基于仿真脑磁共振数据的信噪比在医学影像质量量化上的一致性研究

A Consistency Evaluation of Signal-to-Noise Ratio in Medical Image Quality Assessment:A Simulation Study on Human Brain Magnetic Resonance Images

医学成像过程复杂,且影响成像质量的因素非常多,导致医学影像质量的量化一直非常具有挑战性。信噪比是一种被广泛采用的医学影像质量量化方法,被定义为感兴趣区域的信号均值与背景噪声的标准差的商。它的计算和执行勾画的操作者及勾画的区域密切相关,因此有必要对其一致性进行系统的研究。文章共计使用了324例仿真的磁共振脑影像,分为正常脑和多发性硬化病变两类影像;并将脑脊液区域和白质区域作为感兴趣区域;最后采用Wilcoxon秩和检验来检验操作者间的差异性,并用Pearson相关系数和Spearman秩相关系数来验证感兴趣区域间的一致性。实验结果发现,对相同模态的相同组织区域,不同观察者间的信噪比值没有显著性差异(Wilcoxon秩和检验,P>0.70);对相同模态的相同观察者,Pearson相关系数rp>0.71(P<10^(-5)),而Spearman秩相关系数rs>0.97(P<10-3),即信噪比对磁共振影像质量的评价不受操作者以及感兴趣组织的选择而出现差异。进一步的临床数据验证有利于挖掘信噪比在临床应用中的使用条件和潜在局限。

Medical imaging is a complicated procedure and acquired images are with intrinsic characteristics. As a tool to quantify the image quality, signal-to-noise ratio（SNR） is widely accepted by physicians in clinical situations. It is defined as the quotient of the mean signal intensity in a tissue region of interest and the standard deviation of the signal intensity in a region outside the anatomy of the object imaged. However, insufficient knowledge on its consistency with respect to different observers and tissue regions is known. In this paper, the consistency is studied with 324 simulated MR images of human brain. The consistency of SNR is validated between two observers and between tissues of white matter and cerebral spinal fluid. For the same type of tissues in each modality, Wilcoxon rank sum test suggests no significant difference between two observers（P0.70）. For the same modality and observer, SNR between tissues correlates well（Pearson correlation coefficient rp0.71（P10（-5））, and Spearman＇s rank correlation coefficient rs0.97（P10（-3））. This study indicates that SNR is consistent and robust regarding to different observers and tissues in objective quality assessment of magnetic resonance images. Further research will be carried on clinical images for objective assessment.