MRI Brain Tumor Image Segmentation And Three-dimensional Reconstruction

Brain tumors are one of the diseases that seriously endanger the safety of patients.The brain tissue structure is complicated,which brings great difficulties to doctors' diagnosis and treatment.There are advantages that the magnetic resonance imaging(MRI)technology has no effect of radiation on the human body,good imaging effect on the structure and the ability to achieve arbitrary azimuth tomography.Therefore,doctors often use MRI brain tumor images to analyze and treat brain tumors.Brain tumor image segmentation algorithm and 3D reconstruction technology are beneficial to doctors to observe the shape and size of tumor visually,which is of great significance for disease diagnosis and surgical treatment.In this thesis,sparse subspace clustering algorithm based on sparse representation theory is used to realize tumor segmentation in multimodal brain tumor images,and then the segmented brain tumor images are reconstructed in three dimensions by using the marching cubes(MC)algorithm.The main research contents of this thesis a as follows:(1)The single-mode brain tumor image is difficult to describe the brain tumor comprehensively.In order to know how to effectively fuse tumor information of multimodal brain tumor images,based on the sparse subspace clustering algorithm,multi-modal brain tumor image segmentation is studied from single mode segmentation and multi-modal fusion in this thesis.Experiments show that the multimodal brain tumors image segmentation method based on differential operation achieves better segmentation results both in subjective evaluation and objective evaluation for two-dimensional multi-modal brain tumors images.(2)Different features have different effects on tumor segmentation results.In order to make better use of multimodal brain tumor image information,a multi-modal brain tumor image segmentation method based on feature differentiation is proposed in this thesis.The visual difference image is extracted from multi-modal brain tumor image,then the super-pixel segmentation is performed on the visual difference image to extract the super-pixel feature,the distinguishing ability formula is constructed,the corresponding weight is set according to the distinguishing ability of each feature,the weighted feature matrix is obtained,finally,the sparse subspace clustering algorithm based on the sparse representation is used to complete the segmentation.Experiments are carried out on the proposed method using Brats 2015 competition data.The results show that the proposed method can comprehensively integrate the tumor information of multimodal images and obtain better segmentation results.(3)In order to meet the clinical requirements of 3D reconstruction's rapid realization,the marching cube algorithm in surface rendering to reconstruct the segmentation result image in three dimensions is used in this thesis,and the tumor information is displayed from multi-dimensional,which provides a basis for the design of clinical diagnosis and surgical treatment.