Research On Registration Algorithm For Three-dimensional Multi-modal Medical Images

With the rapid development of computer technology and biological technologyin the medical field, a variety of medical images from different imaging devicesappear, and these images from different devices are called multi-modality images.These multi-modality images can provide the doctors with important information ofan organ or tissue. Medical images gained from a single imaging device cannotprovide enough useful information; usually we need images from a wide variety ofimaging devices to compare with each other to get more comprehensive information.Since information gained from two images acquired in the clinical track of events isusually of a complementary nature, proper integration of useful data obtained fromthe separate images is often desired. Registration of medical Images is theprerequisite for the work above, so it is extremely important.In this paper, we got started with the generic framework of image registrationcomposed of four modules, namely the Transform module, the Similarity Metricmodule, the Interpolator module and the Optimizer module, then the mainalgorithms of each module were introduced. In order to get more precise results,mutual information was used as the similarity measure.The experiments showed that the result gained from the method based onmutual information was more precise than the result gained from the method basedon AM, but the method base on mutual information cost more time. In this paper, byprocessing the reference image before registration, we proposed a new way ofcalculating the mutual information to reduce the amount of computation. The resultof the experiments showed that the accuracy of registration did not change much,whereas the time of calculating the mutual information decreased significantly.In this paper, the genetic algorithm was improved according to the defects ofthe original genetic algorithm, by setting a greater probability for individuals withbetter adaptation to the environment to survive and a greater probability forindividuals with worse adaptation to the environment to be eliminated and varied.The experiments showed that the result gained from the method based on theimproved genetic algorithm was more precise than the result gained from themethod based on the original genetic algorithm.