文章摘要

文章信息/Info

Title:: Magnetic resonance fingerprint optimization algorithm based on extended Kalman filter

Author(s):: HUANG Min; ZHANG Lujian; ZHOU Dao; CHEN Junbo; College of Biomedical Engineering, South-Central University for Nationalities, Wuhan 430074, China

Keywords:: Magnetic resonance fingerprinting; Extended Kalman filter; Dictionary generation; Bloch equation; Parameter quantification

摘要:: “磁共振指纹”技术一次扫描可同时获得定量的T1、T2等参数像，弥补了磁共振成像（MRI）扫描时间长，一次只能得到一种加权像的缺陷。但其经典算法需要对不同序列建立不同的字典，且字典大小直接影响量化结果。我们对基于扩展卡尔曼滤波的磁共振指纹参数量化算法进行优化，不需要建立字典，通过建立成像模型，采用卡尔曼滤波对磁共振指纹信号进行跟踪，反演得到量化参数。该方法减少了卡尔曼迭代次数，加快了参数估计的收敛速度，提升了参数量化图像的质量。

Abstract:: Magnetic resonance fingerprint（MRF） technology can get quantitative T1 and T2 images simultaneously by one scan. It remedies the defects of the traditional MRI with long scanning time and only one weighted image at one MRI scan.However, the classical algorithm needs to establish different dictionaries for different sequences,and the size of the dictionary directly affects the parameter quantization results. We optimized the extended Kalman filter algorithm for MRF. The MRF signals were tracked by Kalman filter without dictionary. Parameters were obtained by back-extrapolation.This method reduces the number of Kalman iteration steps and increases the speed of the convergence for estimation. It effectively improves the quality of parameter quantification maps.