Research On Partial Differential Equation-based Segmentation Algorithms For Brain Tumor MRI Images

Brain tumors have become a serious threat to human health because of their high incidence and mortality.Diagnostic analysis of brain tumors has always been one of the hot research directions at home and abroad.In the treatment of brain tumors,accurate location and segmentation of tumors is the key to further treatment.In the current mature imaging technology,magnetic resonance imaging(MRI)has become the preferred way to obtain information of brain regions because it can obtain high-quality images of the imaging target and the whole process has less damage to the human body.Therefore,the study of brain tumors segmentation on MRI images has become an important topic in the field of brain tumors research,and is also a key step in computer-aided diagnosis of brain tumors and subsequent treatment planning.In this paper,the common image segmentation methods are introduced,and the research status of brain tumors MRI image segmentation is analyzed in this paper.Then three kinds of brain tumors MRI image segmentation algorithms are proposed.Then three methods are proposed,which are based on the level set method.Using one of improved variational level set method makes it suitable for brain tumors MRI image segmentation.Firstly,the salient target extraction technology is used to locate brain tumors.Then,in the local area of brain tumors,the variational level set method of local intensity clustering is modified,so that it only acts on a local area near the zero level set to form a basic automatic segmentation approach.Secondly,a priori shape energy term is introduced into the energy function to construct the energy function.The priori shape of brain tumors is obtained by sparse shapes compose in independent component space.At the same time,by approximately calculating the parameters of affine transformation,the shape can be kept unchanged in translation,rotation and scaling.It can reduce the influence of the initial contour position and size on the first method ins some ways,and avoid that the evolution curve may stay in the local area of the image.Finally,the energy function is constructed using the priori shape from the depth network.Using the advantages of U-Net network,such as fast operation speed,less sample size and simple network structure.The approximate shape and location information of brain tumors are obtained by U-Net network,and then the shape is used as a priori shape to segment brain tumors accurately using improved local intensity clustering algorithm.This method can avoid the unsatisfactory reconstruction results of complex shapes which are different from the average shape of sample set when constructing priori shapes by sparse shape combination.By using the image in BRATS 2017 data set,the segmentation effect of the proposed method is tested and the advancement and effectiveness of the proposed algorithm are proved.