In high field MRI, fast and robust in vivo B₁⁺ mapping is an essential prerequisite for many quantitative or parallel-transmit-based applications (1-3). However, most B₁⁺ mapping techniques are too slow for seamless integration into the clinical workflow. Recently, the DREAM (Dual Refocusing Echo Acquisition Mode) B₁⁺ mapping approach has been introduced (4), allowing RF shimming in a couple of seconds at low SAR without compromising the RF shimming performance (5). The present study aims at further refining the DREAM-based B₁⁺ calibration towards a silent, hidden preparation phase. For this purpose, the sequence is streamlined with respect to acoustic noise reduction. Moreover, a respiratory motion sensor is used to synchronize the short single-shot acquisitions of the individual tx-channel maps with successive expiratory phases. The feasibility of the approach is shown for in vivo B₁⁺ calibration of a clinical dual-channel MRI system at 3T.

The DREAM sequence employs a STEAM preparation module for B₁⁺ encoding and a low-angle pulse train for the quasi-simultaneous acquisition of free-induction decay (FID) and stimulated echo (STE) signals as gradient-recalled echoes under a single readout gradient lobe, separated in time by appropriate gradient encoding (Fig.1). Thereby, the polarity of the STEAM dephaser gradient G_m2 determines the temporal order of the FID and STE signals. Given a fixed echo time of the FID (TE_FID=2.3 ms for water-fat in-phase at 3T), the STE-first order generally results in a shorter TR of the sequence. However, strong dephaser and spoiler gradient areas G_m1 and G_m3, respectively, are required for proper encoding and spoiling (Fig.1a). In combination with the tight timing, this results in strong peak gradient strengths. In contrast, the more relaxed timing and the smaller dephaser and spoiler gradient areas of the FID-first order (Fig.1b) may be utilized to decrease the gradient demands of the sequence, thus reducing the acoustic noise.

Volunteer experiments were performed on a clinical 3T MRI system (Ingenia, Philips Healthcare, Best, The Netherlands). Abdominal DREAM B₁⁺ calibration scans were acquired in 2D single shot acquisition for both echo orders (transversal orientation, slice width = 20 mm, FOV = 530x460 mm², scan matrix = 64x56, STEAM/imaging flip angle = 50°/10°, shot duration ≈ 250 ms). The respiratory bellows of the MRI scanner was placed on the abdomen of the volunteer and used to synchronize the short single shot DREAM acquisition with end-expiration. Thus, the total scan interval for the two tx channels was two respiratory cycles during free-breathing (≈ 10 s). The acoustic noise in terms of the sound pressure was measured using a linear PCM recorder (TASCAM DR-05) . The relative sound pressure level L_p (in decibel) for the two echo orders was determined according to:

$$L_{p}=20\cdot\log_{10}{\frac{p_{FID\_first}}{p_{STE\_first}}}$$

where p_{FID_first} and p_{STE_first} denote the root-mean-square sound pressures for the the respective echo orders. In addition, the measured in vivo B₁⁺ maps were visually inspected and compared.

The FID-first order reduced the required peak gradient strength from 20 mT/m to 4 mT/m, which resulted in a significant reduction of the sound pressure level by approx. 10 dB (Fig.2). The increased TR (from 3.8 ms to 4.6 ms) led to a moderate increase of the shot duration from 213 ms to 252 ms. Moreover, the SAR level was slightly reduced due to the stretched RF pulses (STE-first: local torso SAR < 16%, whole body SAR < 0.5 W/kg, FID-first: local torso SAR < 11%, whole body SAR < 0.4 W/kg). For both timing schemes, the in vivo B₁⁺ maps show the typical shading pattern characteristic for the employed dual channel system (Fig.3), and no significant differences could be observed.

The proposed sequence enables fast, free-breathing, low-SAR and almost silent B₁⁺ calibration on a dual transmit MRI system. Hence, it potentially provides all required ingredients for turning multi-channel B₁⁺ calibration into a robust preparation phase, without adding significant overhead to the examination. Moreover, the preparation phase can be hidden from both, patient and MR technician. Thus, patient discomfort and image quality problems related to breath-holding, acoustic noise, and warming can be reduced, which represents a strong improvement in workflow over the existing B₁⁺ mapping technique. The individual DREAM B₁⁺ maps are acquired in only a fraction of a second, and hence, freeze respiratory quite efficiently. Thus, the motion sensor is merely used to acquire all B1+ maps in roughly the same respiratory state, and therefore, other motion sensor concepts could be used likewise.