多谱CT成像是通过不同谱段的CT图像表征检测对象中的不同组分。 为了便于在同一视图中显示所有组分的信息, 需要研究多谱CT序列的融合方法; 但是常用融合方法如加权平均法、 小波变换融合法等都是针对图像细节信息的优化, 不能表达组分的物理特性, 从而导致融合图像的灰度不具有物理表征性, 影响CT的定量检测。 为此, 结合具有物理表征特征的数据约束模型(DCM), 开展了基于先验组分的多谱CT序列DCM融合算法研究。 首先通过能谱滤波分离的成像方法获得多个能谱范围内的多能投影数据, 采用TV-OSEM算法重建不同能谱段的CT序列; 其次, 利用传统DCM模型和改进DCM模型分别对多谱CT序列进行融合, 传统DCM模型是严格单能的, 由于滤波后能谱的非严格单能特性, 其融合结果不能表征出对象序列中的全部组分。 针对此问题提出了改进DCM模型。 改进DCM模型选择了新的体元定义, 并且在多谱CT序列融合中引入先验组分作为参照, 通过先验物质对融合结果中其他物质进行校准, 实现检测对象中各组分位置的准确分布。 仿真实验表明, 该方法可从物理表征正确性的角度, 实现多谱CT序列融合, 在满足CT序列中不同组分区分的同时, 其融合图像的灰度具有物理可参照性, 有利于后续的CT定量检测。

Multi-spectral CT imaging is to characterize the different components in the CT images of different spectral ranges. For more convenient displaying all components physical information in an image, it is necessary to study the fusion method of multi-spectral CT sequence. But the commonly used fusion methods, such as weighted average method and wavelet transform fusion method, are mainly for the optimization of image information. The physical properties of the components can not be expressed, so that the gray scale of the fused image without physical representation affects the quantitative detection of CT. A multi-spectral CT sequence DCM fusion algorithm based on priori components with physical characterization was presented. First, we got multi-spectrum projection sequence by Imaging method separated by energy spectrum filtering and the CT sequence with different energy spectrum can be obtained by TV-OSEM reconstruction algorithm. Second, the traditional DCM model and the improved DCM model were used to fuse the multi-spectral CT sequences. The traditional DCM model was strictly single energy, considering the non-strict monocular characteristic of the filtered energy spectrum. The fusion result can not accurately characterized all the components in the object sequence. To solve this problem, an improved DCM model was proposed. In the improved DCM model, a new voxel definition was selected and a metallic priori component was introduced as a reference substance in the multi-spectral CT sequence. The Prior component was used to calibrate other substances in the fusion results. Thus accurate distribution of each component in the CT sequence was achieved by calibration of the fusion results. Simulation result, the method can realize multi-spectral CT sequence fusion from the perspective of the physical representation. while satisfying the different components distinction of the CT sequence, the gray scale of its fused image had physical reference. This method is beneficial to the subsequent CT quantitative detection.