TGV在部分磁共振图像重建中的应用

发布时间：2018-04-11 23:01

本文选题：磁共振成像 + 部分k空间　；参考：《上海交通大学》2012年硕士论文

【摘要】：磁共振成像(Magnetic Resonance Imaging, MRI)是一种新型的，非侵入的成像方式。在临床应用中，由于磁共振具有很好的软组织对比度、成像参数多、三维空间的分辨能力高、可任意方向断层、不会产生致电离辐射的内在特点以及易为患者接受等优点[1][2]，而越来越受欢迎，得到了广泛的应用。当代磁共振成像应用对空间分辨率(Spatial Resolution)，瞬时清晰度(Temporal Resolution)，信噪比(Signal-to-Noise Ratio, SNR)及成像速度有相当高的要求[3][4]。然而常规的磁共振成像时间过长，同时被验者身体中的生理性运动会产生影像的模糊和对比度的失真，无法应用于运动器官以及神经系统等的研究。有效提高成像速度已成为磁共振领域一项非常重要的研究，在缩短检查时间的同时提高成像质量和改善病人被检查时的舒适性。事实上，成像速度一直以来就是磁共振进一步广泛应用的瓶颈问题。为了提高成像速度，除了改善磁共振设备硬件性能之外，目前最常用的方法就是部分k空间数据重建。其优势在于不需要改变现有硬件设备结构和扫描方式，是一种低成本的快速成像方法。整体变分(Total Variation, TV)方法最初被提出用于图像去噪，目前已发展成为磁共振重建成像领域一种比较常用的方法。它通过对图像模型变分最小的处理，目的就是在移除图像中噪声及不需要的小尺度细节的同时，保护那些突变的不连续区域。然而TV法是基于图像是由连续片段组成的假定，在一些实际情况中效果往往不太理想。针对这种情况，近几年发展推广出了一种广义全局变量(Total GeneralizedVariation, TGV)方法，用于解决这类问题。本文对这两种方法做了大量实验，其中TV方法同时应用了牛顿法和共轭梯度两种不同的求解方法。在对组成结构比较复杂的图像（例如T2大脑图）进行重建时，TGV方法要优于TV方法，尤其在高度欠采样的情况下；而对于图像结构相对简单的图像（比如Shepp-Logan模型），TGV方法没有明显的优势，甚至效果更不理想，，而且所需时间更长。
[Abstract]:Magnetic Resonance imaging (MRI) is a new, non-invasive imaging method.In clinical application, magnetic resonance imaging has good contrast of soft tissue, many imaging parameters, high resolution of three-dimensional space, and can be used in any direction.The inherent characteristics of non-ionizing radiation and the advantages of being easily accepted by patients [1] [2] have become more and more popular and have been widely used.The application of modern magnetic resonance imaging has quite high requirements for spatial resolution spatial resolution, instantaneous resolution and temporal resolution, signal-to-noise ratio (SNR) and imaging speed [3] [4].However, the conventional magnetic resonance imaging (MRI) is too long, and the physiologic motion in the body of the subject produces blurred image and distortion of contrast, so it can not be applied to the study of motor organs and nervous system and so on.Improving imaging speed effectively has become a very important research in the field of magnetic resonance. It can shorten the inspection time and improve the quality of imaging and the comfort of patients when they are examined.In fact, imaging speed has always been a bottleneck in the further application of MRI.In order to improve the imaging speed, in addition to improving the hardware performance of magnetic resonance devices, the most commonly used method is partial k spatial data reconstruction.Its advantage is that it does not need to change the existing hardware structure and scanning mode, it is a low-cost fast imaging method.The global variational Total variation (TV) method was originally proposed for image denoising and has been developed into a common method in the field of magnetic resonance imaging (MRR).By processing the minimum variation of the image model, the aim is to protect the discontinuous region of the mutation while removing the noise and the unnecessary small-scale details of the image.However, the TV method is based on the assumption that the image is composed of continuous segments.In view of this situation, a generalized global variable Total Generalized variation (TGV) method has been developed in recent years to solve this kind of problems.In this paper, a large number of experiments have been done on these two methods, among which Newton method and conjugate gradient method are used in TV method.The TGV method is superior to the TV method in the reconstruction of complex images (such as T2 brain map), especially in the case of highly under-sampling.However, for images with relatively simple image structure (such as Shepp-Logan model / TGV method), there is no obvious advantage, even less effective, and the time required is longer.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R318.0

【参考文献】