Multimodality Registration Of CT And MRI Head Images Using Principal Axes Transformation And Simulated Annealing Algorithm

Different modility of medical images provides different medical information, so we can not get comprehensive information from only one modility image. But at present doctors acquire total medical information merely through then- ability of integrated imagination, so the result is not accurate and some important information is likely to be overlooked.The aim of medical image registration is to align the corresponding point in two medical images. Then image fusion technique can be used, and a new image containing all medical information of both images can be provided to doctors, through which medical information in different modility image can reinforce each other.Given the head image's characteristic of CT and MRI, we present a new registration strategy which include a coarse registration and accurateregistration. Firstly, we attain coarse registration by matching the contour of two images, and get the parameters range for the accurate. Then we adopt MI (Mutual Information) as cost function, PV (Partial Volume Interpolation) as interpolation, a new combined optimization algorithm that we invited as our optimization algorithm to search the best registration parameters. Theories and experiments indicated that this method have the advantages of high precision, good robust and good ability of resisting noise. In the second part, after discussing the cost function, image resampling algorithm and image trasformation method, we proposed a registration based on mutual information, PV interpolation and particular image trasformation. In the third part, image registration based on principal axes algorithm is studied, we can attain coarse registration after matching the contour. In the fourth part, optimization algorithm is studied: we presented a new algorithm which combines the advantages of both local optimization algorithm(model seach algorithm) and whole optimization algorithm (simulated annealing), Which not only overcome the problem of local extremum, but also improve the speed and precision of registration. In the fifth part, a new registration strategy is presented. In the last part, the evaluations were given.