计算机应用 ›› 2020, Vol. 40 ›› Issue (10): 3054-3059.DOI: 10.11772/j.issn.1001-9081.2020030285

• 虚拟现实与多媒体计算 • 上一篇    下一篇

基于深度先验及非局部相似性的压缩感知核磁共振成像

宗春梅1, 张月琴2, 曹建芳1, 赵青杉1   

  1. 1. 忻州师范学院 计算机系, 山西 忻州 034000;
    2. 太原理工大学 计算机科学与技术学院, 太原 030024
  • 收稿日期:2020-03-16 修回日期:2020-05-14 出版日期:2020-10-10 发布日期:2020-06-01
  • 通讯作者: 宗春梅
  • 作者简介:宗春梅(1977-),女,山西忻州人,讲师,硕士,CCF会员,主要研究方向:深度学习、图像处理;张月琴(1963-),女,山西太原人,教授,硕士,主要研究方向:数据挖掘、深度学习;曹建芳(1976-),女,山西忻州人,教授,博士,主要研究方向:图像处理、机器学习;赵青杉(1972-),男,山西忻州人,教授,硕士,主要研究方向:大数据、深度学习。
  • 基金资助:
    国家自然科学基金资助项目(61876124);忻州师范学院“1331工程”科研项目(2019ky02)。

Compressed sensing magnetic resonance imaging based on deep priors and non-local similarity

ZONG Chunmei1, ZHANG Yueqin2, CAO Jianfang1, ZHAO Qingshan1   

  1. 1. Department of Computer Science, Xinzhou Teachers University, Xinzhou Shanxi 034000, China;
    2. College of Computer Science and Technology, Taiyuan University of Technology, Taiyuan Shanxi 030024, China
  • Received:2020-03-16 Revised:2020-05-14 Online:2020-10-10 Published:2020-06-01
  • Supported by:
    This work is partially supported by National Natural Science Foundation of China (61876124), the Scientific Research Project of "1331 Program" of Xinzhou Teachers University (2019ky02).

摘要: 针对现有压缩感知核磁共振成像(CSMRI)算法在低采样率下重构质量低的问题,提出一种融合深度先验及非局部相似性的成像方法。首先,利用深度去噪器和块匹配三维滤波(BM3D)去噪器构建能够融合多种图像先验知识的稀疏表示模型;其次,将该模型作为正则化项,利用高度欠采样的k空间数据构建压缩感知核磁共振成像优化模型;最后,利用交替优化方法求解构建的优化问题。所提出的算法不仅能够通过深度去噪器利用深度先验,还能够通过BM3D去噪器利用图像的非局部相似性来进行图像重建。实验结果表明,与基于BM3D的重建算法相比,该算法在采样率为0.02、0.06、0.09及0.13情况下重构的平均峰值信噪比高出约1 dB;此外,从视觉角度,与现有的基于小波树稀疏性的核磁共振成像算法WaTMRI、基于字典学习的核磁共振成像算法DLMRI、基于字典更新及块匹配和三维滤波的核磁共振成像算法DUMRI-BM3D等相比,所提算法重构的图像包含大量纹理信息,与原始图像最接近。

关键词: 压缩感知, 核磁共振成像, 深度先验, 非局部相似性, 稀疏表示

Abstract: Aiming at the problem of low reconstruction quality of the existing Compressed Sensing Magnetic Resonance Imaging (CSMRI) algorithms at low sampling rates, an imaging method combining deep priors and non-local similarity was proposed. Firstly, a deep denoiser and Block Matching and 3D filtering (BM3D) denoiser were used to construct a sparse representation model that can fuse multiple priori knowledge of images. Secondly, the undersampled k-space data was used to construct a compressed sensing magnetic resonance imaging optimization model. Finally, an alternative optimization method was used to solve the constructed optimization problem. The proposed algorithm can not only use the deep priors through the deep denoiser, but also use the non-local similarity of the image through the BM3D denoiser to reconstruct the image. Compared with the reconstruction algorithms based on BM3D, experimental results show that the proposed algorithm has the average peak signal-to-noise ratio of reconstruction increased about 1 dB at the sampling rates of 0.02, 0.06, 0.09 and 0.13. Compared with the existing MRI algorithm WaTMRI (Magnetic Resonance Imaging with Wavelet Tree sparsity),DLMRI (Dictionary Learning for Magnetic Resonance Imaging), DUMRI-BM3D (Magnetic Resonance Imaging based on Dictionary Updating and Block Matching and 3D filtering), etc, the images reconstructed by the proposed algorithm contain a lot of texture information, which are the closest to the original images.

Key words: Compressed Sensing (CS), Magnetic Resonance Imaging (MRI), deep prior, non-local similarity, sparse representation

中图分类号: