shape of them is similar with cerebrovascular),isolated noises and the borders of the image.These enhancements will seriously affect the following Blood vessels segmentation and interfere with the results of the blood vessels segmentation and result in over segmentation at nasal tissues,the boundary and the isolated noise points.What’s worse,they will seriously affect the accuracy of the segmentation.To solve this problem,we propose a method that combine Hessian multi-scale and image gray level information.We add gray scale factor to the original vascular similarity function.The gray factor will weakened the enhanced effect at borders of images,nasal soft tissues and isolated noise points.The results show that the improved Hessian multi-scale filtering method proposed in this paper can effectively reduce the enhancement at nasal tissues and the borders of image and can also reducing the isolated noise points generated by the enhancement for MRA cerebrovascular data.The processing of CT pulmonary vascular data can effectively reduce the isolated noises points,the enhancement effect of image boundary and the enhancement effect of the boundary of cerebrospinal fluid and brain gray matter.3.Design and implementation of MRA fuzzy c-means vessel segmentation algorithm based on the vascular feature.In the case of image segmentation of tubular structure,Fuzzy C-Means(FCM)algorithm is difficult to get good results only by the Euclidean distance between cluster centers to determine the fuzzy membership classification.It will produce false segmentation on vessels with having little gap in the grey value between vessels and background.Due to the effect of noises,the blur of vascular morphology and the low contrast between target and background in MRA images,the FCM algorithm cannot get a good segmentation result.In order to solve this problem,this paper proposes an MRA fuzzy c-means vessel segmentation algorithm based on the vascular feature.Experiment results show that the proposed method can effectively extract the vessel structure and get more information of blood vessels comparing with the traditional FCM.