基于间隔采样和区域增长的磁共振水脂分离方法研究
发布时间：2018-08-19 09:39
【摘要】：磁共振成像作为主要的医学影像技术之一,为辅助临床诊断治疗提供重要的图像信息和量化信息。由于脂肪在磁共振成像中长T2,短T1的特性,导致脂肪在磁共振图像中是高亮信号,进而会影响图像对比度,并可能会掩盖一些潜在的病变,从而影响临床诊断。因此,在磁共振图像中抑制脂肪信号有很大的意义。Dixon技术不仅能够抑制脂肪信号,同时还能得到脂肪图。两点Dixon技术因扫描效率较高、并能灵活选择回波时间,所以在临床上有较广泛的应用。然而在图像噪声较大、存在伪影和运动等情况下,基于区域增长算法的两点Dixon技术会因为误差传播和累积导致最终的水和脂肪分离发生错误。本文提出了一种基于间隔采样和区域增长的两点Dixon水脂分离新方法,能有效减少误差传播和累积,得到更准确的水脂分离结果。提出方法的主要步骤为:首先对求得的场向量图进行间隔采样获得四对子向量图,目的是降低突变相位对结果的影响;然后对四对子向量图分别运用区域增长算法,得到四幅子向量图;随后再对四幅子向量图进行平滑校正操作,使子向量图的空间相位更加准确;最后,把四幅子向量图合并校正得到最终的需要求解的向量图。本文方法的创新之处在于:1.利用间隔采样操作同时得到四幅子图像,在后续处理中可相互约束并且可利用的图像信息更加丰富;2.通过间隔采样,把突变相位分配到四幅子图,使得四幅子图的相同区域同时出现突变相位的几率变低;3.区域增长算法和间隔采样相结合,使得因突变相位导致的误差传播和累积可限制在单幅子图中,降低了误差传播对全局相位估计的影响;4.利用子图像的独立性可进行并行运算,减少算法运行时间。本文分别用了仿真实验和多组临床数据对算法进行验证。仿真相位的实验结果表明,在存在多处相位突变的情况下,本文提出方法发生错误的像素点明显比原区域增长算法要少。噪声仿真实验结果表明本文方法能降低噪声对水脂分离结果的影响。真实数据水脂分离结果表明本文提出方法更稳定和准确。算法运行时间是原基于区域增长的两点Dixon方法的三分之一左右,该优势在图像数据矩阵较大时将更明显。
[Abstract]:As one of the main medical imaging techniques, magnetic resonance imaging (MRI) provides important image information and quantitative information for clinical diagnosis and treatment. Because of the long T _ 2 and short T _ 1 characteristics of fat in magnetic resonance imaging, fat is a highlight signal in magnetic resonance imaging, which will affect the contrast of the image, and may cover up some potential lesions, thus affecting the clinical diagnosis. Therefore, it is of great significance to suppress fat signal in magnetic resonance imaging. Dixon technique can not only suppress fat signal, but also obtain fat map. Two-point Dixon is widely used in clinic because of its high scanning efficiency and flexible choice of echo time. However, in the case of large image noise, artifact and motion, the error in the final separation of water and fat will occur due to the error propagation and accumulation of the two-point Dixon technique based on the region growth algorithm. In this paper, a new two-point Dixon water-lipid separation method based on interval sampling and regional growth is proposed, which can effectively reduce the error propagation and accumulation, and obtain more accurate results of water-lipid separation. The main steps of the proposed method are as follows: firstly, four pairs of subvector graphs are obtained by interval sampling of the obtained field vector graphs, the purpose of which is to reduce the effect of the abrupt phase on the results, and then the region growth algorithm is applied to the four pairs of subvector graphs. Four subvector graphs are obtained, and then the four subvector graphs are smoothed and corrected to make the spatial phase of the subvector graphs more accurate. Finally, the four subvector graphs are combined and corrected to obtain the final vector graphs that need to be solved. The innovation of the method of this paper lies in: 1. Four sub-images can be obtained simultaneously by interval sampling operation, which can constrain each other and enrich the available image information in subsequent processing. Through interval sampling, the abrupt phase is allocated to four subgraphs, which makes the probability of abrupt phase appearing in the same region of the four subgraphs to be lower. The combination of region growth algorithm and interval sampling makes the error propagation and accumulation caused by the abrupt phase can be restricted to a single subgraph, which reduces the influence of error propagation on the global phase estimation. Using the independence of the subimage, the parallel operation can be carried out, and the running time of the algorithm can be reduced. In this paper, simulation experiments and multiple sets of clinical data are used to verify the algorithm. The experimental results show that in the presence of multiple phase mutations, the number of pixels with errors in the proposed method is obviously less than that of the original region growth algorithm. The results of noise simulation show that the proposed method can reduce the effect of noise on the separation of water and lipid. The real data show that the proposed method is more stable and accurate. The running time of the algorithm is about 1/3 of the original two-point Dixon method based on region growth, which will be more obvious when the image data matrix is large.
【学位授予单位】：南方医科大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;R445.2
本文编号：2191276
[Abstract]:As one of the main medical imaging techniques, magnetic resonance imaging (MRI) provides important image information and quantitative information for clinical diagnosis and treatment. Because of the long T _ 2 and short T _ 1 characteristics of fat in magnetic resonance imaging, fat is a highlight signal in magnetic resonance imaging, which will affect the contrast of the image, and may cover up some potential lesions, thus affecting the clinical diagnosis. Therefore, it is of great significance to suppress fat signal in magnetic resonance imaging. Dixon technique can not only suppress fat signal, but also obtain fat map. Two-point Dixon is widely used in clinic because of its high scanning efficiency and flexible choice of echo time. However, in the case of large image noise, artifact and motion, the error in the final separation of water and fat will occur due to the error propagation and accumulation of the two-point Dixon technique based on the region growth algorithm. In this paper, a new two-point Dixon water-lipid separation method based on interval sampling and regional growth is proposed, which can effectively reduce the error propagation and accumulation, and obtain more accurate results of water-lipid separation. The main steps of the proposed method are as follows: firstly, four pairs of subvector graphs are obtained by interval sampling of the obtained field vector graphs, the purpose of which is to reduce the effect of the abrupt phase on the results, and then the region growth algorithm is applied to the four pairs of subvector graphs. Four subvector graphs are obtained, and then the four subvector graphs are smoothed and corrected to make the spatial phase of the subvector graphs more accurate. Finally, the four subvector graphs are combined and corrected to obtain the final vector graphs that need to be solved. The innovation of the method of this paper lies in: 1. Four sub-images can be obtained simultaneously by interval sampling operation, which can constrain each other and enrich the available image information in subsequent processing. Through interval sampling, the abrupt phase is allocated to four subgraphs, which makes the probability of abrupt phase appearing in the same region of the four subgraphs to be lower. The combination of region growth algorithm and interval sampling makes the error propagation and accumulation caused by the abrupt phase can be restricted to a single subgraph, which reduces the influence of error propagation on the global phase estimation. Using the independence of the subimage, the parallel operation can be carried out, and the running time of the algorithm can be reduced. In this paper, simulation experiments and multiple sets of clinical data are used to verify the algorithm. The experimental results show that in the presence of multiple phase mutations, the number of pixels with errors in the proposed method is obviously less than that of the original region growth algorithm. The results of noise simulation show that the proposed method can reduce the effect of noise on the separation of water and lipid. The real data show that the proposed method is more stable and accurate. The running time of the algorithm is about 1/3 of the original two-point Dixon method based on region growth, which will be more obvious when the image data matrix is large.
【学位授予单位】：南方医科大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;R445.2
【参考文献】
相关期刊论文 前2条
1 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
2 马旭东;以无隙而入有间——浅谈磁共振成像系统[J];国外科技动态;2001年09期
，本文编号：2191276
本文链接：https://www.wllwen.com/yixuelunwen/fangshe/2191276.html
