关键词: PET/CT, 图像融合, 双树复小波, 高斯隶属度函数, 自适应

Abstract: PET/CT medical image fusion has very important application value for medical image analysis and diseases diagnosis. It is useful to improve the image content and accuracy by fusing PET/CT images. Aiming at PET/CT fusion problem, this paper proposes a self-adaption fusion algorithm of PET/CT based on dual-tree complex wavelet transform. Firstly, source PET and CT images after registration are decomposed low and high frequency sub-images using dual-tree complex wavelet transform (DTCWT). Secondly, according to the characteristics of low frequency sub-images concentrating the majority energy of the source image and determining the image contour, a fusion rule based on self-adaption Gaussian membership function is adopted in low frequency sub-band coefficients. Thirdly, in high frequency sub-images, according to the relation among region pixels and fuzziness, in the first layer of high-frequency component, Gaussian membership function and 3×3 field windows are used to fuse the high-frequency dual-tree complex wavelet coefficients. In the second layer of high-frequency component, regional variance fusion rule is used. Finally, in order to verify the validity and feasibility of the proposed algorithm, three experiments are done, comparison experiment of the proposed algorithm and other pixel-level fusion algorithms, the fusion effect evaluation experiment with information entropy, mean, standard deviation and mutual information, and comparison experiment with different fusion rules of dual-tree complex wavelet transform. The experimental results show that the proposed algorithm can improve the information entropy by 7.23%, and mutual information by 17.98%. That is to say the algorithm is an efficient fusion method of multimode medical image.

Key words: PET/CT, image fusion, dual-tree complex wavelet, Gaussian membership function, self-adaption