文章摘要

文章信息/Info

Title:: Fast Group Matching Method based on Singular Value Decomposition Decomposition for Magnetic Resonance Fingerprinting

作者:: 黄敏1△; 范玲玲1; 干博文2; 陈军波1; 1.中南民族大学生物医学工程学院，武汉 430074；2.华中科技大学光学与电子信息学院，武汉 430074

Author(s):: HUANG Min1; FAN Lingling1; GAN Bowen2; CHEN Junbo1; 1.College of Biomedical Engineering, South-central University for Nationalities, Wuhan 430074,China；2.School of Optical and Electronic Information, Huazhong University of Science & Technology, Wuhan 430074

Keywords:: Magnetic resonance fingerprinting; Singular value decomposition; Group matching; Dictionary; Parameter map

摘要:: 传统MRI一次扫描只能得到一种加权像，“磁共振指纹”成像新方法可同时获得组织的T1，T2及质子密度图像，但它需要采用全新的数据处理方式。直接匹配法将每个体素的信号与“字典”中的所有信号进行逐条匹配，得到参数值所需的时间很长。我们采用基于SVD的“磁共振指纹”快速分组匹配法来提高匹配效率。首先建立头部和模型“字典”，从“字典”选取一条时间信号，根据该信号与字典的相关系数对字典进行分组，以大幅压缩字典的大小；再对分组后的字典进行奇异值分解，提取字典的特征信息，加速匹配速度。采用大脑和模型数据进行测试的结果表明，该方法可以快速准确地得到各种组织参数图像。

Abstract:: The traditional MRI can only acquire a weighted map in a single scan. Magnetic resonance fingerprinting is a new technique requiring a novel data processing method. It can simultaneously generate quantitative maps of different tissue parameters, such as the relaxation time T1, T2 and proton density. The direct matching algorithm is to do inner-product operation between the observed signal and the signals of each entry in the dictionary. Though the direct matching algorithm has shown the MR parameters of interest accurately, matching time is very long. We put forward a more efficient method called fast group matching method based on the singular value decomposition (SVD). Two dictionaries with different sets of characteristic parameters of T1 and T2 were simulated based on the brain and the model. Then a random dictionary entry element that is normalized as an initial signal was choosen. According to the correlation coefficient from inner-product between the signal and all dictionary entries, dictionary could be assigned to different groups. SVD was made on each group to get feature information, which compresses the size of dictionary and speeding up the matching algorithm. The results of tests on the brain and model data show that multi tissue parametric maps can acquire accurately and efficiently.