Research On Computer-aided Diagnosis Of Moyamoya Disease Based On Magnetic Resonance Imaging

With the continuous development of computer-aided diagnosis technology,Medical imaging equipment such as MRI,CT and related image data post-processing work play an increasingly important role in the detection and analysis of cerebrovascular diseases.Moyamoya disease is a cerebrovascular disease,the incidence is not high but it is very harmful.As a non-invasive imaging technique,magnetic resonance technology plays an important role in the diagnosis of Moyamoya disease.However,due to imaging differences between magnetic resonance image sequences,magnetic resonance angiography images are insensitive to intracranial hemorrhage,and susceptibility weighted imaging images can effectively identify hemorrhage areas but do not clearly show intracranial vascular structures.The location of the hemorrhage areas and the correlation with blood vessels are the focus of diagnosis of Moyamoya disease.Quick access to the location of the hemorrhage areas and vascular structure information for Moyamoya disease is very helpful for the doctor's diagnosis.In order to solve this problem,aiming at pathogenic features of hemorrhagic Moyamoya disease,according to the image characteristics of different sequences in the magnetic resonance image,this paper cooperates with hospitals to deal with the image data of patients with Moyamoya disease,which is intended to provide corresponding assistance for the diagnosis of doctors.Through tests and analysis of actual clinical data,the feasibility and rationality of the design scheme in this paper are verified.The main work of this thesis includes the following three aspects:(1)Segmentation of the hemorrhage areas of Moyamoya disease through SWI sequence image.In order to ensure the accuracy and stability of the segmentation results,two semiautomatic segmentation methods of human interaction are studied and implemented: Region growing method based on gradient response function and Canny edge operator,Active contour model based on gradient vector flow field(GVF-Snake).In this paper,Combined with experimental results,the applicability and robustness of the two methods are analyzed,and the advantages and disadvantages of the two methods and the practical application scenarios are compared and explained.(2)Segmentation of the intracranial vessels of Moyamoya disease magnetic resonance angiography(MRA).Firstly,a series of preprocessing methods are used for image filtering,skull stripping and segmentation of lateral ventricle.Then a multi-threshold OTSU classification and multi-scale vessels enhancement based on Hessian matrix eigenvalues are used to segment cerebral vascular voxel.This method acquires the three-dimensional structure of the cerebral vessels successfully.(3)Registration of the MRA sequence and the SWI sequence according to the relationship between these sequences.Frameworks such as VTK,ITK and algorithms are integrated in software in this paper.The volume rendering method is used to render the hemorrhage area and cerebral vessels for the three-dimensional visualization,which can visually display the spatial and internal information between the brain tissues.Combined with the volume of the hemorrhage areas,this research provides some support for research and diagnosis.