Research On Intelligent Analysis And Visual Treatment Of Cerebral Arteriovenous Malformations

Cardiovascular and cerebrovascular disease is a serious threat to human health.It has the characteristics of high morbidity,high disability and high mortality.Cerebral arteriovenous malformation(AVM)is a serious cerebrovascular disease.Magnetic resonance angiography(MRA)is a standard imaging method commonly used in the diagnosis of AVM.The pathological characteristics of cerebral arteriovenous malformations are that there are no capillaries between arteries and veins,but a group of abnormal vascular structure with uneven diameter and thickness of vessel wall,which lacks elastic layer and muscular layer and is prone to rupture and bleed.At present,there are many challenges in the diagnosis and treatment of AVM,which can be summed up as follows: 1)In the aspect of Theoretical method research: there are very few algorithms to analyze the internal structure of AVM in detail,and these algorithms generally rely on a large number of artificial initial markings and experience operations,with low accuracy.In addition,most of them assume that AVM is a compact spherical structure,and they perform poorly on AVM data with relatively sparse vessels.2)In the aspect of clinical diagnosis and treatment technology research and development: the effect of interventional diagnosis and treatment of cerebral arteriovenous malformation is closely related to the accuracy of focus anatomical structure knowledge and interventional path planning.The existing imaging methods are still not enough for the visual tracking and operation of vessel path.The realization of 3D visualization of the complex structure of AVM can facilitate doctors to understand the patient's vascular structure and surrounding tissues,and complete the interventional operation planning.In view of the above theoretical and clinical problems,there are two innovations in this paper.Firstly,this paper proposes an AVM localization and segmentation method based on graph theory,which can automatically segment the AVM nidus,feeding arteries and draining veins according to the topological relationship.Secondly,for AVM visualization,we designed and developed the AVM Interventional Surgery Visualization system,which can not only understand the length and key position of the interventional path through the combination of 3D vascular model and centerline display,but also map the tip position of the guidewire collected during the operation to the image space through the registration algorithm.The main work includes the following parts:1.Using the finite mixture model and Markov random field to segment the cerebral vascular structure from the MRA images,using the skeleton algorithm to generate the vascular skeleton centerline.Based on the relevant knowledge of graph theory,the skeleton centerline is transformed into a skeleton map composed of vertices and edges.2.In this work,we proposed a novel weight-based breadth-first search method for AVM location and segmentation.The AVM nidus was located in the spanning tree of the skeleton image using our method.A topological path was established to segment the feeding artery and draining vein connected to the AVM nidus.3.This paper designed and developed the AVM Interventional Surgery Visualization system.Through the combination of 3D vascular model and different centerline branches,we can know the length and key position of the interventional path very vividly and concretely.The real-time vascular roaming is realized in combination with the preoperative interventional operation planning path,which is convenient for doctors to understand the position of the guidewire tip during the operation and avoid the wrong branches.In a word,we have analyzed the problems of AVM segmentation and visualization comprehensively,and made innovations from two aspects: algorithm and system.The accuracy of nidus anatomy structure and intervention path planning is helpful to improve the interventional diagnosis and treatment effect of cerebral arteriovenous malformation.We used clinical data and phantom data to test,which proved that the algorithm proposed in this paper can automatically and more accurately extract AVM structure,make operation plan,and realize real-time intravascular roaming,which is helpful for comprehensive interventional diagnosis and auxiliary treatment combined with extravascular and intravascular tissue structure.