基于虚拟线圈和卷积神经网络的多层同时激发图像重建  

Reconstruction of Simultaneous Multi-Slice MRI Data by Combining Virtual Conjugate Coil Technology and Convolutional Neural Network

作　　者：王婉婷 苏适[1] 贾森 梁栋[1,2,3] 王海峰 WANG Wan-ting;SU Shi;JIA Sen;LIANG Dong;WANG Hai-feng(Paul C.Lauterbur Research Center for Biomedical Imaging,Shenzhen Institutes of Advanced Technology,Chinese Academy of Sciences,Shenzhen 518055,China;University of Chinese Academy of Sciences,Beijing 100049,China;Research Center for Medical AI,Shenzhen Institutes of Advanced Technology,Chinese Academy of Sciences,Shenzhen 518055,China)

机构地区：[1]保罗C.劳特伯生物医学成像研究中心(中国科学院深圳先进技术研究院),广东深圳518055 [2]中国科学院大学,北京100049 [3]医学人工智能研究中心(中国科学院深圳先进技术研究院),广东深圳518055

出　　处：《波谱学杂志》2020年第4期407-421,共15页Chinese Journal of Magnetic Resonance

基　　金：深圳市孔雀团队资助项目(KQTD20180413181834876);中国科学院B类先导专项资助项目(XDB25000000);中国科学院影像技术和装备工程实验室资助项目(KFJ-PTXM-012);广东省珠江人才计划资助项目(2019QN01Y986);广东省自然科学基金资助项目(2018A0303130132);国家自然科学基金资助项目(61771463,81830056,U1805261,81971611,61871373,81729003,81901736)。

摘　　要：本文提出一种基于虚拟共轭线圈(Virtual Coil Concept,VCC)技术和k空间插值鲁棒人工神经网络(Robust Artificial-neural-networks for k-space Interpolation,RAKI)的图像重建方法,用于磁共振多层同时激发成像(Simultaneous Multi-Slice imaging,SMS),该方法能够有效提升重建图像的质量,被命名为VIRGINIA(VIRtual conjuGate colls Neural-networks InterpolAtion).为了得到更高质量的SMS图像,本文提出的VIRGINIA方法利用磁共振线圈数据的复数共轭对称性质扩展了SMS所获取的多通道数据,并将扩展后的数据用于RAKI网络的训练,利用训练后的网络实现高质量的SMS图像重建.本文将VIRGINIA方法和其他SMS图像重建方法(RAKI和Slice-GRAPPA方法)进行了对比,并采用结构相似指数(Structural Similarity Index,SSIM)、峰值信噪比(Peak Signal-to-Noise Ratio,PSNR)和均方根误差(Root Mean Square Error,RMSE)对不同方法的重建图像进行了量化对比分析.结果显示,在相同的SMS加速倍数下,使用VIRGINIA方法进行重建的图像质量均好于RAKI方法,且远好于传统Slice-GRAPPA方法.This paper proposes an image reconstruction method for simultaneous multi-slice imaging(SMS)by combining the virtual conjugate coil(VCC)technology and robust artificial-neural-networks for k-space interpolation(RAKI).This method can effectively improve the reconstruction quality,and is named VIRGINIA(VIRtual conjuGate coIls Neural-networks InterpolAtion).VIRGINIA utilizes the complex conjugate symmetry property of the virtual coil concept to generate virtual coil data for training,and obtains better image quality by applying the trained network to the original aliased SMS data.With experimental data,the VIRGINIA method was compared to other reconstruction methods(i.e.,RAKI only and slice-GRAPPA)in terms of quantitative indices such as structural similarity index(SSIM),peak signal-to-noise ratio(PSNR)and root mean square error(RMSE).The results demonstrated that,under some certain slice-acceleration factors,VIRGINIA produced better reconstruction quality than those obtainable by Slice-GRAPPA and RAKI.

关 键 词：磁共振图像重建 多层同时成像 k空间插值鲁棒人工神经网络(RAKI) 虚拟线圈 卷积神经网络(CNN) 

分 类 号：O482.53[理学—固体物理]