Synopsis

Optimising CEST sequences for clinical use is difficult due to the lack of representative phantoms. Healthy volunteers do not show the variation in pH or concentration that these sequences seek to detect. However, in this work we show how the inherent T₁ sensitivity of CEST sequences [1] can be exploited to optimise them in healthy volunteers. We demonstrate that the sequence conditions that maximise the grey/white matter contrast in exchange maps are also the parameter conditions that maximise the exchange sensitivity. This method provides an effective way to optimise in vivo CEST sequences without the need for phantoms or simulations.

Purpose/Background

Optimising CEST sequences for clinical use is difficult due to the lack of representative phantoms. Testing development sequences on patients is not always feasible, and healthy volunteers do not show the variation in pH or concentration that these sequences seek to detect. As a result, much of the sequence optimisation is done in phantoms or by numerical optimisation. These alternatives often fail to capture the in vivo environment as exact exchange parameters are unknown or hard to match using phantoms.

Here, we show how the inherent T₁ sensitivity of CEST sequences [1], which is often seen as an unwanted effect, can be exploited to optimise CEST sensitivity directly in vivo using healthy volunteers.

The amount of CEST signal encoded in the longitudinal magnetisation of the bulk water per unit time depends on the T₁ of the water. This generates an additional T₁ sensitivity, encoded in the longitudinal magnetisation, that is specific to exchange processes [1]. The perceived CEST effect is often corrected for these T₁ variations.

MTR_asymmetry expresses the decreased longitudinal signal, due to exchange processes, as a fraction of the unperturbed magnetisation [1] (see Fig.1). Maximization of MTR_asymmetry is commonly used to optimise CEST sequences, as an increase in MTR_asymmetry reflects an increase in the exchange sensitivity and/or decrease of direct water saturation (DWS) of the sequence. The increase in MTR_asymmetry is larger for voxels with longer T₁, due to the T₁ dependence [1].

Although differences in CEST exchange rates or concentrations are hard to detect in healthy volunteers, the difference in grey and white matter T₁ is easily detected. We hypothesise that for exchange metrics which scale linearly with this additional T₁ sensitivity, such as MTR_asymmetry, the sequence conditions that maximize this CEST contrast metric are also the conditions that maximise the contrast between the grey and white matter in these maps.

Methods

A conventional 2D CEST sequence was implemented (3T Siemens Verio) with the following parameters; TR/TE=5000/23ms; CEST duty cycle=50%; fifty 20ms Gaussian CEST pulses 180°; Readout pulse 90°; 33-offset z-spectrum samples; Acquisition time 3:25min.

MTR_asymmetry was calculated from CEST data acquired in a healthy volunteer to separate the contribution of the exchange processes from that of the DWS effect. Conventional T₁ and T₂ weighting of the images is removed by the normalisation process used to calculate MTR_asymmetry.

Using FSL, grey matter, white matter and CSF masks were obtained from a structural scan. These masks were applied to the MTR_asymmetry maps and the mean values and variances over several hundred voxels of the three types of tissue were calculated. This process was repeated for different CEST flip angles.

Conventional in silica optimisation using two-pool Bloch-McConnell simulations was also performed for comparison [2].

Results and Discussion

Fig. 2,3B confirms our hypothesis and shows that maximum contrast between the grey and white matter in the MTR_asymmetry maps is obtained for the same flip angle that maximises the contribution of the exchange processes as obtained through numerical simulations (Fig. 3A). In line with theory, the CSF, which contains few exchangeable protons, shows little signal. Residual variation in the CSF signal is most likely due to the low SNR of this signal and to partial volume effects. For FA values close to 0° and 360°, where little exchange is expected, the MTR_asymmetry contrast between the grey- and white matter disappears.

MTR_asymmetry does not distinguish between NOE, MT and CEST. Using alternative definitions for the exchange maps, e.g. Lorentzian fitting, it is possible to also use this method to optimise sequences for specific exchange mechanisms (see Fig. 1). However, many applications (such as stroke) optimise only for maximum exchange sensitivity relative to DWS. This to boost the low CNR, which is expected to disappear in regions of low pH, generating maximum contrast. Post-acquisition model fitting methods are used to separate the different exchange contributions. As such, this method is not limited to MTR_asymmetry but MTR_asymmetry often suffices.

Factors other than T₁, can induce a contrast between the grey and white matter in the MTR_asymmetry maps. However, the origin of the contrast does not matter. As long as there is a detectable contrast that scales with the exchange process, optimisation and comparison of sequences in the same subject is possible.

Conclusion

Acknowledgements

References

[1] P.C.M. van Zijl, N.N. Yadav, Chemical exchange saturation transfer (CEST): what is in a name and what isn’t?, Magn. Reson. Med. 65 (2011) 927–48. doi:10.1002/mrm.22761.

[2] K. Murase, N. Tanki, Numerical solutions to the time-dependent Bloch equations revisited., Magn. Reson. Imaging. 29 (2011) 126–31. doi:10.1016/j.mri.2010.07.003.