基于多孔小波融合的PET/CT 图像增强及其评价

目的:为解决融合图像视觉效果增强与量化信息损失之间的矛盾,本文提出一种基于非降采样的多孔小波(àtrous wavelet)分解的PET/CT图像融合方法,使得融合图像既有利于肿瘤诊断又能用于放疗靶区勾画和放射性定量分析。方法:对PET和CT图像分别进行多孔小波分解,以包含肿瘤目标的适当大小的感兴趣区域的清晰度为目标函数,采用Nelder-Mead算法对PET和CT图像高频分解系数之比进行优化获得最终的融合系数,使融合图像充分增加解剖学信息的同时又尽量保持PET图像原有的局部和整体灰度信息。结果:融合图像质量评价表明,本文方法能将有价值的PET功能信息与精确的CT解剖信息结合在一起,并克服传统小波融合损失图像量化信息的不足。结论:基于多孔小波融合的PET/CT图像既能用于肿瘤诊断,又能同时用于肿瘤学放射性计算和适形放疗计划制定等量化研究。

Edge Enhancement and of PET/CT Imaging Using à Trous Wavelet Based Image Fusion

Objective: To overcome the contradiction of image visual enhancement and the loss of quantitative information,we proposed a new method to fuse PET and CT image based on non-subsampled à trous wavelet decomposition.Methods:The PET and CT images were decomposed using à trous wavelet respectively;the edge-sharpness of an tumor-enclosed area was selected as an objective function,then the Nelder-Mead algorithm was employed to optimize the initial fusion coefficient computed from the ratio of the decomposed highest frequency between PET and CT image to yield a fused image with maximum edge-sharpness around the tumor.Results: The proposed technique can not only extract the anatomical information from CT image and thus enhance the edge-sharpness of fused PET/CT image,but also maximally preserve its functional quantitative information both around the tumor area and background.Conclusion: Besides the application of tumor diagnosis,the resulted fused image can thus be used in radiotherapy planning in which the quantitative analysis is imperative.