基于稀疏重建的磁共振图像尖峰噪声消除方法

磁共振图像K空间中的尖峰噪声会严重影响图像质量.该文在磁共振图像压缩感知的共轭梯度重建法的基础上,提出一种新的利用磁共振图像稀疏性进行尖峰噪声修复的方法.传统的共轭梯度重建是通过小波域迭代进行的,对于K空间的尖峰噪声的消除不是最适合.首先提出压缩感知的K空间重建算法,该算法与小波域重建等效.在此基础上,提出可以较好地修复尖峰噪声的K空间部分重建算法.即在迭代过程中,以图像的稀疏性作为约束条件,仅修改尖峰噪声所遮盖区域的数据,其他位置的数据保持不变.该算法与传统的插值算法及共轭梯度算法相比,能够更好地修复K空间尖峰噪声点,减少图像伪影,同时降低了对尖峰噪声定位准确性的要求.

Spike Noise Removal for Magnetic Resonance Imaging Based on Sparse Reconstruction

It is well-known that, in magnetic resonance imaging (MRI), the presence of spike noise in the K-space will degrade the quality of reconstructed images. In this work, we proposed a method to remove spike noises based on the non-linear conjugate gradient (NLCG) reconstruction algorithm of compressed sensing (CS). The traditional CG algorithm reconstructs images in the wave-domain, making it difficult to remove spike noises. The proposed algorithm is a partial K-space reconstruction algorithm. Using image sparsity as a restrain, the algorithm reconstructs only the data which is covered by spike noises. Compared with the interpolation and NLCG algorithm, the proposed algorithm was shown to yield better images with less artifacts without the need to know the accurate localization of the spike noises.