基于小波和各向异性扩散的PET图像MLEM重建算法

MLEM RECONSTRUCTION ALGORITHM FOR PET IMAGE BASED ON WAVELET AND ANISOTROPIC DIFFUSION

在正电子发射断层成像中,经典的MLEM(Maximum Likelihood Expectation Maximization)算法具有收敛速度慢、不能有效抑制噪声的不足。为了解决该问题,通常在迭代过程中加入正则项来改善MLEM的重建性能。提出一种新的基于小波收缩和各向异性扩散的去噪算法,将该算法与MLEM算法结合起来形成一种新的PET(Positron Emission Tomography)重建方法。实验结果表明,该算法在降低复杂性、保持较高收敛速度的同时,能获得较高的信噪比和较好的图像视觉效果。

In positron emission tomography （PET） imaging, traditional maximum likelihood expectation maximisation （MLME） algorithm has the deficiencies of converging slowly and can not suppress noise effectively. To address this problem, usually the regularisation term would be introduced to iterative process to improve the reconstruction performance of MLEM. In this paper, we propose a new denoising algorithm which is based on wavelet shrinkage and anisotropic diffusion, and combine this algorithm with the MLEM algorithm to form a novel PET reconstruction method. Experimental results show that this algorithm can obtain higher SNR and superior visual effect on images while reducing the complexity and keeping higher convergence rate.