Comparison of 3D CEST acquisition schemes: steady state versus pseudo-steady state

Vitaliy Khlebnikov¹, Nicolas Geades², Dennis WJ Klomp¹, Hans Hoogduin¹, Penny Gowland², and Olivier Mougin²

¹Radiology, University Medical Center Utrecht, Utrecht, Netherlands, ²Sir Peter Mansfield Imaging Centre, University of Nottingham, Nottingham, Nottinghamshire, United Kingdom, Nottingham, United Kingdom

Synopsis

Chemical Exchange Saturation Transfer (CEST) has gained much popularity due to its unmatched sensitivity to dilute labile protons when compared to other MRI techniques. Of particular interest are two CEST effects: Amide Proton Transfer (APT, CEST of amides) and Nuclear Overhauser Enhancement (NOE). Fast-paced developments for CEST resulted in the design of multiple CEST imaging sequences. This raises the obvious question as to which sequence to use and in what particular applications. Two pulsed volumetric CEST acquisition schemes are currently available in the literature. The first is based on the standard Magnetization Transfer imaging technique: a steady-state (SS) acquisition with alternating brief saturation and image acquisition. The second is based on the preparation of the magnetization before a long readout, where the prolonged saturation reaches a pseudo-steady state (PS) before the image acquisition. In this report, these two CEST acquisition schemes, optimized for maximum sensitivity to APT and NOE effects through Bloch-McConnell simulations, were systematically compared for the same spatial resolution, brain coverage and scan time.

Target audience

Those interested in the development of new Chemical Exchange Saturation Transfer (CEST) acquisition strategies.

Introduction

Two 3D CEST acquisition schemes are currently available in literature. The first is based on a steady-state (SS) acquisition with alternating saturation and readout [1]. The second uses a prolonged saturation to reach a pseudo-steady (PS) state followed by a readout [2,3]. The aim of this report was to compare both SS and PS, optimized through Bloch-McConnell simulations, for the same spatial resolution, brain coverage and scan time. The comparison was done in terms of sequence sensitivity to APT and NOE effects, sensitivity of those effects to B1-inhomogeneity, and SNR.

Methods

All experiments were performed on a 7T Philips MR system. Five healthy, consented subjects were scanned using two CEST protocols at three nominal B1 amplitudes: 1µT, 2µT and 3µT and 37 frequency offsets. SS protocol was as follows [1]: saturation prepulse (a single RF-spoiled 25ms sinc-gauss pulse followed by a 50mT/m spoiler of 10ms) interleaved with a readout (sagittal acquisition, segmented EPI, EPI factor 13 with a binomial pulse for water only excitation, TR/TE/FA=58ms/6ms/10°, k-space center-weighted acquisition); scan time 10min27s. PS protocol was as follows [2,3]: saturation prepulse (train of 40 RF-spoiled 25ms sinc-gauss pulses interleaved with a sinusoidal-modulated GR-spoiler, duty cycle 50%); readout (axial acquisition, three-shot TFE, TFE factor of 550, inter-shot recovery 2s, TR/TE/FA=2.3ms/1.05ms/10°, acquisition starts in the center of k-space); 2dummy scans, scan time 10min18s. The parameters for both SS and PS techniques were optimized to match the total acquisition scan time as close as possible, with a FOV of 150x224x208mm3, voxel size of 2mm isotropic with SENSE factor 2. For SNR estimations, a noise image was obtained by switching off all of RF pulses and gradients. All simulations were performed for four-pools (APT, NOE, water and magnetization transfer-MT).

Data processing

Z-spectra data were motion corrected (FSL). A T1-weighted anatomical scan was used to create white matter (WM) and grey matter (GM) masks (FSL), used to average APT and NOE effects. B0 correction was done using WASSR [4]. In experiments, APT and NOE were quantified as the area obtained using three-point method [5] (3 to 4 ppm and -5 to -2 ppm, respectively). The maps of traditional asymmetry (MTRasym) [6] were calculated in the region of 3 to 4 ppm. In simulations, APT (3.5ppm) and NOE (-3.5ppm) were quantified by subtracting a three-pool spectrum (APT or NOE pool parameters set to 0) from a four-pool spectrum.

Results and Discussion

Based on water-T1(T1w) values expected in the healthy human brain, the optimum number of saturation pulses (Fig.1) was determined to be 152 (to reach the center of k-space) and 40 (due to proximity to maximum) for SS and PS, respectively.The SS scheme has maximum sensitivity to the effect size of APT (Fig. 2 A) at a B1 of 1µT and 43% duty cycle (hardware-limited) and NOE (Fig. 2 B) at a B1 of 2µT and 43% duty cycle (hardware-limited); whereas PS scheme has maximum sensitivity to the effect size of APT (Fig. 2 C) at a B1 of 1µT and 50% duty cycle and NOE (Fig. 2 D) at a B1 of 2µT and 50% duty cycle.Both SS and PS show maximum APT and NOE effects at 1µT and 2µT, respectively (Fig. 3). For the optimum saturation parameters, the contrast change was found to be 30% and 6% for a B1 dropout of 40% for APT and NOE, respectively. NOE effect is relatively insensitive to B1-inhomogeneity effects compared to APT. No difference for SS and PS in their sensitivity to the effects of B1-inhomogeneity was found.In experiments (Fig. 5), no statistically significant differences (p>0.05) were found between SS and PS for APT and NOE effects in WM. In GM, however, SS is more sensitive to APT (p<0.05) and PS is more sensitive to NOE (p<0.05) at both 2µT and 3µT. For the parameters used both SS and PS produced images with poor spatial resolution (Fig. 5) due to a long TFE readout and EPI distortions, respectively, although the PS sequence would not normally be used to acquire data at such low resolution [3]

Conclusions

For the same spatial resolution, brain coverage and scan time, SS is more sensitive to APT effect, whereas PS is more sensitive to NOE effect, and has on average 130% more SNR (due to an inter-shot T1recovery) that can be traded for a higher spatial resolution. Both SS and PS have similar sensitivity to B1-inhomogeneity effects and similar SAR.

Acknowledgements

This work was supported by the European Commission (FP7-PEOPLE-2012-ITN-316716).

References

[1] Jones CK et al. MRM 2012. [2] Zhu H et al. MRM 2010. [3] Mougin O et al. NMR Biomed 2013. [4] Kim M et al. MRM 2009. [5] Jin T et al. MRM 2013. [6] Guivel-Scharen V et al. JMagnReson 1998.

Figures

Fig. 1. Simulated approach to steady state for SS (A – APT, B – NOE) and PS (C – APT, D – NOE), respectively, as a function of the number of saturation pulses (also recalculated as time-per-volume or dynamic scan time) at a B1 of 1µT and various water-T1 relaxation times.

Fig. 2. The simulated 3D plots of APT (A and C for SS and PS, respectively) and NOE (B and D for SS and PS, respectively) versus B1 amplitude and duty cycle. The optimum saturation pulse parameters were determined from Fig. 1.

Fig. 3. The simulated B1 dependence of APT and NOE effect sizes in WM for both SS and PS schemes.

Fig. 4. The experimental B1 dependence of APT (A) and NOE (B) effect sizes in WM and GM for both SS and PS schemes. The error bars represent standard deviation across 5 subjects.

Fig. 5. (A) and (B) Saturated images for one anatomical slice as a function of off-resonance frequency saturation for SS and PS, respectively. Middle images are dark due to effective on-resonance water saturation. (C) and (D) MTRasym maps were calculated based on the images in (A) and (B), respectively.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)

2900