Multi-channel DTI Registration Based On Improved Learning Factor PSO And Active Demons Algorithm

Alzheimer's disease,as an irreversible chronic disease,cannot be cured once it becomes ill,so the treatment of Alzheimer's disease is mainly prevention.Before the onset of Alzheimer's disease,the abnormal white matter structure corresponding to specific functions was abnormal,which is very similar to the diseased part of the patient after diagnosis.The study of the white matter structure is of great significance for controlling the deterioration of the disease and the pathogenesis.Diffuse tensor imaging is an important tool for non-invasive analysis of the internal structure of the brain.It is very sensitive to the changes in the white matter microstructure.It makes it possible to image the white matter fiber bundles in the living body by utilizing the diffusion characteristics of water molecules in human tissue.The registration of diffuse tensor images of the brain of patients with Alzheimer's disease can enable doctors to accurately determine the changes in white matter and control the deterioration of the disease.The registration of brain images of patients at different stages of disease is of great value for studying the development of the disease and accurately determining the location of the disease.The multi-channel registration algorithm based on Active Demons is suitable for registration of continuous grayscale diffusion tensor images.The registration accuracy is high and the registration of large parts of the image deformation area can be better.But three key parameters : the optimal values of the balance coefficient,the homogenization coefficient,and the elastic coefficient are difficult to determine.This paper proposes an optimal parameter determination method based on the extreme value method of MSE,which obtains experimental data by fixing two parameters and changing one parameter,Then according to the functions of the balance coefficient and the mean square error MSE,the averaging coefficient and the mean square error MSE,the elasticity coefficient and the mean square error MSE,the optimal values of the three parameters of the DTI image registration are obtained by using the method of obtaining the extreme value of the function.However,the optimal value determination method based on the relationship between the evaluation indexand the parameter function only considers the influence of a single parameter,and does not optimize the combination of multiple parameters.This paper proposes to introduce particle swarm optimization algorithm into Active Demons algorithm for DTI image registration,and realizes three key parameters of automatic optimization.The introduction of particle swarm algorithm improves the convergence speed and accuracy while solving the problem that parameters need to be manually set.The particle swarm algorithm is easy to fall into a local optimum when dealing with complex problems.In this paper,an improved learning factor PSO algorithm is proposed to improve the registration accuracy and convergence speed while increasing the registration efficiency.This paper uses genetic algorithm,cuckoo algorithm and particle swarm algorithm to optimize Active Demons algorithm for DTI image registration.Experiments prove that particle swarm optimization algorithm has better optimization ability.Then,the MSE extreme value Active Demons algorithm,the improved learning factor PSO algorithm Active Demons algorithm,the basic PSO algorithm Active Demons algorithm,the Demons algorithm,and the variable parameter Active Demons algorithm are applied to the registration of DTI images.Experimental results show that the proposed algorithm has higher convergence speed and registration accuracy.