基于未知输入的PET医学图像重构

Using unknown input to improve PET image reconstruction

正电子发射断层成像(PET)扫描技术是一项能够反映组织代谢水平的显像技术,对实体肿瘤的定性诊断和病灶转移的检查准确率较高。本文提出了一种PET图像的重建算法来提高PET图像的质量。考虑到大多已有重建算法的图像质量严重依赖于PET的线性测量模型,而该模型与实际情况不符,通过引入未知输入项,本文提出了一种新的测量模型。该未知输入项由2部分组成一部分是未知输入的输入矩阵,用以描述测量模型中投影矩阵的不确定性;另一部分是未知输入,用于刻画被干扰了的示踪剂浓度和一些未建模信息。在此新模型的基础上,利用最优估计理论,提出了一种不同于已有算法的重建算法,同时重构出了未知输入和示踪剂的浓度分布。最后通过实验验证了该算法比其他常用算法能更好地重建PET图像。

Positron emission tomography(PET)is a kind of imaging technique that can reflect the metabolic situation of tissues.It enjoys a higher accuracy for qualitative diagnosis and metastatic of cancers.A reconstruction algorithm is proposed to improve the quality of PET images.Given that the most existing reconstruction algorithms depends heavily on the linear measurement model of PET,which is impossible in the real situation,this paper starts with proposing a new measurement model by introducing an unknown input terms,which consists of two parts.One is the input matrix of the unknown input,which describes the uncertainty of the projection matrix in the measurement model;the other is the unknown input,which is regarded as the disturbed tracer concentration as well as some unmodeled information.Based on the proposed model,it utilizes the optimal estimation theory to present a reconstruction algorithm different from those existing,which can reconstruct the unknown input and the concentration distribution of the tracer simultaneously.Finally,the experimental results show that the image provided by the algorithm is superior to that generated by those typical algorithms.