| The eye is the most important sensory organ of the human body.It is filled with aqueous humor.The normal flow of aqueous humor can provide the necessary nutrients for the eye tissues and maintain normal physiological activities.However,the pathogenesis,development process,and drug treatment of many ophthalmic diseases are closely related to the flow of aqueous humor.If the flow law of aqueous humor is abnormal,which may lead to the occurrence of eye diseases.Therefore,systematic and in-depth study of the aqueous humor dynamics of the human eye and exploring the key factors of its impact are of great practical significance for enhancing people's understanding of the pathogenesis and development process of ophthalmic diseases,as well as improving the therapeutic effect of eye diseases.The anatomical structure of the human eye is complex,and the flow of aqueous humor is diverse.The use of theoretical and experimental methods to study the dynamics of aqueous humor is subject to many restrictions and is not easy to achieve.The numerical method has the advantages of short time period,good repeatability,and low calculation cost.Many researchers have used numerical methods to study the dynamics of aqueous humor,and have achieved many results.However,the study of the real three-dimensional human eye aqueous humor dynamics involves the complex anatomical structure of the human eye,the difficulty of boundary processing,and the huge amount of calculations.Therefore,compared to traditional numerical methods,the Lattice Boltzmann Method(LBM)with the advantages of easy multi-physics coupling,simple boundary processing,and easy parallelism is a promising new way to study the dynamics of aqueous humor.For this reason,this paper carried out a three-dimensional human eye aqueous humor dynamics research based on GPU parallel technology and LBM.Main tasks as follows:(1)This paper firstly uses the square cavity natural convection to numerically verify the three-dimensional double distribution function LB model,and proves that the model can accurately simulate the thermal convection phenomenon and can be used to simulate the convection diffusion of aqueous humor.Then,a three-dimensional geometric model of the anterior segment of the eye was constructed from the anatomical structure data of the human eye,using a LB evolution equation to solve the flow field composed of aqueous humor flow,and using another LB evolution equation to solve the temperature field composed of the temperature difference of the intraocular tissue,using Darcy's law solves the aqueous seepage at the trabecular meshwork,and uses the Boussinesq approximation to couple the flow field and the temperature field to establish a three-dimensional double distribution function LB model of the anterior segment aqueous humor dynamics of human.Then,the established model was used to simulate the normal intraocular temperature difference and the velocity and temperature distribution of aqueous humor in the anterior segment of the eye under different eye orientations,and the feasibility of the model was verified by comparison with relevant literature.(2)Using the validated model,the influence of different factors on the flow of aqueous humor was further studied.The experimental results show that natural convection is the main flow mode of aqueous humor,and the difference is obvious under different eye positions,the aqueous humor flows faster when standing;The intraocular temperature difference has the greatest effect on the flow of aqueous humor,indicating that the flow of aqueous humor is mainly driven by the buoyancy force dominated by the temperature difference;the pupil size has little effect on the flow of aqueous humor;and the rate of secretion of aqueous humor and trabecular meshwork permeability have The influence of internal pressure(IOP)is relatively large,in which IOP increases with the increase of aqueous humor secretion rate,but decreases with the increase of trabecular meshwork permeability.(3)Finally,the current model was used to simulate the aqueous humor dynamics of human eye after laser iridotomy.The experimental results show that when the pupil is blocked,IOP is increased,and laser iridotomy can change the flow path of aqueous humor to achieve the purpose of reducing IOP;In addition,the farther the position of the laser iris opening from the pupil axis,the greater the wall shear stress on the corneal endothelial cells.Therefore,the position of the laser iris opening can reduce the risk of corneal endothelial cells falling off due to excessive wall shear stress,And effectively improve the harm caused by pupil block.(4)In this paper,in order to solve the huge bottleneck problem involved in the calculation of 3D human eye aqueous humor dynamics model,GPU technology is used to design and optimize the 3D human eye aqueous humor dynamics model in parallel.First,based on the conventional GPU parallel technology,a three-dimensional human eye aqueous humor dynamics model LBM parallel algorithm is proposed,and the acceleration effect is about 470 times.In order to further improve the performance of the algorithm,the acceleration effects of optimization strategies such as the division of thread block dimensions and the use of shared memory are discussed.The experimental results show that the performance of one-dimensional Grid and one-dimensional Block schemes is the highest.At the same time,the reasonable use of shared memory can further improve the performance of the algorithm.(5)The flow field of the 3D human eye aqueous humor dynamics model has sparse characteristics,including fluid lattice points(only 38%)that need to be calculated composed of aqueous humor,and solid lattice points composed of cornea,iris,and lens that do not need to be calculated.In response to this feature,this paper proposes a part addressing algorithm based on indirect addressing and a part sparse matrix algorithm based on sparse matrix.These algorithms can save more than half of memory consumption and greatly improve the performance of the model.Among them,the part sparse matrix algorithm based on sparse matrix has the best performance.After synthesizing the various optimization strategies proposed in this paper,the acceleration ratio of the GPU program can reach 866 times,and the performance is 1541.89 MLUPS,which effectively improves the computing performance of the current model. |
PDF Full Text Request