Gradient-Based Low Rank Method for Highly Undersampled Magnetic Resonance Imaging Reconstruction

doi:10.1007/s12204-018-1927-8

Abstract

Abstract: Recently, exploiting low rank property of the data accomplished by the non-convex optimization has shown great potential to decrease measurements for compressed sensing. In this paper, the low rank regularization is adopted to gradient similarity minimization, and applied for highly undersampled magnetic resonance imaging (MRI) reconstruction, termed gradient-based low rank MRI reconstruction (GLRMRI). In the proposed method, by incorporating the spatially adaptive iterative singular-value thresholding (SAIST) to optimize our gradient scheme, the deterministic annealing iterates the procedure e±ciently and superior reconstruction performance is achieved. Extensive experimental results have consistently demonstrated that GLRMRI recovers both real- valued MR images and complex-valued MR data accurately, especially in the edge preserving perspective, and outperforms the current state-of-the-art approaches in terms of higher peak signal to noise ratio (PSNR) and lower high-frequency error norm (HFEN) values.

Key words: magnetic resonance imaging (MRI)| low rank| image gradients| sparse representation| deterministic annealing

摘要： Recently, exploiting low rank property of the data accomplished by the non-convex optimization has shown great potential to decrease measurements for compressed sensing. In this paper, the low rank regularization is adopted to gradient similarity minimization, and applied for highly undersampled magnetic resonance imaging (MRI) reconstruction, termed gradient-based low rank MRI reconstruction (GLRMRI). In the proposed method, by incorporating the spatially adaptive iterative singular-value thresholding (SAIST) to optimize our gradient scheme, the deterministic annealing iterates the procedure e±ciently and superior reconstruction performance is achieved. Extensive experimental results have consistently demonstrated that GLRMRI recovers both real- valued MR images and complex-valued MR data accurately, especially in the edge preserving perspective, and outperforms the current state-of-the-art approaches in terms of higher peak signal to noise ratio (PSNR) and lower high-frequency error norm (HFEN) values.

关键词: magnetic resonance imaging (MRI)| low rank| image gradients| sparse representation| deterministic annealing

CLC Number:

TN 911.73

XU Xiaoling (徐晓玲), LIU Yiling (刘沂玲), LIU Qiegen (刘且根),LU Hongyang (卢红阳), ZHANG Minghui (张明辉). Gradient-Based Low Rank Method for Highly Undersampled Magnetic Resonance Imaging Reconstruction[J]. Journal of Shanghai Jiao Tong University (Science), 2018, 23(3): 384-.

References 21

[1]	LUSTIG M, DONOHO D, PAULY J M. Sparse MRI:The application of compressed sensing for rapid MRimaging [J]. Magnetic Resonance in Medicine, 2007,58: 1182-1195.
[2]	HALDAR J P, HERNANDO D, LIANG Z P.Compressed-sensing MRI with random encoding [J].IEEE Transactions on Medical Imaging, 2011, 30(4):893-903.
[3]	SMITH D S, WELCH E B. Non-sparse phan-tom for compressed sensing MRI reconstruction[C]//International Society for Magnetic Resonance inMedicine 19th Scientiˉc Meeting. [s.l.]: ISMEM, 2011:2845.
[4]	BILGIN A, KIM Y, LIU F, et al. Dictionary design forcompressed sensing MRI [C]//Proceedings of the 18thScientiˉc Meeting of ISMRM. Stockholm: ISMRM,2010: 4887.
[5]	RAVISHANKAR S, BRESLER Y. MR image recon-struction from highly undersampled k-space data bydictionary learning [J]. IEEE Transactions on MedicalImaging, 2011, 30: 1028-1041.
[6]	LIU Q, WANG S, YANG K, et al. Highly under-sampled magnetic resonance image reconstruction us-ing two-level Bregman method with dictionary updat-ing [J]. IEEE Transactions on Medical Imaging, 2013,32(7): 1290-1301.
[7]	MAIRAL J, BACH F, PONCE J, et al. Non-localsparse models for image restoration [C]//2009 IEEE12th International Conference on Computer Vision.[s.l.]: IEEE, 2009: 2272-2279.
[8]	BADRI H, YAHIA H. A non-local lowrank approach toenforce integrability [J]. IEEE Transactions on ImageProcessing, 2016, 25(8): 3562-3571.
[9]	REN W, CAO X, PAN J, et al. Image deblurring viaenhanced low-rank prior [J]. IEEE Transactions onImage Processing, 2016, 25(7): 3426-3437.
[10]	DONG W, SHI G, LI X. Nonlocal image restorationwith bilateral variance estimation: A low-rank ap-proach [J]. IEEE Transactions on Image Processing,2013, 22(2): 700-711.
[11]	PATEL V M, MALEH R, GILBERT A, et al.Gradient-based image recovery methods from incom-plete fourier measurements [J]. IEEE Transactions onImage Processing, 2012, 21(1): 94-105.
[12]	YANG J, ZHANG Y, YIN W. A fast alternating direc-tion method for TVL1-L2 signal reconstruction frompartial Fourier data [J]. IEEE Journal of Selected Top-ics in Signal Processing, 2010, 4(2): 288{297.
[13]	LIU Q, WANG S, YING L, et al. Adaptive dictionarylearning in sparse gradient domain for image recov-ery [J]. IEEE Transactions on Image Processing, 2013,22(12): 4652-4663.
[14]	HU Z, LIU Q, PENG X, et al. Image reconstructionfrom few-view CT data by gradient-domain dictionarylearning [J]. Journal of X-Ray Science and Technology,2016, 24(4): 627-638.
[15]	LUO X, SUO Z, LIU Q, et al. E±cient InSAR phasenoise ˉltering based on adaptive dictionary learning ingradient vector domain [J]. Journal of Xidian Univer-sity, 2016, 43(1): 18-23 (in Chinese).
[16]	CAI J F, CANDES E J, SHEN Z. A singular valuethresholding algorithm for matrix completion [J].SIAM Journal on Optimization, 2010, 20(4): 1956-1982.
[17]	ARIAS P, CASEL V, SAPIRO G. A variational frame-work for non-local image inpainting [C]//InternationalWorkshop on Energy Minimization Methods inComputer Vision and Pattern Recognition. Berlin:Springer, 2009: 345-358.
[18]	ARIAS P, FACCIOLO G, CASELLES V, et al.A variational framework for exemplar-based imageinpainting [J]. International Journal of ComputerVision, 2011, 93(3): 319-347.
[19]	P'EREZ P, GANGNET M, BLAKE A. Poisson imageediting [J]. InACM Transactions on Graphics (TOG),2003, 22(3): 313-318.
[20]	HORE A, ZIOU D. Image quality metrics: PSNRvs. SSIM [C]//International Conference on PatternRecognition. [s.l.]: IEEE, 2010: 2366-2369.
[21]	SHAH M A. Distributed continuous media streaming-Using redundant hierarchy (RED-Hi) servers [D]. Fam-agusta: Eastern Mediterranean University (EMU),2014.