一种基于指数型先验分布的正电子发射断层图像重建算法**★◆

背景：最大似然估计算法是正电子发射断层图像重建的经典算法，能够在信息量不足的情况下获得分辨率和噪声特性均优于滤波反投影重建的重建结果。但是MLEM算法具有不稳定性，即随迭代次数的增加，图像噪声反而会增加。 目的：针对MLEM算法的图像噪声问题，提出一种基于指数型先验分布约束的MAP重建算法。 方法：将指数先验分布代替传统MAP重建中的高斯先验，并用信噪比和归一化均方误差来判断重建质量。 结果与结论：实验证明，该算法不仅能够抑制噪声，而且能够保持重建图像的边缘，不会造成过分平滑。

Positron emission tomography image reconstruction algorithm based on an exponential Markov random field prior model

BACKGROUND: Maximum likelihood expectation maximization (MLEM) algorithm, the classical algorithm in PET reconstruction, is superior to filtered back projection reconstruction with its better performance in resolution and noise characteristic. However, with the increasing iterations, the noisy influence of reconstruction image increases. OBJECTIVE: To propose a maximum a posteriori (MAP) reconstruction algorithm based on exponential prior distribution for noise suppression METHODS: Exponent prior distribution replaces the Gaussian of traditional MAP, and the reconstruction image is tested with signal-to-noise and root mean squared error. RESULTS AND CONCLUSION: Results show that the proposed method performed well for noise suppression, and preferably keep the edges of reconstruction image without excessive smoothing.