Research On Automatic Segmentation Of Cerebral Vessels Based On Convolutional Neural Network

In the modern society,cerebrovascular disease has become one of the important diseases that threaten people's health and life.And the research of cerebrovascular disease has become increasingly important.In addition,with the development of the angiography image technology and computer technology,the method of computer aided diagnosis and treatment with angiography image has become a research focus and difficulty,the vascular angiography image segmentation is one of the key techniques in diagnosis.The segmentation of blood vessels is to separate the part of the blood vessels in the image,which is of great significance to the diagnosis and treatment of vascular diseases.At present,because of the complexity of imaging methods and the diversity of vascular structure,although a lot of segmentation methods for brain vessels are proposed,but there is not a general method with high segmentation accuracy,which is suitable for a wide variety of angiography images.This paper mainly studies the method of automatic segmentation of brain blood vessels based on convolutional neural network,and applies it to the segmentation of brain CTA images.Specifically,the main contents of this paper are as follows:(1)First,the significance and importance of the study on the segmentation of cerebral angiography images are introduced.At the same time,the research status of convolution neural network at home and abroad is summarized.(2)Introduce the related knowledge of cerebral vascular segmentation,including that study and summarize angiographic imaging technology and medical image storage technology,introduce the classification and theory of the conventional segmentation methods of cerebrovascular vessels,and the theoretical study of convolutional neural network(3)The segmentation technique based on multi-modality convolutional neural network is studied.First three modalities are added to the original convolutional neural network by Gauss filtering,Laplace filtering and Gabor filtering.Then train the model by parallel convolutional neural and fuse the segmentation results.At last,the segmentation results are processed by Gauss mixture model and fuzzy C mean methods.The contrast experiment of the real brain CTA images indicates that the combination of multiple modalities has more advantages than the single modality in the multi-modality convolutional neural network proposed by this paper.And compared with the conventional segmentation algorithm for cerebrovascular vessels,the multi-modality convolutional neural network achieves more accurate segmentation results.