Evaluation of Primal-Dual Splitting Algorithm for MRI Reconstruction Using Spatio-Temporal Structure Tensor and L1-2 Norm

Authors

Mia Rizkinia, Department of Electrical Engineering, Faculty of Engineering, Universitas Indonesia, Depok 16424, IndonesiaFollow
Masahiro Okuda, Faculty of Environmental Engineering, The University of Kitakyushu, Kitakyushu 808-0135, Japan

Abstract

Magnetic resonance imaging (MRI) is an essential medical imaging technique which is widely used for medical research and diagnosis. Dynamic MRI provides the observed object visualization through time and results in a spatiotemporal signal. The image sequences often contain redundant information in both spatial and temporal domains. To utilize this characteristic, we propose a spatio-temporal reconstruction approach based on compressive sensing theory. We apply spatio-temporal structure tensor using nuclear norm, in addition to the wavelet sparsity regularization. The spatio-temporal structure tensor is a matrix that consists of gradient components of the MRI data w.r.t the spatial and temporal domains. For the wavelet sparsity, we use L1 – L2 instead of L1 norm. We propose the algorithm using primaldual splitting (PDS) approach to solve the convex optimization problem. In the experiment, we investigate the potential benefit of adding the two regularizations to the compressive sensing problem. The algorithm is compared with PDSbased algorithm using conventional regularizations, i.e., wavelet sparsity and total variation. Our proposed algorithm performs superior results in terms of reconstruction accuracy and visual quality.

Bahasa Abstract

Rekonstruksi Compressive Sensing MRI menggunakan Spatial-Temporal Structure Tensor melalui Metode Primal-Dual Splitting. Magnetic resonance imaging (MRI) merupakan sebuah tehnik penting dalam pencitraan medis yang digunakan luas untuk penelitian dan diagnosa medis. Dynamic MRI menyediakan visualisasi objek yang diamati terhadap waktu dan menghasilkan sinyal spasial-temporal. Dengan memanfaatkan karakteristik ini, kami mengajukan sebuah pendekatan rekonstruksi spasial-temporal berdasarkan pada teori compressive sensing. Kami menerapkan spatio-temporal structure tensor menggunakan nuclear norm, selain regularisasi wavelet sparsity. Spatio-temporal structure tensor adalah sebuah matriks yang terdiri atas komponen gradien dari data MRI berkenaan dengan domain spasial dan temporal. Untuk wavelet sparsity, kami menggunakan L1 – L2 daripada L1 norm. Kami mengajukan algoritma menggunakan pendekatan primal-dual splitting (PDS) untuk menyelesaikan masalah optimisasi convex. Dalam eksperimen, kami menginvestigasi potensi kelebihan penambahan kedua regularisasi ke masalah compressive sensing. Algoritma ini dibandingkan dengan algoritma PDS yang menggunakan regularisasi konvensional yaitu, wavelet sparsity dan total variation. Algoritma yang kami tawarkan menunjukkan hasil yang superior dalam hal akurasi rekonstruksi dan kualitas visual citra.

References

1. G. Oliveri, M. Salucci, N. Anselmi, A. Massa, IEEE Antennas Propagation Mag. 59/5 (2017) 34.
2. S. Qaisar, R.M. Bilal, W. Iqbal, M. Naureen, S. Lee, J. Comm. Netw. 15/5 (2014) 443.
3. E.J. Candes, M.B. Wakin, IEEE Signal Proc. Mag. 25/2 (2008) 21.
4. M. Lustig, D.L. Donoho, J.M. Santos, J.M. Pauly, IEEE Signal Proc. Mag. 25/2 (2008) 72.
5. C. Chen, J. Huang, Magn. Reson. Imag. 32/10 (2014) 1377.
6. X. Qu, W. Zhang, D. Guo, C. Cai, S. Chai, Z. Chen, Inver. Problem Sci. Eng, 18/6 (2010) 737.
7. J. Huang, S. Zhang, D. Metaxas, Proc. the 13th Int. Conf. Med. Imag. Comp. Computer-Assisted Int.: Part I (MICCAI'10), 2010, p.135.
8. C. Chen, Y. Li, J. Huang, IEEE Trans. Signal Proc. 62/11 (2014) 2803.
9. R. Otazo, E. Candes, D.K. Sodickson, Magn. Reson. Med. 73/3 (2015) 1125.
10. C. Chen, Y. Li, L. Axel, J. Huang, Magn. Reson. Imag. 34/4 (2016) 473.
11. C. Ulas, P.A. Gómez, J.I. Sperl, C. Preibisch, B.H. Menze, Proc. IEEE 13th Int. Sym. on Biomed. Imag. (ISBI), 2016, p.306.
12. P. Montesinos, J.F.P.J. Abascal, L. Cussó, J.J. Vaquero, M. Desco, Magn. Reson. Med. 72/2 (2013) 369.
13. Y. Lou, M. Yan, J. Sci. Comput. 74/2 (2018) 767.
14. L. Condat, J. Optim. Theory Appl. 158/2 (2013) 460.
15. P.L. Combettes, J.C. Pesquet, Fixed-Point Algorithm. Inver. Prob. Sci. Eng. Springer, 2011, p.185.
16. S. Vaishali, K.K. Rao, G.V.S. Rao, Int. Conf. Sig. Proc. Comm. Eng. Sys. (SPACES), 2015, p. 363.
17. Y. Lou, S. Osher, J. Xin, Adv. Int. Syst. Comput. 359 (2015) 169.
18. Y. Lou, T. Zeng, Stanley J. Osher, Jack Xin, SIAM J. Imag. Sci. 8/3 (2015) 1798.
19. W. Wang, J. Wang, Z. Zhang, IEEE Signal Proc. Lett. 24/3 (2017) 304.
20. R. Kurihara, S. Ono, K. Shirai, M. Okuda, 25th Eur. Signal Proc. Conf. (EUSIPCO), 2017, pp.488.
21. A. Chambolle. T. Pock, J. Math. Imag. Vis. 40/1 (2010) 120.
22. P.L Combettes, J.C. Pesquet, Set-Valued Variation Anal. 20/2 (2012) 307.
23. C. Chen, Y. Li, L. Axel, J. Huang, Magn. Reson. Imag. 34 (2016) 473.
24. R. Otazo, D. Kim, L. Axel, D.K. Sodickson, Magn. Reson. Med. 64/3 (2010) 767.
25. R. Otazo, L. Feng, H. Chandarana, T. Block, L. Axel, D.K. Sodickson, Proc. Int. Symp. Biomed. Imag., 2012, p.980.

Recommended Citation

Rizkinia, Mia and Okuda, Masahiro (2019) "Evaluation of Primal-Dual Splitting Algorithm for MRI Reconstruction Using Spatio-Temporal Structure Tensor and L1-2 Norm," Makara Journal of Technology: Vol. 23: Iss. 3, Article 4.
DOI: 10.7454/mst.v23i3.3892
Available at: https://scholarhub.ui.ac.id/mjt/vol23/iss3/4