Research On Optimization Problem Of Medical Image Registration

Medical image registration refers to the search for a(or a series of)spatial transformation of a medical image to make it spatially consistent with a corresponding point on another medical image.As the basic and key technology in medical image processing and analysis,it has important theoretical research and clinical application value.The registration process of medical images is essentially a problem of optimizing the multi-parameter similarity measure.The work of this paper is to improve the similarity measure and the optimization algorithm so as to improve the performance of medical image registration.Specifically,the effect of medical image registration under the different similarity measure condition of the same optimization algorithm is studied,and the effect of medical image registration under the same similarity measure condition of different optimization algorithms is studied.Among them,based on the gray-scale multi-modal rigid body registration method,the introduction of interpolation often leads to more or less interpolation artifacts of the objective function,which causes the similarity measurement curve to become unsmooth,so that the follow-up optimization algorithm easily falls into local extremum,eventually leading to registration failures.Based on this,this paper proposes an improved Sum of Conditional Variance(SCV)method that uses a new interpolation function instead of a high-order interpolation function to make the similarity measure curve smoother.This effectively prevents the registration process from falling into local extremum and obtains more accurate registration results.The problem of grayscale-based medical image registration is essentially the problem of the optimization of the registration function.One of the main tasks is to find an optimization algorithm with better accuracy and convergence.This paper introduces and analyzes the basic principle of differential search algorithm(DSA),and points out that it has the characteristics of simple structure,fast running speed,and is not easy to fall into local extremes.However,this algorithm has the disadvantage that it is difficult to find the peak at the later stage of iteration.It is not competent for the work of algorithm optimization in medical image registration.Based on this,this paper proposes an improved differential search algorithm(MDSA)and a hybrid differential search algorithm(HDSA).Compared with the original differential search algorithm,these two improved optimization algorithms combine the breadth of search and the depth of exploration,balancing the global and local nature of the algorithm,making the optimization process more stable and efficient.Experiments show that the improved optimization algorithm has a stronger global search capability and is more suitable for the optimization of medical image registration.