光子计数X射线真彩色CT关键技术研究

Photon-counting based x-ray spectral/true-color CT plays an important role in medical diagnosis, industry detection, security and other fields, because it can provide color images to increase contrast resolution and material specificity. However, it is difficult for the current photon counting detector to balance the energy resolution and the image noise within each bin, and this limits the applications of spectral CT. To overcome this limitation, this project makes efforts in some aspects, such as energy bins division, noise reduction and color representation. First, an optimal model for energy bin division is set up by studying relationship between the noise and the number, width and position of the energy bins. Second, multi-constraint based reconstruction algorithm is built for narrow spectral CT through the research of narrow spectrum noise statistical characteristics, combining with compressed sensing theory, prior information utilization and the goals of image reconstruction. Third, the true-color representation of material components is achieved by establishing the relationship between the spectral information and RGB color channel, combined with spectrum evaluation and compensation. This project can help to improve the material discrimination of spectral CT and satisfy the requirements of clinical applications for medical CT and the engineering demand for industrial CT. It also is of great significance to promote the development of true-color CT and the development of quantitative CT detection technology in China.

基于光子计数探测技术的X射线多能CT，能够提供彩色图像，具有较高的图像对比度和材料组分的识别能力,在医学诊断、工业检测和安全检查等领域发挥重要作用。然而，目前的光子计数探测器很难平衡系统的能量分辨率和能量通道内的噪声，限制了谱CT的应用。针对此问题，本项目从能量通道划分，噪声抑制和彩色表征等几个方面进行突破。首先，基于蒙特卡罗模拟方法，通过研究能量通道的数量、宽度、位置与噪声的关系，建立最优能量通道的划分模型；其次，在研究窄谱噪声统计性质的基础上，结合压缩感知、先验信息利用和重建目标等，构建多约束窄谱CT重建算法模型；进一步研究谱信息与RGB色彩通道的模型关系，结合能谱评估与补偿，实现材料组分的真彩色表征。项目的研究能够提高谱CT系统的材料辨别能力，满足医学CT临床应用和工业CT工程化的需求，对推动我国彩色CT和定量CT技术的发展具有重要意义。

基于光子计数探测器的X射线多能CT可以提高图像对比度和材料组分的识别能力，对推动我国彩色CT和定量CT技术的研究具有重要意义。本项目针对光子计数探测器难以平衡系统的能量分辨率和能量通道内噪声的问题，对光子计数X射线真彩色CT关键技术展开研究。通过搭建蒙特卡罗模拟平台，模拟不同能量阈值和宽度下的投影过程，分析高斯、瑞丽混合噪声与能谱划分的关系。基于压缩感知框架，构建多约束窄谱CT降噪重建算法，研究稀疏先验以及能谱先验对重建算法的影响，仿真实验和实际实验结果验证了提出算法的有效性和可行性。推导了能谱CT解析重建算法，为能谱CT重建算法的评价提供了一个比较的平台。通过多能投影域分解的方法，实现了造影剂物质的识别，同时在降噪能谱CT重建算法的基础上，利用主成分分析和多能CT最小二乘分割的方法实现了能谱CT的彩色表征。本课题在国内外重要学术期刊及会议上发表论文15篇，其中SCI/EI期刊论文6篇，会议论文1篇，申请专利2项。