基于增量优化传递的正电子发射断层图像重建

PET image reconstruction algorithm by incremental optimization transfer

将增量优化传递算法用于正电子发射断层图像重建中,得到一种新的最大后验概率图像重建算法,实验结果表明,相对于滤波反投影和最大似然期望值最大化算法,有序子集可分离的抛物面型替代函数算法和发射增量优化传递算法得到的图像归一化均方误差较小,且边界得到一定的保持.在迭代的初期,有序子集可分离的抛物面型替代函数算法收敛较快,误差小;当迭代到一定的次数时,发射增量优化传递算法误差小,且能够有效收敛,对发射扫描断层图像重建产生重要的影响.

A new maximum a posterior probability algorithm, named as emission incremental optimization transfer （EMIOT）, was obtained by introducing incremental optimization transfer method into positron emission tomography （PET） image reconstruction. The results show that the images reconstructed by ordered subsets separable paraboloidal surrogates （OS-SPS） and EMIOT are better than those by traditional image reconstruction algorithm such as filtered back projection （FBP） and maximum likelihood expectation maximization （ML-EM）, in terms of normalized root square error and edge-preserving ability. At the early iteration, OS-SPS showed better convergence, however, after certain iterations, EMIOT is better than OS-SPS. Therefore, EMIOT has important influence on emission image reconstruction.