节点文献

基于压缩感知的磁共振图像重建方法研究

Research on Magnetic Resonance Image Reconstruction Method Based on Compressed Sensing

【作者】 刘静

【导师】 申艳;

【作者基本信息】 北京交通大学 , 电子科学与技术, 2019, 硕士

【摘要】 磁共振成像(Magnetic Resonance Imaging,MRI)技术是重要的医学辅助手段之一。该技术具有无电离辐射,对软组织灵敏度高,成像方位多样化等优点。但是MRI的数据扫描时间较长,数据的采集时间和重建图像质量之间的关系难以平衡。增加扫描时间意味着可以采集更多的数据,获得更高质量的重建图像;但是扫描时间的增加可能会引入运动伪影,引起患者的不适。而降低扫描时间将减少采集的数据量,进而导致重建图像质量的下降。因此,利用少量的数据重建清晰的MRI图像具有重要的研究意义。本文以压缩感知为理论基础,研究了从稀疏采样中重建出完整MRI图像的方法,分别建立了静态MRI和动态MRI的稀疏重建模型,具体内容包括:针对静态磁共振成像中扫描时间较长的问题,提出一种基于Shearlet变换和非局部结构相似性的静态MRI重建算法。该算法通过Shearlet变换对图像进行多尺度多方向分解,充分地提取图像特征,对图像进行更稀疏的表示;并结合图像的非局部结构相似性,在算法迭代过程中通过对图像块加权求和来保留图像的细节信息。使用公开的静态MRI图像进行实验,实验结果表明,该算法可以用10%的数据重建静态MRI图像,并且能够较好地保持图像的细节信息,与其他方法相比重建图像的PSNR和SSIM分别提高3dB和0.2。在动态磁共振成像中,提出一种基于时间和空间稀疏的动态MRI稀疏重建算法,以缩短单帧图像的扫描时间,从而获取器官在更多时刻的运动状态。该算法利用动态MRI图像在时间上的冗余性,通过动态全变差对图像序列进行稀疏表示,同时结合图像在双树复小波变换域的空间稀疏性,更好地表示图像的特征信息,有效地保留细节信息。使用公开的动态MRI图像进行实验,实验结果表明,该算法能够用14%的数据重建动态MRI图像,并且在保证与其他方法同等重建质量的情况下,重建时间降低0.4秒。

【Abstract】 Magnetic resonance imaging(MRI)is one of the important medical aids for its nonionizing radiation,excellent depiction of soft tissues and the arbitrary imaging.However,the scan time of MRI is relatively long,and it is hard to balance the relationship between the quality of reconstructed images and the scan time.Increasing the scan time means more data for reconstruction and higher quality images,but it may cause a number of problems including increasing the susceptibility to physiological motion artifacts and adding discomfort of the patients.Nevertheless,reducing the amount of data will lead to the decline of image quality.Therefore,the reconstruction of clear MRI images with less data is of great clinical significance.Based on compressed sensing,the sparse reconstruction algorithms of static MRI and dynamic MRI are studied separately,details are as follows:In order to solve the problem of long scan time,a static MRI reconstruction algorithm based on Shearlet and nonlocal structural similarity is proposed.In the algorithm,Shearlet transform is used to decompose images in multiple scales and directions,with which the features of images can be extracted more fully and the images can be represented more sparsely.In addition,combining the nonlocal structural similarity,the blocks of the image are weighted and summed in the iteration to ensure the detail information.Experimental results on the public static MRI dataset show that the proposed algorithm can reconstruct static MRI images with 10%data as well as ensure details,and the PSNR and SSIM of the reconstructed image are improved by 3dB and 0.2 respectively.In dynamic magnetic resonance imaging,a dynamic MRI sparse reconstruction algorithm based on temporal and spatial sparsity is proposed to shorten the scan time of a single frame image so as to obtain the motion states of organs.In the algorithm,the dynamic total variation is applied to represent the image sequences sparsely,which utilizes the temporal redundancy of dynamic MRI images.In addition,the spatial sparsity of images in the double tree complex wavelet transform domain is applied to better represent the features of images.Experimental results on the public dynamic MRI dataset show that the proposed algorithm can reconstruct images with 14%data and reduce the time by 0.4 seconds with the same reconstruction quality as other methods.

  • 【分类号】TP391.41;R445.2
  • 【被引频次】1
  • 【下载频次】245
  • 攻读期成果
节点文献中: 

本文链接的文献网络图示:

本文的引文网络