从ML-EM 重建算法入手，分析了贝叶斯模型的一些关键点，针对采用传统方法求解MAP问题的局限性，提出一种用于正电子成像的贝叶斯神经网络重建算法，为了保留边缘信息，引入了二进制的保边缘变量，并应用共轭神经网络求解，模拟的重建结果表明，应用这种算法可以得到比ML-EM 算法更好的重建图像@-From the ML-EM reconstruction algorithm start with an analysis of Bayesian model some of the key points, the use of traditional methods for solving the problem of the limitations of MAP, a positron imaging for the Bayesian neural network algorithm, in order to to retain the edge information, the introduction of binary variables edge security, and application of neural network for solving the conjugate, the reconstruction of simulated results show that the algorithm can be more than ML-EM image reconstruction algorithm better @