基于CEST理论模型的优化采样方案研究
发布时间：2018-11-09 21:34
【摘要】：化学交换饱和转移成像技术(chemical exchange saturation transfer MRI contrast,CEST)是一种全新的核磁共振对比成像技术。CEST不同于传统的MR对比剂,它通过两种或多种处于不同化学环境下的质子之间的化学交换,产生成像对比。CEST的理论建模基于Bloch-Mc Connell方程。在实际的CEST成像中,通常在频率轴上等间隔施加预饱和脉冲序列,采集信号得到的z谱会和CEST的理论模型进行匹配,从而估计我们需要了解的生理参数。这种采集信号的方式被称为EDS(evenly distributed sampling)方式。EDS方式并不是最佳的采样方式。相关研究证明,在某些特定频率点上z谱信号对理论模型参数的微小变化具有相对更高的敏感度,如果在这些特定的频率点上多放置采样点,可以期待参数估计的精确度得到提高。基于以上认识,Y.K.Tee等研究者针对DIACEST提出了优化采样方案(osptimal sampling schedule,以下简称为OSS),并在DIACEST的双池模型上进行了仿真验证。结果表明,OSS下的参数估计精度相对于EDS确实有很大的提高。本文在此基础上,进行了以下几个方面的研究:第一,进一步改进了寻找OSS的算法。本文引入概率分布描述CEST重要模型参数的先验知识,提出了基于参数先验概率分布的平均归一化敏感度方程。第二,利用平均归一化敏感度方程提出了PARACEST双池模型和DIACEST三池模型的优化采样方案OSS。第三,通过仿真从更加全面的角度对OSS的性能作出分析。包括OSS性能与模型参数真实值分布的联系,OSS在不同噪声环境中的性能以及在采样点数目不断减少的情况下,OSS的性能变化趋势。第四,试图利用与Fisher信息矩阵相联系的Hessian矩阵行列式来预测OSS的性能表现。本研究发现:第一,OSS在CEST对比剂化学位移附近的采样点明显增加,与平均归一化敏感度曲线在此处的峰值相对应。第二,OSS的性能与CEST模型参数真实值的分布有密切联系,这主要是由于z谱信号在不同的模型参数真实值附近对模型参数的微小变化具有不同的敏感度。Hessian矩阵行列式可以定性地描述这种联系。第三,在仿真限定的条件下,OSS的性能在不同的噪声环境下相对于EDS都具有优越性。第四,OSS相对EDS更能抵抗采样点数目减少带来的影响。
[Abstract]:Chemical exchange saturation transfer imaging (chemical exchange saturation transfer MRI contrast,CEST) is a new nuclear magnetic resonance imaging technique. CEST is different from traditional MR contrast agents. It produces imaging contrast by chemical exchange between two or more protons in different chemical environments. The theoretical modeling of CEST is based on Bloch-Mc Connell equation. In actual CEST imaging, presaturation pulse sequences are usually applied at equal intervals on the frequency axis, and the Z spectrum of the collected signals will match with the theoretical model of CEST to estimate the physiological parameters we need to know. This way of collecting signals is called EDS (evenly distributed sampling). EDS is not the best sampling method. Studies have shown that z spectrum signals are more sensitive to minor changes in theoretical model parameters at certain frequency points, if sampling points are placed at these particular frequency points, The accuracy of parameter estimation can be expected to be improved. Based on the above understanding, Y.K.Tee and other researchers proposed an optimized sampling scheme for DIACEST. (osptimal sampling schedule, is referred to as OSS), for short, and the simulation is carried out on the double-cell model of DIACEST. The results show that the precision of parameter estimation under OSS is much higher than that of EDS. On this basis, this paper studies the following aspects: first, the algorithm of finding OSS is further improved. In this paper, we introduce a priori knowledge of probability distribution to describe the parameters of important CEST model, and propose an average normalized sensitivity equation based on a priori probability distribution of parameters. Secondly, using the average normalized sensitivity equation, an optimal sampling scheme, OSS., for PARACEST two-cell model and DIACEST three-cell model is proposed. Thirdly, the performance of OSS is analyzed from a more comprehensive point of view through simulation. Including the relationship between OSS performance and real value distribution of model parameters, the performance of OSS in different noise environments and the trend of OSS performance under the condition of decreasing number of sampling points. Fourthly, the determinant of Hessian matrix associated with Fisher information matrix is used to predict the performance of OSS. It is found that: first, the sampling points of OSS near the chemical shift of CEST contrast agent increase obviously, corresponding to the peak value of the average normalized sensitivity curve here. Secondly, the performance of OSS is closely related to the distribution of real values of CEST model parameters. This is mainly due to the fact that the z spectrum signal has different sensitivity to the slight variation of the model parameters near the real values of different model parameters. The Hessian matrix determinant can describe this relation qualitatively. Thirdly, under the limited conditions of simulation, the performance of OSS is superior to that of EDS in different noise environments. Fourth, OSS is more resistant to the impact of a decrease in the number of sampling points than EDS.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R445.2
本文编号：2321617
[Abstract]:Chemical exchange saturation transfer imaging (chemical exchange saturation transfer MRI contrast,CEST) is a new nuclear magnetic resonance imaging technique. CEST is different from traditional MR contrast agents. It produces imaging contrast by chemical exchange between two or more protons in different chemical environments. The theoretical modeling of CEST is based on Bloch-Mc Connell equation. In actual CEST imaging, presaturation pulse sequences are usually applied at equal intervals on the frequency axis, and the Z spectrum of the collected signals will match with the theoretical model of CEST to estimate the physiological parameters we need to know. This way of collecting signals is called EDS (evenly distributed sampling). EDS is not the best sampling method. Studies have shown that z spectrum signals are more sensitive to minor changes in theoretical model parameters at certain frequency points, if sampling points are placed at these particular frequency points, The accuracy of parameter estimation can be expected to be improved. Based on the above understanding, Y.K.Tee and other researchers proposed an optimized sampling scheme for DIACEST. (osptimal sampling schedule, is referred to as OSS), for short, and the simulation is carried out on the double-cell model of DIACEST. The results show that the precision of parameter estimation under OSS is much higher than that of EDS. On this basis, this paper studies the following aspects: first, the algorithm of finding OSS is further improved. In this paper, we introduce a priori knowledge of probability distribution to describe the parameters of important CEST model, and propose an average normalized sensitivity equation based on a priori probability distribution of parameters. Secondly, using the average normalized sensitivity equation, an optimal sampling scheme, OSS., for PARACEST two-cell model and DIACEST three-cell model is proposed. Thirdly, the performance of OSS is analyzed from a more comprehensive point of view through simulation. Including the relationship between OSS performance and real value distribution of model parameters, the performance of OSS in different noise environments and the trend of OSS performance under the condition of decreasing number of sampling points. Fourthly, the determinant of Hessian matrix associated with Fisher information matrix is used to predict the performance of OSS. It is found that: first, the sampling points of OSS near the chemical shift of CEST contrast agent increase obviously, corresponding to the peak value of the average normalized sensitivity curve here. Secondly, the performance of OSS is closely related to the distribution of real values of CEST model parameters. This is mainly due to the fact that the z spectrum signal has different sensitivity to the slight variation of the model parameters near the real values of different model parameters. The Hessian matrix determinant can describe this relation qualitatively. Thirdly, under the limited conditions of simulation, the performance of OSS is superior to that of EDS in different noise environments. Fourth, OSS is more resistant to the impact of a decrease in the number of sampling points than EDS.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R445.2
【共引文献】
相关博士学位论文 前3条
1 卢建华;化学交换饱和转移及其在MRI中的应用[D];厦门大学;2014年
2 洪晓华;用酰胺质子转移磁共振成像技术评价鼠脑胶质瘤放疗疗效[D];华中科技大学;2014年
3 朱筱磊;环境响应型CEST MRI造影剂的构建及其性质研究[D];中国科学院研究生院(武汉物理与数学研究所);2015年
，本文编号：2321617
本文链接：https://www.wllwen.com/yixuelunwen/fangshe/2321617.html
