Introduction

At higher field strength, fast and accurate B₁⁺-mapping is an essential prerequisite for many applications as e.g. parallel transmit MRI or quantitative methods like MR fingerprinting. The DREAM approach (Dual Refocusing Echo Acquisition Method, (1)) allows volumetric B₁⁺ mapping, covering the whole transmit coil volume in a few seconds, which is more than ten times faster than other B₁⁺ mapping approaches. However, DREAM is based on a STEAM-preparation module, and requires short shot durations to avoid T₁-related signal recovery artefacts. Moreover, short T₁ reduces the accurate working range of the method. Hence, a careful trim of the sequence parameters is required (1), or an appropriate filtering of the FID signal has to be performed in reconstruction (2). In this work, a T₁-compensated phase-cycling scheme based on adiabatic RF inversion pulses, dubbed iDREAM, is presented, which removes the T₁-sensitivity of the DREAM approach completely. iDREAM B₁⁺ mapping was performed on phantoms and in vivo on the brain at 3T to study the performance of this new approach.

Methods

A DREAM phase cycling scheme has been implemented, where signals from two averages, acquired with or without adiabatic inversion pulse, respectively (Figure 1), were subtracted by switching the phase of the RF imaging pulse between 0 and 180 degrees for the two averages. Alternatively, one could also toggle the receiver phase. These measures eliminate the potential T₁ bias in DREAM (Figure 2). An adiabatic inversion pulse was chosen to account for a broad range of Tx sensitivities (pulse angle = 750 deg). B₁⁺-mapping experiments were performed on phantoms (bottle filled with mineral oil) and in vivo (brain) on a 3T MRI system. A DREAM protocol with a long acquisition train was chosen on purpose to increase T₁ sensitivity (scan matrix = 120x90, pixel size = 2.5mm, slice thickness = 7mm, TR = 7.6ms, acq. train duration = 660 ms). A shot delay of 5s was used to allow full T₁ recovery. For the phantom experiments, single-slice T₁-compensated DREAM B₁⁺ maps were acquired for nominal STEAM flip angles between 5 and 90 degrees in steps of 5 degrees. A 10x10 pixel ROI in the centre of the phantom was used to average STE and FID signals, and to derive the STEAM angle according to Eq.4. The flip angles derived by DREAM were plotted against the nominal angles, which were regarded as accurate reference for an oil bottle at 1.5T. The experiments and the analysis were repeated for conventional DREAM without inversion pulse for comparison. For the in-vivo experiments, a transversal slice through the brain (intersecting the ventricle) was chosen and DREAM B₁⁺ maps with and without T₁ compensation were acquired. In addition, also combined T₁- and T₂-compensation was applied by using the virtual stimulated echo (3) in combination with inversion pulse cycling.

Results

The experimental results on the oil phantom show, that the standard DREAM B₁⁺ map starts to degrade above 50 degrees (Figure 3). This is due to the short T₁ of the oil (about 100ms) and the long acquisition duration of the high-resolution DREAM protocol, which was deliberately chosen to emphasize the T₁ bias. In contrast, the (T₁-compensated) iDREAM B₁⁺ maps are accurate up to 85 degrees. Thus, the iDREAM sequence covers a much larger portion of the theoretical working range than the standard DREAM sequence without a bias. In vivo, without T₁ compensation, a strong T₁ contrast between ventricle, grey and white matter and the rim of the head is visible. In case of T₁ compensation, the contrast between rim, grey and white matter disappears, and only slight contrast of the ventricle remains. This tiny remaining contrast disappears in case of both T₁ and T₂ compensation, yielding a perfectly smooth B₁⁺ map (Figure 4). The profile plots show that the T₁ bias of standard DREAM results in an underestimation of B₁⁺ up to 5% in grey and white matter. Note that this T₁ bias was emphasized by the elongated shot duration.

Discussion

The proposed refinement makes the DREAM sequence more accurate, robust, and versatile. The acquisition train duration can be prolonged without performance degradation. Hence, also larger scan matrices may still be acquired in a single shot. Moreover, other view orders than centric, like linear, may be used if appropriate. Compared to advanced filtering performed in reconstruction (2), the DREAM Equation (Eq.4) preserves its simple analytical form. However, as a drawback, acquisition time is doubled. Nevertheless, due to the high speed of the DREAM acquisition, this is acceptable for many applications, especially for quantitative MRI applications, where a better accuracy and a gain in SNR of the B₁⁺ maps are appreciated.

Acknowledgements

No acknowledgement found.