Research On The Registration Method Of Craniocerebral CT-MRI Images

The cerebral lesions are difficult to be observed directly because of their closed characteristics.CT and MRI tomography can be used to diagnose and treat intracranial diseases effectively.CT images have high-resolution in bone tissue,while MRI images have highresolution in soft tissue imaging.Based on the fusion images after CT-MRI registration,multimodality information can be used to better observe intracranial lesions and accurately diagnose diseases.CT and MRI images can’t be directly fused due to differences in imaging locations and equipment coordinate systems.Therefore,registration of CT and MRI images is the premise of accurate diagnosis and treatment based on multi-modality information.In order to improve the speed and accuracy of cerebral CT-MRI image registration,an Improved Brain Storm Optimization(IBSO)combined with Simplex was proposed.Firstly,the optimal clustering center was used as a cue to generate new individuals to replace the worst clustering center to improve the brain storm optimization algorithm.At the same time,the images to be registered were decomposed into a multi-resolution image pyramid.Then,based on the multi-resolution strategy,the IBSO was used to register the top image of the image pyramid.Then the coarse registration results are used as the starting point of the Simplex to accurately register the middle image and the bottom image of the image pyramid.Experimental results show that,compared with Simplex combined with Particle Swarm Optimization,Powell combined with Differential Evolution,Powell combined with Brain Storm Optimization,the average error and the maximum error of the proposed algorithm in the mono-modality experiment are the smallest,and the four metrics,mutual information,normalized mutual information,cross-cumulative residual entropy and normalized cross-correlation,of the proposed algorithm in the multi-modality experiment are all superior to other algorithms.In order to solve the misregistration problem of low-resolution images under multiresolution strategy,a registration method based on Co-Evolutionary Brain Storm Optimization(CEBSO)was proposed.Firstly,the BSO is improved according to the fitness value of the objective space,the elite population and three normal populations are divided to replace the clustering process in the original algorithm;Different evolutionary strategies were selected to increase the population diversity of three normal individual populations.Then the proposed algorithm was registered with Genetic Algorithm,Differential Evolution,Particle Swarm Optimization and BSO respectively.In the mono-modality experiment,the mean square error of the registration method based on the proposed algorithm was reduced by 2.35%,2.53%,2.24% and 0.72% respectively,and the translational error was the least.In the multi-modality experiment,the registration results of the proposed algorithm were generally better than comparison algorithms.The IBSO combined with Simplex registration method can fully combine the global search ability of IBSO with the local search advantage of the Simplex,so that the registration speed and accuracy of high-resolution images can be improved.The registration method based on CEBSO solves the misregistration problem of low-resolution image registration with multiresolution strategy,and makes the registration error significantly reduced.The proposed registration methods solve the problem of low registration accuracy of CT-MRI images,and can provide high-precision registration results for subsequent image analysis.