CS-MRI中评价随机欠采样矩阵的新方法

发布时间：2018-04-08 12:02

本文选题：磁共振成像(MRI)　切入点：压缩感知　出处：《波谱学杂志》2015年04期

【摘要】：在压缩感知-磁共振成像(CS-MRI)中,随机欠采样矩阵与重建图像质量密切相关.而选取随机欠采样矩阵一般是通过计算点扩散函数(PSF),以可能产生的伪影的最大值为评价参数,评估欠采样对图像重建的影响,然而最大值只反应了伪影的最坏情况.该文引入了两种新的统计学评价参数平均值(MV)和标准差(SD),其中平均值评估了伪影的平均大小,标准差可以反映伪影的波动情况.该文分别使用这3种参数对小鼠和人体脑部MRI数据以不同的采样比率进行CS图像重建,实验结果表明,当采样比率不低于4倍稀疏度时,使用平均值获得了质量更优的重建图像.因此,通过稀疏度先验知识指导合理选取采样比率,并以平均值为评价参数选取随机欠采样矩阵,能够获得更优的CS-MRI重建图像.
[Abstract]:In compression perceptron magnetic resonance imaging (CS-MRI), the random under-sampling matrix is closely related to the quality of reconstructed images.The random under-sampling matrix is usually selected to evaluate the effect of under-sampling on image reconstruction by calculating the point diffusion function and using the maximum value of the artifact as the evaluation parameter. However, the maximum value only reflects the worst case of the artifact.In this paper, two new statistical evaluation parameters (MV) and standard deviation (SDV) are introduced, in which the average value of the artifact is evaluated, and the standard deviation can reflect the fluctuation of the artifact.In this paper, the MRI data of mouse and human brain were reconstructed with different sampling ratios using these three parameters. The experimental results showed that when the sampling ratio was not less than 4 times sparse,The reconstructed image with better quality is obtained by using the average value.Therefore, through the prior knowledge of sparsity to guide the reasonable selection of sampling ratio and the selection of random under-sampling matrix with the average value as the evaluation parameter, a better CS-MRI reconstruction image can be obtained.
【作者单位】：波谱与原子分子物理国家重点实验室中国科学院生物磁共振分析重点实验室武汉磁共振中心(中国科学院武汉物理与数学研究所);中国科学院大学;
【基金】：国家自然科学基金资助项目(81227902;11174327)
【分类号】：R445.2;O482.532

【参考文献】