Introduction

CEST MRI signal is usually contaminated by magnetization transfer (MT) from semi-solids and overlapping faster, broad exchanges. Numerous solutions have been proposed to remove confounding effects but these methods often either increase acquisition time or sacrifice sensitivity^1-4. We propose an Average Saturation Efficiency Filter (ASEF) which uses two pulse trains with similar average saturation power but highly unequal duty cycles, so that saturation transfer effects from fast chemical exchange species and semisolid macromolecules are similar between the two pulse trains, but drastically different for slow exchange species. Thus, their difference becomes an exchange rate filter suppressing MT and fast exchange signals. In this work, the signal properties of ASEF were evaluated by computer simulation and validated by phantom experiments.

Methods

Simulations: Assuming a train of block pulses where the RF in each repeating unit has a bipolar pulse (which minimizes the rotation effect ²) with duration t_p (Fig. 1). ASEF takes the difference between CEST signals measured by a continuous wave (or high duty cycle) saturation and a low duty cycle (DC_l) one. CEST signals were simulated by Bloch-McConnel Equations which include 3 exchanging pools of free water protons, labile protons, and bound water protons, assuming a chemical shift between the labile proton and water of 1.9 ppm, a fraction of labile proton of 0.001, the T₁ (T₂) of water, labile proton, and bound water protons is 2 s (66.6 ms), 2 s (66.6 ms), and 2 s (10 ms), respectively.
Phantom experiments: MR experiments were performed at 9.4 T. Seven phantoms were prepared in 10% Bovine Serum Albumin (BSA). 50 mM Creatine was added to six of them and they were then titrated to pH = 6.15, 6.55, 7.04, 7.44, 7.85, 8.25, and 7.0 for the BSA only phantom. These phantoms were heated to 95ºC to denature the BSA and measured at room temperature. The saturation preparation schemes consisted of either a single CW block pulse or a DC_l = 15% and t_p = 24 ms bipolar pulse train, with a duration of 6-s. The B1 was 0.8 μT for the CW and a nominal average B₁ = 0.82 μT was determined for the DC_l pulse train. This 2.4% fudge factor was obtained by matching the saturated signal at 6 ppm (where the CEST effect is minimal) to minimize the mismatch due to our RF power linearity/stability and the mismatch in the MT effect expected from two saturations with highly unequal duty cycles ⁵. Images were acquired by single slice spin-echo EPI.

Results

Fig. 1B shows simulated exchange rate filtering of ASEF. The CEST contrast of the DC_l pulse train is like that of CW at fast exchange rates but is much smaller at slower exchange rates. As a result, the sensitivity of their difference, i.e., the ASEF signal, is only slightly lower than CW for slow exchange rates but is suppressed at k_ex > 2000 s^-1. For larger DCl values, the sensitivity of ASEF decreases but the exchange rate filtering (e.g., peak position and the linewidth) remains the same (Fig. 1C). With a higher B_{1, avg} = 1.6 µT, the ASEF signal peak shifts to a higher k_ex, and CEST contrast is suppressed for k_ex > 4000 s^-1 (Fig. 1D). For creatine in heated-denatured BSA, the Z-spectra of the pH = 7.04 phantom show larger differences in the 1.9 ppm dips for CW than DC_l pulse train saturation (Fig.2A). The spectra match well for offsets > 3.5 ppm indicating the MT effect can be effectively minimized by ASEF. In Fig. 2B, the CEST contrast of CW saturation is calculated by subtracting the signals of phantoms with creatine from that of BSA only. The ASEF signal is only slightly lower than that of CW at lower pH values, but the difference is much larger at higher pH. Note MTRasym showed a negative baseline of about -2% due to the intrinsic asymmetry of the MT effect, as shown in the BSA-only phantom (Fig. 2C). In contrast, the ASEF signal is minimal for the phantom of BSA only and is small for pH = 8.24 (Fig. 2D), indicating that signals from both the MT contrast and from fast chemical exchange can be effectively suppressed.

Discussions

ASEF can improve the CEST signal specificity for slow to intermediate exchanges (e.g., k_ex <2000 s^-1) with only a small reduction in the sensitivity. It can be acquired at as few as only one frequency, i.e., the Larmor frequency of the labile proton of interest. ASEF has three parameters that can be adjusted: the B_{1, avg} determines the range of exchange rate filtering, the duty cycles of the pulse trains determine ASEF sensitivity, and t_p which is sensitive to direct water saturation. The MT mismatch⁵ between the low and high DC saturations may be minimized by careful selection of these parameters, as well as a fudge factor of the B_{1, avg} between the two saturations.

Conclusion

ASEF is a simple method that can minimize the MT effect and provide fast exchange rate filtering for CEST MRI with a relatively small reduction in sensitivity. It can be a highly useful tool for CEST study in the slow to intermediate exchange regime.

Acknowledgements

This work is supported by NIH grant NS100703.