To better use the phase information to compensate the influence of blood flow, the phase unwrapping problem in susceptibility weighted imaging (SWI) is studied in this paper. In order to improve the accuracy of unwrapping, this paper proposes a magnitude image-guided phase unwrapping algorithm of SWI. The basic idea is as follows: (1) reduce the influence of noise by improving the rotational invariant non-local principal component analysis method (PRI-NL-PCA); (2) extract the corresponding solid region in the phase image to avoid the influence of the background noise on the phase unwrapping method; (3) use the phase compensation method to constrain the phase image reconstructed by the K-space. Finally, the reliability of the unwrapping method is evaluated by using four kinds of statistics as quantification index: the number, mean (M), variance (Var), and positive percentage (Pos) and negative percentage (Neg) of phasic error points. By comparing the simulated data with 226 sets of true head SWI data, the results show that the proposed algorithm has high accuracy compared with the classical branch cut method and the least squares method.
为了更好地利用相位信息补偿血流造成的影响，本文对磁敏感加权成像（SWI）中存在的相位缠绕问题展开了研究。为提高解缠绕的准确性，本文提出了幅度图像引导的磁敏感加权图像相位解缠算法。基本思路如下：① 通过改进旋转不变非局部主成分分析滤波（PRI-NL-PCA）降低噪声影响；② 结合 C-V 模型水平集提取相位图像中对应的实性组织区域，从而规避背景噪声对相位解缠方法的影响；③ 采用相位补偿的方法约束 K 空间重建出的相位图像。最后，利用四种统计量作为量化指标，评价解缠绕方法的可靠性：相位误差的突变点个数、均值（M）、方差（Var），以及阳性百分比（Pos）和阴性百分比（Neg）。通过对比仿真数据和 226 组真实头部磁敏感数据，结果表明，本文算法相对于经典的枝切法和最小二乘法，解缠绕结果具有较高的准确性。.
Keywords: magnitude image; phase error; phase unwrapping; susceptibility weighted imaging.