Research On SVM Algorithm For CT / PET Medical Image Registration Based On

With the rapid development of science and technology and medical imaging technology, used for check and diagnosis of human organs of various medical equipment constantly arise, for the development of modern medical provides more hardware support. The development of medical image processing is faced with the challenge of comprehensive application of multimode multi-source information, through integration between the various imaging techniques, greatly improve the accuracy of medical diagnosis and treatment, all these need with high efficiency, high precision between the medical image registration, it has become one of the hot technology of extensive research.Based on the above background, this paper analyze and research the existing image registration method, on this basis, the SVM algorithm is proposed based on support vector machine (SVM)of PET/CT medical image registration method.Firstly, the wavelet transform is introduced, the characteristics of CT image are extracted on the multi-scale, build the new CT image edge extraction algorithm. The wavelet transform feature extraction algorithm runs fast, it can suppress edge noise effectively, there is a lot of practical value for improve accuracy of edge detection.Secondly, using support vector machine (SVM) algorithm as the basis of image registration. Establish a parameter optimization model of support vector machine (SVM) algorithm for characteristic vector of the sample set by the return of the least squares support vector machine(SVM)algorithm performance to transform estimates of the model, finally find the optimal transformation model finally and improve the robustness of the algorithm.Finally, to improve the smoothness of image registration, the method of using gray level interpolation of image pixel value estimate, determine the final image registration. And compared with the traditional mutual information registration method, proceed the simulation experiment and analysis.Based on the SVM algorithm of CT/PET image registration method has very prominent advantages on Compared with the method of mutual information in computation time and memory space are needed, it’s an effective and practical application value of medical image registration method.