Accuracy of PRFS Temperature Mapping using Fast Interleaved Sequences for MR-HIFU Thermal Therapies
Steffen Weiss1, Jochen Keupp1, and Edwin Heijman2

1Tomographic Imaging, Philips Research, Hamburg, Germany, 2Oncology Solutions, Philips Research, Eindhoven, Netherlands

Synopsis

MR-HIFU therapies require stable PRFS-based temperature monitoring over the treatment duration (>30min). Drift effects may be corrected using independent temperature information e.g. based on T2 or T1 maps and simultaneously obtained using fast interleaved sequences. As interleaves may impact PRFS-accuracy, a dynamic on-off-interleaved sequence pattern is introduced to selectively measure their influence on the PRFS phase. The mechanism of the phase variations is explored using multiple variants of the interleaved sequence. It is shown that eddy currents induced during the interleaved sequences represent the main cause of the phase variation. The spatio-temporal characteristics of the effect are explored as a first step towards compensation.

Objective

MR guided high-intensity focused ultrasound (MR-HIFU) has been established as a treatment option that elegantly combines two non-invasive technologies. It is performed for ablation, hyperthermia and hyperthermia-mediated local drug delivery and commonly monitored by proton resonance frequency shift (PRFS)-based sequences. Especially hyperthermia requires a temperature accuracy of 1K for at least 30min. Baseline drift and other phase errors (e.g. by motion) may be corrected using an additional sequence to measure absolute temperature information. Recently, ultra-fast interleaving of arbitrary sequences was demonstrated on clinical scanners [1] and applied to interleaved PRFS- and T2-based temperature mapping [2]. Fast interleaved sequences may compromise the accuracy of the PRFS method, e.g. via spin effects or residual eddy currents. Here, various methodological effects of such fast interleaved sequences on the PRFS phase are examined that could limit temperature accuracy.

Methods

The influence of interleaving on the PRFS phase was examined by MRI of a spherical phantom (270mm diameter) using the body-coil of a 1.5T scanner (Philips Achieva). Individual dynamics of a multi-slice T1w-FFE-EPI PRFS sequence (three slices acquired in 5.4s, TE/TR=19.5ms/55ms, 288² reconstruction matrix, (1.4mm)² x 4mm voxels) were interleaved with TFE-shots of a saturation-prepared Look-Locker sequence (single slice, TFE-shot duration of 1.5s, 10 phases per shot). In order to identify the effect of interleaving, a dedicated on-off pattern was introduced (Fig.1, top). For each dynamic, the average phase in ROIs of 8x8 pixels was calculated per slice. To quantify the contribution of the interleaved sequence to the phase variation, a square wave model was fitted to the average phase values as a function of the dynamic scan number (Fig. 1, bottom). The amplitude p0 of the square wave was evaluated for different experimental conditions to explore the mechanism of the observed phase variation. These included the variation of the slice position of the interleaved sequence relative to the PRFS slices, the time order of the acquisition of the PRFS slices, an extra delay before the PFRS sequence, comparison of the effect in gel and water to test for any magnetization transfer effects, and omitting all RF and/or all gradients for the interleaved sequence.

Results

A high value for p0 was only found in the PRFS slice acquired first in time after the interleave (Fig.2). However, a variation of the position of the interleaved slice relative to the PRFS slices did not change the p0 value markedly. A comparison of measurements in gel and water samples of about equal T1 did not result in any marked differences either. The phase amplitude p0 vanished when all gradients of the interleaved sequence were switched off, but p0 remained unchanged when omitting the RF (Fig.3). Initial delay experiments performed for a selected ROI revealed an exponential decay of p0 (τ ≈ 500ms). An evaluation of p0 for ROIs across the entire phantom revealed a pronounced position dependence (Fig. 4a), which required careful re-evaluation of various earlier experiments. The repetition of this experiment with delays showed that the position dependence is also time dependent (Fig 4 b,c,d). Deliberate detuning of the eddy current compensation of the MR system resulted in a largely amplified effect (Fig 5).

Discussion

Initial experiments had been designed to test the hypothesis that residual magnetization excited by the interleaved sequence contributes to the PRFS phase variations, which however proofed wrong. The gradient on/off experiment clearly hinted to an effect caused by eddy currents. Clinically relevant temperature changes may be mimicked if this effect is not considered (a phase error of 100mrad corresponds to a temperature error of 1.3K). The observed phase errors depend on gradient design and parameters that influence the time between the interleave and the acquisition of the k-space center of a PRFS slice, which provides options for optimization. All measurements so far have been performed on the same 1.5T system, and further experiments are required to explore whether the effect requires calibration per sequence, per system or per system type. Potentially, a further refined eddy current compensation technique can be used to minimize the effect.

Conclusion

Fast interleaving of additional sequences with PRFS temperature mapping can result in phase errors due to eddy currents induced by the interleaved sequence that mimic clinically relevant temperature changes. Further work is required to find adequate techniques to either minimize the effect or to calibrate for it.

Acknowledgements

No acknowledgement found.

References

1. Henningsson M, et al. Magn Reson Med 2015;73:692–696

2. Keupp J, et al. Proc. ISMRM 2015, p4061.

Figures

Fig. 1: An on-off-interleave pattern was introduced to solely identify and to quantify the influences of the interleaved sequence on the PRFS phase. The resulting stepwise variation as a function of the dynamic scan number was evaluated for various experimental conditions to explore the mechanism of the effect.

Fig. 2: For ascending PRFS slice acquisition order (pos1, pos2, pos3), a phase step occurs only at pos1, for descending order (pos3, pos2, pos1) only at pos3. Obviously, only the PRFS slice acquired first in time after the interleave is affected. No influence of the flip angle of the interleaved sequence on the phase variation is observed.

Fig. 3: Phase steps occurred if and only if the gradients of the interleaved sequence were switched on, irrespective of the RF being switched on or off.

Fig. 4: Position and time dependence of the phase error in mrad in a coronal slice at the iso-center. Each voxel corresponds to an 8x8 pixel wide ROI in which the mean phase was evaluated. The green oval ring visible in (a) is due to voxels that have been omitted because of phase wraps.

Fig. 5: Deliberate detuning of the eddy current compensation resulted in a strongly increased phase variation and different spatial characteristics. Note the enlarged range (±800 mrad) with respect to Fig.4.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3617