磁共振成像二维相位解缠方法研究
发布时间：2018-07-25 15:52
【摘要】：磁共振成像得到的信号是复数形式,包含幅度和相位。常规诊断中,往往用到的是磁共振的幅度图像。然而,其相位也包含了大量信息,例如自旋原子核的移动速度、磁场的不均匀度和磁化率变化等。因此磁共振相位可以用来估计主磁场均匀性和获取临床相关的生理参数。 但从复数信号中提取真实相位时,相位值却会被限制在(-π,π)弧度区间内,位于该区间外的真实相位被缠绕到这一区间内。此现象称为相位缠绕,得到的相位称为缠绕相位。从缠绕相位恢复真实相位的过程就叫做相位解缠。噪声、欠采样和物体不连续的存在使相位解缠变得困难。 本文提出了三种新的二维相位解缠方法。模拟数据和实际磁共振相位数据被用来评估这些方法的表现。 第一种方法是基于离散粒子群优化算法的枝切线法。这种方法先将整幅图像的残差分成几组；在每组内使用离散粒子群优化算法对正负极性残差进行配对；用枝切线连接每组内配好对的正负极性残差；最后绕过这些枝切线进行相位解缠。 与最新的基于人工智能的枝切法对比,这种方法能在较短时间内得到合理的枝切线连接,且通过对残差分组进一步降低了计算时间。 第二种相位解缠方法是基于直接求解法的加权最小Lp范数法。它将整个相位图像的解缠相位梯度与缠绕相位梯度之间差值的加权Lp范数作为优化目标函数；将这个目标函数转化成一个方程组,其系数矩阵采用稀疏结构储存和表达；最后使用直接求解法求解方程组。由于方程组的系数矩阵与解缠相位有关,因此采取迭代方式得到最终的解缠结果。 与一些常用方法相比,这种方法能有效减少计算时间且解缠结果更准确。 第三种方法是基于掩码的区域增长法。这种方法采用一种新的掩码提取方式将残差合理地连接起来作为掩码中的零点；将掩码与相位导数方差结合构成最终的质量图,这样连接残差经过的点均被当成零质量(也就是质量最差)的点,会被滞留到最后才被相位解缠；接着根据质量图将整幅图像分成多个区域,在每个区域内单独进行相位解缠,其中质量最差的那个区域从多个方向进行相位加权平均；最后将多个区域融合在一起。 与最新的区域增长相位解缠方法(PHUN)相比,这种方法运算速度快并能够限制误差的传播。
[Abstract]:The signal obtained by magnetic resonance imaging is complex and contains amplitude and phase. In conventional diagnosis, the amplitude image of magnetic resonance is often used. However, its phase also contains a lot of information, such as the moving speed of spin nuclei, the variation of magnetic field inhomogeneity and magnetic susceptibility, etc. Therefore, the magnetic resonance phase can be used to estimate the homogeneity of the main magnetic field and to obtain clinical physiological parameters. However, when the real phase is extracted from the complex signal, the phase value will be limited to (- 蟺, 蟺) radians, and the real phase located outside the complex signal will be entangled in this region. This phenomenon is called phase winding, and the resulting phase is called winding phase. The process of recovering the real phase from the winding phase is called phase unwrapping. The presence of noise, undersampling and discontinuity makes phase unwrapping difficult. In this paper, three new two-dimensional phase unwrapping methods are proposed. Analog data and actual magnetic resonance phase data are used to evaluate the performance of these methods. The first method is branch tangent method based on discrete particle swarm optimization. In this method, the residuals of the whole image are divided into several groups, the discrete particle swarm optimization algorithm is used to match the positive and negative pole residuals in each group, the branch tangent is used to connect the positive and negative pole residuals in each group. Finally, phase unwrapping is carried out by bypassing the tangent of these branches. Compared with the new branch cutting method based on artificial intelligence, this method can get reasonable branch tangent connection in a short time, and further reduce the computing time by grouping the residual error. The second phase unwrapping method is the weighted least LP norm method based on the direct solution method. The weighted LP norm of the difference between the unwrapped phase gradient and the winding phase gradient of the whole phase image is taken as the optimization objective function, the objective function is transformed into a set of equations, and the coefficient matrix is stored and expressed by sparse structure. Finally, the direct solution method is used to solve the equations. Because the coefficient matrix of the equations is related to the unwrapping phase, the final unwrapping results are obtained by iterative method. Compared with some commonly used methods, this method can effectively reduce the calculation time and the unwrapping result is more accurate. The third method is the region growth method based on mask. In this method, a new mask extraction method is used to reasonably connect the residuals as the zeros in the mask, and the mask and the variance of the phase derivative are combined to form the final quality map. In this way, the points through which the link residuals are passed are treated as zero mass (that is, the worst quality points), and they are detained until the last phase unwrapping; then, according to the quality map, the whole image is divided into multiple regions, and the phase unwrapping is carried out separately in each region. The region with the lowest quality is weighted from multiple directions, and the multiple regions are fused together. Compared with the new region growth phase unwrapping method (PHUN), this method is faster and can limit the propagation of errors.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：R310
本文编号：2144306
[Abstract]:The signal obtained by magnetic resonance imaging is complex and contains amplitude and phase. In conventional diagnosis, the amplitude image of magnetic resonance is often used. However, its phase also contains a lot of information, such as the moving speed of spin nuclei, the variation of magnetic field inhomogeneity and magnetic susceptibility, etc. Therefore, the magnetic resonance phase can be used to estimate the homogeneity of the main magnetic field and to obtain clinical physiological parameters. However, when the real phase is extracted from the complex signal, the phase value will be limited to (- 蟺, 蟺) radians, and the real phase located outside the complex signal will be entangled in this region. This phenomenon is called phase winding, and the resulting phase is called winding phase. The process of recovering the real phase from the winding phase is called phase unwrapping. The presence of noise, undersampling and discontinuity makes phase unwrapping difficult. In this paper, three new two-dimensional phase unwrapping methods are proposed. Analog data and actual magnetic resonance phase data are used to evaluate the performance of these methods. The first method is branch tangent method based on discrete particle swarm optimization. In this method, the residuals of the whole image are divided into several groups, the discrete particle swarm optimization algorithm is used to match the positive and negative pole residuals in each group, the branch tangent is used to connect the positive and negative pole residuals in each group. Finally, phase unwrapping is carried out by bypassing the tangent of these branches. Compared with the new branch cutting method based on artificial intelligence, this method can get reasonable branch tangent connection in a short time, and further reduce the computing time by grouping the residual error. The second phase unwrapping method is the weighted least LP norm method based on the direct solution method. The weighted LP norm of the difference between the unwrapped phase gradient and the winding phase gradient of the whole phase image is taken as the optimization objective function, the objective function is transformed into a set of equations, and the coefficient matrix is stored and expressed by sparse structure. Finally, the direct solution method is used to solve the equations. Because the coefficient matrix of the equations is related to the unwrapping phase, the final unwrapping results are obtained by iterative method. Compared with some commonly used methods, this method can effectively reduce the calculation time and the unwrapping result is more accurate. The third method is the region growth method based on mask. In this method, a new mask extraction method is used to reasonably connect the residuals as the zeros in the mask, and the mask and the variance of the phase derivative are combined to form the final quality map. In this way, the points through which the link residuals are passed are treated as zero mass (that is, the worst quality points), and they are detained until the last phase unwrapping; then, according to the quality map, the whole image is divided into multiple regions, and the phase unwrapping is carried out separately in each region. The region with the lowest quality is weighted from multiple directions, and the multiple regions are fused together. Compared with the new region growth phase unwrapping method (PHUN), this method is faster and can limit the propagation of errors.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：R310
【参考文献】
相关期刊论文 前2条
1 杨维,李歧强;粒子群优化算法综述[J];中国工程科学;2004年05期
2 刘亚涛;俎栋林;包尚联;;水、脂分离磁共振成像Dixon方法[J];中国医学物理学杂志;2012年06期
，本文编号：2144306
本文链接：https://www.wllwen.com/yixuelunwen/swyx/2144306.html
