Huajun She^{1}, Bian Li^{1}, Joshua S. Greer^{1,2}, Jochen Keupp^{3}, Ananth Madhuranthakam^{1,4}, Ivan E. Dimitrov^{1,5}, Robert Lenkinski^{1,4}, and Elena Vinogradov^{1,4}

This work investigates accelerating CEST imaging using parallel blind compressed sensing (BCS). BCS method assumes a few functions are enough to represent the dynamic behavior. In CEST imaging, the Z-spectrum performs similar in the same compartment, which is suitable for BCS reconstruction. The traditional BCS method does not consider the coil sensitivity, which is complementary sparse information with spatial-temporal dictionary. The proposed method addresses the coil sensitivity information and the sparsity prior information in CEST and further improves the BCS method, demonstrating a better estimation of the CEST effect for both phantom and in vivo brain data.

[1] van Zijl P, et al. MRM 2011;65:927–948.

[2] Pruessmann KP, et al. MRM 1999;42:952–962.

[3] Heo HY, et al. MRM 2016;10.1002/mrm.26141.

[4] Candes EJ, et al. IEEE TIT 2006;52:489–509.

[5] Lustig M, et al. MRM 2007;58:1182–1195.

[6] Lingala SG, et al. IEEE TMI 2013;32:1132–1145.

[7] She H, et al. ISMRM 2016;2904.

[8] Heo HY, et al. ISMRM 2016;0301.

[9] Uecker M, et al. MRM 2014;71:990–1001.

[10] Kim M, et al. MRM 2009;61:1441–1450.

Fig. 1 compares MTR_{asym} maps reconstruction
of (a) fully sampled, (b) BCS, (c) BCS Error=|MTR_{asym}Full-MTR_{asym}BCS|x5,
(d) PBCS and (e) PBCS Error=|MTR_{asym}Full-MTR_{asym}PBCS|x5
in each region of interest (ROI) for different pH values. The
reconstruction error (magnified 5 times) demonstrates that PBCS results in
better reconstruction, as compared with the BCS method. The Z-spectra (f-h) for
pH = 7.0, 7.5, 8.0 display the mean value in the ROIs. The BCS curve is fairly
accurate for most of the range but underperforms PBCS at the fast changing range
near the zero-frequency offset.

Fig.
2 compares MTR_{asym} maps reconstruction of (a) fully sampled, (b) BCS, (c) BCS Error = |MTR_{asym}Full - MTR_{asym}BCS|, (d) PBCS and (e) PBCS Error = |MTR_{asym}Full - MTR_{asym}PBCS| for in vivo brain of volunteer #1 at reduction factor R = 4. The reconstruction error demonstrates that PBCS gives a better reconstruction quality compared with the BCS method.

Fig. 3 compares MTR_{asym} maps reconstruction of (a) fully sampled, (b) BCS, (c) BCS Error = |MTR_{asym}Full - MTR_{asym}BCS|, (d) PBCS and (e) PBCS Error = |MTR_{asym}Full - MTR_{asym}PBCS| for in vivo brain of a highly cooperative (less motion) volunteer #2 at high reduction factor R = 8. The reconstruction error demonstrates that PBCS gives a better reconstruction quality compared with the BCS method.

Fig.
4 compares correlation plots of the MTR_{asym} estimation of (a) BCS and (b) PBCS for in vivo brain of another volunteer #3 at reduction factor R = 4 for different saturation power 0.7 μT (red), 1.2 μT (green), 1.6 μT (blue). From the correlation plots and r^{2} values it is clear that
the PBCS leads to higher correlation estimation with the fully sampled case than
the BCS method.