Research On Artificial Microvascular Tree Generation Method Based On Generative Adversarial Networks

Investigation of the structure of microvascular tree plays a vital role in studying the pathophysiological process of circulatory system.Modeling and simulation of the hemodynamics are important approaches for studying the function of microvascular trees.However,it is difficult to obtain the real microcirculatory vascular tree structures.Therefore,the researchers attempted to generate microvascular trees artificially.Conventional methods are mostly based on statistics and fractal theory to generate microvascular trees.The generated microvascular trees exhibit high homogeneity and cannot reflect the real microvascular tree structure with high heterogeneity.Based on these homogeneous microvascular trees,it is difficult to analyze the physiological and pathological problems such as vascular growth and tissue perfusion.These microvascular trees are also difficult to be used as reasonable boundary conditions in modeling and simulation.Therefore,generation of microvascular trees in line with the actual physiological structure and topological characteristics is of great research value.In view of the shortage of conventional methods,the actual characteristics of microcirculation system cannot be accurately expressed.In this thesis,an artificial microvascular tree generation method based on generative adversarial network(GAN)is proposed.The main research contents of this paper are as follows:(1)A data preprocessing method for vascular tree is proposed,which caneffectively conduct deep learning training based on the preprocessing data and generate vascular data from the trained model.(2)A microvascular bifurcation generation model is established based on generative adversarial network.The model integrates the techniques of conditional generative adversarial network and spectral standardization generative adversarial network,and makes a series of improvements to the generator and discriminator,and finally generates the microvascular bifurcation that accords with the topological characteristics of real microvascular tree.(3)Design the strategy of microvascular tree splicing and construct the microvascular tree automatically.The results of qualitative and quantitative analysis of the microvascular tree show that the proposed method is superior to the conventional method,which proves the effectiveness of the proposed method.In summary,the recently popular GAN model was applied to the task of microvascular tree generation.The results show that the performance is better than the traditional method.After further optimization,this method is expected to conduct modeling and simulation research on the auxiliary circulatory system,which can provide support for exploring the physiological and pathological mechanism of cardiovascular system.