Introduction

Chemical exchange saturation transfer (CEST) imaging is a novel MRI technique that allows indirect detection of exchangeable protons in the water pool. In recent years, several studies have explored various CEST applications in the liver, such as amide proton transfer weighted (APTw) imaging^1-4, glycogen CEST (glycoCEST) imaging^2,3, and glycogen NOE (glygoNOE) imaging⁵. However, CEST MRI of the liver in human studies is still challenging. Breath-holding is currently needed to reduce motion artifacts^1-4, limiting spatial resolution and coverage. Even with multiple breath holds, 3D coverage would require an impractically long scan time.

Steady-state CEST (ss-CEST) is a method that performs pre-saturation and k-space sampling in an interleaved pattern with repeated modules⁶. Recently, we developed a ss-CEST method in the brain using MR Multitasking to accelerate image acquisition and enhance image quality^7,8. In this work, we propose a novel respiration-resolved 3D abdominal Multitasking ss-CEST technique at 3.0 T, which enables whole-liver coverage with free-breathing acquisition. The feasibility of APTw and glycoCEST imaging were evaluated in healthy volunteers.

Methods

Image acquisition
Fig. 1 illustrates the proposed pulse sequence and k-space sampling pattern⁷. Each ss-CEST module contains a single-lobe Gaussian saturation pulse (t_sat = 30 ms, effective B₁ = 0.93 μT), followed by a spoiler gradient and eight 5° FLASH readouts. The center k-space line in the superior-inferior direction was acquired as training data.

Image reconstruction
The abdominal CEST image is modeled as a 6D image $$$a(\textbf{x},z,\tau,r)$$$, with voxel location $$$\textbf{x}$$$, pre-saturation frequency offset $$$z$$$, time within each frequency offset $$$\tau$$$, and respiratory motion phase $$$r$$$. Due to high correlation along all dimensions, it can be modeled as a 4-way low-rank tensor $$$\mathcal{A}$$$: $$\mathcal{A}=\mathcal{G}\times_1{\textbf{U}_{\textbf{x}}}\times_2\textbf{U}_{z}\times_3\textbf{U}_{\tau}\times_4\textbf{U}_{r}$$ where columns of each $$$\textbf{U}_{\textbf{x}}$$$, $$$\textbf{U}_{z}$$$, $$$\textbf{U}_{\tau}$$$ and $$$\textbf{U}_{r}$$$ contain basis functions for space, frequency offset, approach to steady-state, and respiratory motion respectively. $$$\mathcal{G}$$$ denotes the core tensor.

The respiratory motion was binned into 5 states with the help of recorded PMU data extracted from Siemens raw data files. $$$\textbf{U}_{\tau}$$$ was predetermined from a Bloch-simulated dictionary. The core tensor $$$\mathcal{G}$$$ and remaining non-spatial basis matrices $$$\textbf{U}_{z}$$$ and $$$\textbf{U}_{r}$$$ were recovered by low-rank tensor completion of the training data⁷. Then, the spatial bases $$$\textbf{U}_{\textbf{x}}$$$ were reconstructed using the remaining imaging data by wavelet-regularized least-squares fitting to the measured data.
For respiratory motion state $$$i$$$, 3D images of the last time point within each offset frequency (closest to steady-state), i.e., $$$\tilde{a}_{i}(\textbf{x},z)=a(\textbf{x},z,\tau_\max, i)$$$, were extracted for further analysis.

CEST analysis
The central part (|Δω| < 1.0 ppm) of the Z-spectrum was first used to correct for the B₀ field inhomogeneity. The CEST effect was quantified with the magnetization transfer asymmetry ratio (MTR_asym): $$MTR_{asym}(\Delta\omega) = Z(-\Delta\omega) - Z(\Delta\omega)$$ APTw and glycoCEST values were quantified as MTR_asym at 3.5 ppm and 1.0 ppm respectively.

In-vivo experiments
Data were acquired in three healthy volunteers on a 3T MR system (MAGNETOM Vida, Siemens Healthineers). Imaging parameters were: FOV = 384 x 320 x 192 mm³, matrix size = 192 x 160 x 32, spatial resolution = 2.0 x 2.0 x 6.0 mm³. The acquisition time was 11.2 s for each frequency. Images of 53 frequency offsets (from -100 to 100 ppm) were acquired, with prolonged unsaturated acquisition (S₀) at the beginning and the end. The total imaging time was 11 min.

Results

Fig. 2 and Fig. 3 represent (A) CEST image at 3.5 ppm, (B) B₀ map, (C) APTw map (MTR_asym at 3.5 ppm) and (D) glycoCEST map (MTR_asym at 1.0 ppm) at the end-expiration state in a healthy volunteer. The axial view and coronal view are displayed in Fig. 2 and Fig. 3 respectively.

Fig. 4 shows the distribution of APTw and glycoCEST values within the liver region of the central slice in all three volunteers. The APTw was consistent among different volunteers, but there was more intra-subject variance in glycoCEST values.

Discussion

The Multitasking ss-CEST technique was able to acquire 3D liver CEST images without breath-holds, unlike previous liver CEST methods. The in-plane spatial resolution of 2.0 mm was comparable with previous 2D studies^1-3, but the Multitasking 3D slice coverage of 192 mm was much larger, covering the whole liver. APTw values were consistent across healthy volunteers. GlycoCEST values were variable, potentially due to B₀ inhomogeneities, since 1.0 ppm is very close to the water frequency. An averaging method with denser frequency offset sampling around 1.0 ppm may reduce glycoCEST variability.

These results establish the feasibility of free-breathing 3D CEST in the liver at 3.0 T. Future work will focus on optimizing the saturation power B₁, evaluating non-Cartesian readout trajectories to reduce artifacts, incorporating fat suppression modules, and developing application-specific frequency offset sampling patterns. Advanced CEST quantification methods, such as multi-pool fitting or the Lorentzian difference method, will be investigated. Further reduction of total scan time and improvement of the spatial resolution will also be explored.

Conclusion

In conclusion, the proposed 3D abdominal steady-state CEST method using MR Multitasking can generate CEST images of the entire liver during free breathing. It has the potential to push the liver CEST technique towards practical clinical use.

Acknowledgements

This work was supported in part by NIH R01 EB028146, R01 AR066517, and R01 HL156818.