Optimization Of Adaptive Mesh Refinement And The Application In Thrombosis Simulation

Computer-based simulation and tracking for moving interface are widely used due to its affordability,simple operation as well as guarantees for experimental completeness and comprehensiveness.Ever since the mechanism of reactive biological cascade leading to thrombogenesis has been confirmed,many coagulation-based thrombus simulation models have been proposed.Compared with conventional research on thrombus simulation,the computational model not only has less requirement for experiment operators and instruments,but it can also better describe the morphological varieties of blood vessels and thrombi.In thrombus simulation experiments,the growing surface of a thrombus is in and of itself a moving interface.Therefore,computations for interface tracking require the mesh of the zone in question to be solved.Among the mesh-based implementation method,the level-set method(LSM)does not require parameterization for curves or curved surfaces.What's more,combining LSM with adaptive Cartesian grids can not only realize graded refinement,but also avoid excessive memory consumption due to global refinement,while also improving the resolution of key regions.But such uneven refinement will result in hanging nodes,which will make calculation more difficult if a complete neighboring node set cannot be found.Moreover,iterative calculation will also gradually expand data size,which will increase both the calculating duration of the level set and the required storage space.In response to this problem,this study tried to improve the level set-based adaptive mesh refinement from two aspects and later verified the improved algorithm via a thrombus simulation.The main work of this paper is as follows:(1)The paper proposes to use red-black tree data structure instead of quadtree/octree data structure to store cartesian adaptive grid,which solves the problems of large storage space consumption and low computational efficiency.The two/three-dimensional space is usually segmented with a quad/octree.Although such structures can readily correspond with the physical space and storage structure,it may result in wastage of storage space and protracted duration for access of the designated node.For this reason,this paper selected the red-black tree data structure for storage and introduces Morton for node encoding to ensure high efficiency and express the adjacency between nodes.(2)This paper improve the red-green refinement and apply to cartesian adaptive mesh,which solves the problem that hanging nodes may cause calculation difficulties in hierarchical refinement.Bisection is usually used for refining quadrilateral mesh.Although the method is simple and can ensure consistency of the sub-mesh shape with the original mesh,it may result in hanging nodes between different refinement grades and in turn increase calculation difficulty.To keep the advantage while eliminating this shortcoming,the paper attempted to raise the efficiency by introducing and improving the Red-green refinement strategy.(3)The paper built a coagulation-based initial thrombus mixed growth model based on the thrombogenesis mechanism and used both the above improvement methods in the mixed model for the subsequent experiment.The mixed model is made up of the plasma sub-model,platelet sub-model,and thrombus sub-model.Blood vessel models with and without stenosis were established,then relevant measurements,e.g.,thrombus growth rate,thrombus disintegration duration,thrombus volume during disintegration,were collected under different blood flow velocities.Lastly,the experimental results were compared with other clinical results to verify simulation accuracy.