结构先验约束的动态PET图像重建

Dynamic PET image reconstruction with Geometrical structure prior constraints

为提高图像质量,提出一种利用其他高质量解剖模板作为先验,引导动态正电子发射断层(PET)成像重建的方法.该方法基于状态空间理论体系,将与生理特性相关的解剖模板耦合于状态方程中,并与成像模型相结合构成完善的状态空间方程组,运用具有鲁棒性的H∞滤波法求解,从而构建一种适合于符合计数率低、噪声影响显著的动态PET图像的重建框架.蒙特卡洛模拟实验结果表明,与其他传统方法相比,本方法在能够适应实际动态PET成像中统计特性和系统特性不确定的基础上,进一步抑制了噪声,并保持了图像边缘和细节信息.

In order to improve the image quality,an improved algorithm for dynamic positron emission tomography（PET） image reconstruction was proposed by using segmented anatomical template that provided by other high quality imaging technology.Based on state space theory,a dynamic PET image reconstruction framework for low count rate and high noise environment was formulated with the observation equation of detectors and a modified evolution equation incorporating structural constraint which was generated to guide the reconstruction process,and H∞ filtering principle was employed to solve the above two equations.Compared with other algorithms,experiments conducted by Monte Carlo simulations indicate a persuasive assessment that the proposed strategy was particularly applicable for real-world situations with the uncertainties of system and statistical properties,suppresses noise well,while the boundary information and other details remain clear.