Application Of Mutual Information And Coarse-grained Parallel Genetic Algorithm In Medical Image Registration

Medical image registration technology is an important aspect of medical imageprocessing, through an organic combination of different modal medical images, it canprovide doctors with richer diagnostic information. Medical image registration based onmutual information has been widely used, traditional local optimization algorithmeasily falls into the local extremum, and because of the premature convergence problem,the genetic algorithm with global optimization may also fall into the local extremum,ultimately the method cannot get the global optimization solution. These optimizationmethods have lower registration accuracy when they are applied to the actual medicalimage registration.On the basis of the optimization algorithm of medical image registration based onmutual information, an improved coarse-grained parallel genetic algorithm is proposedand applied to the medical image registration. The improved algorithm makes use of thestructure of coarse-grained parallel evolution and adaptively adjusts the geneticparameters. Consequently, the population diversity is improved and the excellentindividual is protected. Moreover, the mind of different optimal values searched bydifferent sub-populations is truly realized and the algorithm can effectively overcomethe defect of local extrema. Meanwhile, combined with the termination conditions ofthe standard genetic algorithm, the convergence criteria of coarse-grained parallelgenetic algorithm is designed.The main contents of this paper can be summarized as follows:â‘ Detailedly discusses and analyses the research background, the research statusand the basics of medical image registration including the main technology and process.â‘¡The similarity measure and optimization algorithm of the medical imageregistration are introduced in detail. They are the most important two steps in imageregistration and determine the accuracy of the result. By the comparison of three classicoptimization algorithms, the migration operation can increase the population diversityof the coarse-grained parallel genetic algorithm, so the algorithm has better global andlocal research capabilities which are useful for avoiding premature convergence.â‘¢By combining the existing adaptive crossover and mutation rate with cosinefunction, a new adaptive formula is designed, and an improved coarse-grained parallelgenetic algorithm is proposed. As a result, the population diversity is improved, the excellent individual is protected and the mind of different optimal values searched bydifferent sub-populations is truly realized, so the algorithm can effectively overcome thedefect of local extrema and improve the accuracy of image registration. Combined withthe common termination conditions of the standard genetic algorithm, the convergencecriteria of coarse-grained parallel genetic algorithm is designed.â‘£The experiment of the method proposed by the paper: realizes the registrationalgorithm which is applied to single-modality and multi-modality medical imageregistration and compares the experimental result between the traditional algorithm andimproved algorithm. The experiment shows that the improved method can effectivelyreduce the local extremum phenomenon and has greater stability and higher registrationaccuracy, which can be well applied to the medical image registration.