An in vivo comparison of the DREAM sequence with current RF shim technology

Abstract

Object

In the present study the performance of the dual refocusing echo acquisition mode (DREAM) B₁ ⁺ mapping sequence is evaluated for RF shimming in the abdomen at 3 T and validated against existing RF shim technology.

Materials and methods

In vivo experiments were performed on 19 normal volunteers using a clinical 3 T dual channel MRI system. For each volunteer three different B₁ ⁺ mapping techniques [DREAM, actual flip angle imaging (AFI) and saturated double angle method (SDAM)] were employed for RF shimming of the liver and to subsequently assess the quality of the obtained RF shim settings in terms of the achieved B₁ ⁺ homogeneity and accuracy of the mean B₁ ⁺.

Results

DREAM-based B₁ ⁺ calibration led to an average homogeneity improvement of 39.1 % (AFI = 38.7 %, SDAM = 38.1 %) and a mean B₁ ⁺ of 90.9 % of the prescribed B₁ ⁺ (AFI = 88.9 %, SDAM = 92.0 %). The duration of the B₁ ⁺ calibration scan was reduced from 30 s (AFI) and 15 s (SDAM) to 2.5 s (DREAM).

Conclusion

DREAM accelerates RF shimming of the liver by an order of magnitude without compromising RF shimming performance.

Introduction

High-field MRI (≥3 T) is prone to B₁ ⁺ inhomogeneities in the body resulting from the dielectric shortening of the RF wavelength. Advanced MRI systems can address these problems by RF shimming based on parallel RF transmission [1–4]. Thus, significant improvements in diagnostic image quality have been obtained by RF shimming for body MRI applications at 3 T with, for example, the liver as the target anatomy [5]. However, RF shimming typically has to be performed in a patient-specific manner requiring an additional calibration scan where B₁ ⁺ maps of the individual transmit coil elements are measured. Most B₁ ⁺ mapping techniques determining the transmit field from the magnitudes [6–20] or phases [21, 22] of two or more acquired images are relatively slow. Therefore, depending on the employed B₁ ⁺ mapping technique multi-channel B₁ ⁺ calibration in the body may take several long breath-holds which has a negative impact on work flow and patient comfort.

The recently introduced ultra-fast DREAM B₁ ⁺ mapping sequence allows RF shimming of a multi-transmit system to be performed in a few seconds [23] thus representing an attractive alternative to current RF shim technology. The general potential of the DREAM technique has been demonstrated extensively at different field strengths and for different anatomies in various studies [23–25].

It is the aim of the present study to provide a careful evaluation of the DREAM approach with respect to RF shimming on a clinical dual channel MRI system. For this purpose a DREAM B₁ ⁺ calibration scan protocol is tailored for a liver MRI, which represents one of the important body MRI applications at 3 T. Furthermore, DREAM RF shimming performance is validated against two well-established B₁ ⁺ mapping sequences, i.e., AFI [17] and SDAM [15], in an extensive in vivo study on 19 subjects.

Methods

MRI system and RF coils

A 3 T clinical MRI scanner (Ingenia, Philips Healthcare, Best, The Netherlands) equipped with a dual-channel Tx/Rx body coil was used. The two linear modes of the body coil are driven by two independent RF amplifiers and RF waveform generators. For the present study the standard RF shimming functionality of the system was used, applying RF pulses with individually adjustable complex weight factors for the two transmit channels. In the following the term “quadrature excitation” is used to refer to weight factors equal to “1,” resulting in an equal magnitude and phase of the circular transmit field components of the two channels, B ⁺_1,ch1 and B ⁺_1,ch2 , in the centre of the empty coil.

B₁ ⁺ mapping sequences

For B₁ ⁺ mapping three different techniques were applied, AFI, SDAM and DREAM, where the first two were provided as a standard functionality by the scan software of the MRI system employed. For the DREAM approach the sequence implementation and B₁ ⁺ map reconstruction as previously described [23–25] has been used as a research software patch. 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. The flip angle α of the STEAM preparation module is derived from the ratio of the FID and STE signals (I _STE, I _FID) according to:

$$\alpha = \arctan \sqrt {{{2I_{{\text{STE}}} } \mathord{\left/ {\vphantom {{2I_{{\text{STE}}} } {I_{{\text{FID}}} }}} \right. \kern-0pt} {I_{{\text{FID}}} }}}$$

(1)

In this work the “STE first” DREAM sequence version was used (cf. Fig. 1 in Ref. [24]) because it generally allows shorter repetition times (TR) of the imaging sequence than the “FID first” version. The echo times were adjusted to achieve spectral encoding times of 2.3 ms (i.e., water-fat in-phase at 3 T) for both STE and FID gradient echoes. The “STE first” version was favored over the fully T₂* (including T₂) compensated “STE* first” version based on the virtual STE because of the more relaxed timing of the STEAM preparation module, resulting in an increased flexibility of the sequence.

Slice profile imperfections potentially impairing the accuracy of B₁ ⁺ mapping [26] were addressed for all three approaches. For AFI [17] 3D spatial encoding was performed using only the central slice of the imaging slab for RF shimming. For SDAM [15] non-selective block-shaped RF pulses were used for RF excitation, and hence B₁ ⁺ encoding, while slice-selective Sinc-Gaussian-shaped refocussing pulses were employed for the applied turbo spin echo (TSE) readout. For DREAM non-selective block-shaped RF pulses were used for STEAM B₁ ⁺ encoding and slice-selective Sinc-Gaussian-shaped RF pulses for gradient echo readout [23, 24].

B₁ ⁺ calibration scans and RF shimming

The three B₁ ⁺ mappers considered were first used as B₁ ⁺ calibration scans to supply complex B₁ ⁺ (AFI) maps of the two transmit channels for RF shimming. For this purpose, the B₁ ⁺ magnitude was determined from the underlying source images according to the B₁ ⁺ encoding principle of the individual techniques [cf. Eq. (1) for DREAM] and the relative transmit phase was directly adopted from one of the source images. The RF shimming routine employed estimates the superposed B₁ ⁺ as a function of a global scale factor f and the relative gain g and phase φ between the two transmit channels,

$$\text{B}_{1,(g,\varphi )}^{ + } ({\mathbf{r}}) = f \cdot (\text{B}_{{1,\text{ch}1}}^{ + } ({\mathbf{r}}) + e^{i\varphi + g \cdot \ln (10)/20} \text{B}_{{1,\text{ch}2}}^{ + } ({\mathbf{r}}))$$

(2)

where the gain is given in decibels, which explains the divisor 20 in the formula. The B₁ ⁺ inhomogeneity expected for certain (g, φ) values is determined in terms of the coefficient of variation (CV) defined as the ratio of the standard deviation σ to the mean μ of the superposed B₁ ⁺ field:

$$CV_{(g,\varphi )} = \frac{{\sigma (\text{B}_{1,(g,\varphi )}^{ + } ({\mathbf{r}}))_{{{\mathbf{r}}\; \in \;M}} }}{{\mu (\text{B}_{1,(g,\varphi )}^{ + } ({\mathbf{r}}))_{{{\mathbf{r}}\; \in \;M}} }}$$

(3)

where M denotes the shim mask, defining the set of valid pixels (or voxels) used for RF shimming. The shim mask is derived from a user-selected region of interest (ROI) and from an intensity-based threshold. The latter identifies B₁ ⁺ measures likely degraded by insufficient signal (e.g., air, lungs, bowels) or vanishing B₁ ⁺ by their low intensity in the underlying source images. From Eq. (3) the optimum gain and phase values are determined by a SAR-constrained exhaustive search in the 2D (g, φ) parameter space (about 10,000 grid search points),

$$\left( {g,\varphi } \right)_{{\text{opt}}} = \mathop {\arg \hbox{min} }\limits_{g,\varphi \; \in \;S} \{ CV_{(g,\varphi )} \}$$

(4)

where S denotes the set of shim values allowed by the SAR model of the system. Finally, to meet the target B₁ ⁺ the global gain factor f is determined from the ratio of the nominal field, B₁ ⁺ _{, nom} and the predicted mean field, μ, according to

$$f_{(r,\varphi )} = \frac{{\text{B}_{{1,{\text{nom}}}}^{ + } }}{{\mu (\text{B}_{1,(r,\varphi )}^{ + } )}}_{{{\mathbf{r}} \in M}} \,$$

(5)

Experimental assessment of RF shim settings

In a second step the three B₁ ⁺ mappers considered were used for direct experimental assessment of the RF shim settings (f, g, φ) obtained from the different B₁ ⁺ calibration scans by acquiring RF shimmed B₁ ⁺ maps. The CV of the measured B₁ ⁺ maps was determined as an inhomogeneity measure and the mean B₁ ⁺ value μ, given in percent of the nominal B₁ ⁺, was considered as a measure for the accuracy of the B₁ ⁺ field. For comparison additional B₁ ⁺ maps were acquired using default quadrature transmission. To improve statistics and facilitate comparison each RF shim setting was assessed by all three B₁ ⁺ mapping methods. Moreover, the CV values measured in B₁ ⁺ mapping scans were compared with the CV values predicted by the calibration scans [cf. Eq. (3)] to check the self-consistency of the different methods.

The homogeneity improvement by RF shimming, in the following dubbed ζ, was quantified as:

$$ \zeta \left[ {\text{\%}} \right] \, = 100 \cdot \left( {1 - \frac{{CV_{\text{shimmed}} }}{{CV_{\text{quad}} }}} \right) $$

(6)

where CV _shimmed and CV _quad indicate the B₁ ⁺ inhomogeneity of the B₁ ⁺ maps acquired with RF shimmed or quadrature transmission, respectively. Hence, ζ = 0 % corresponds to no homogeneity improvement at all and ζ = 100 % corresponds to a perfectly uniform B₁ ⁺ map.

In vivo experiments

Experiments were performed on 19 normal volunteers (18 males, 1 female, min/average/max age = 31/42/58 years, body length = 169/185/198 cm, weight = 56/81/110 kg). All human studies were approved by the appropriate ethics committee and have therefore been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. All volunteers gave their informed consent prior to their inclusion in this study. The volunteer group covers a reasonable range of patient sizes with a body mass index (BMI) from 20 to 30.

For each subject the three B₁ ⁺ mappers were employed for B₁ ⁺ calibration and B₁ ⁺ homogeneity assessment as described in the previous section. Hence, the examination included three B₁ ⁺ calibration scans and twelve B₁ ⁺ mapping scans (see workflow diagram, Fig. 1a).

Fig. 1

Workflow and scan geometry. The employed scan protocol is shown schematically (a). For simplicity the scan order was not varied (pseudo-randomized) throughout the study. The yellow frames shown in (b) indicate the region of interest used for RF shimming

Protocol parameters of the employed calibration scans were:

• 1. AFI: 3D acquisition, FOV = 530 × 464 × 24 mm³, measured voxel size = 8.3 × 10.1 × 8 mm³, TR₁/TR₂/TE = 20/100/2.3 ms, nominal flip angle = 60°, whole body SAR level = 0.5 W/kg, pixel/readout bandwidth (BW) = 0.38/24.17 kHz, scan time = 31.4 s, breath-holds = 2.

• 2. SDAM: 2D acquisition, FOV = 530 × 464 mm², slice thickness = 20 mm, measured pixel size = 8.3 × 10 mm², TR/TE = 755/40 ms, TSE factor = 23, nominal excitation flip angle α/2α = 130°/260°, whole body SAR level = 2.2 W/kg, pixel/readout BW = 0.22/14.17 kHz, scan time = 15.1 s, breath-holds = 1.

• 3. DREAM: 2D acquisition, FOV = 530 × 464 mm², measured pixel size = 8.3 × 8.3 mm², slice thickness = 20 mm, TR/TE_STE/TE_FID = 3.8/1.1/2.3 ms, nominal imaging/STEAM flip angle = 20°/100°, imaging/STEAM RF pulse duration = 0.6/0.4 ms, whole body SAR level = 0.1 W/kg, pixel/readout BW = 0.16/10.28 kHz, shot duration = 0.2 s, shot delay = 2 s, scan time = 2.4 s, breath-holds = 1.

Prior to B₁ ⁺ calibration the standard transmitter calibration of the system was performed on a central slice. Therefore, the indicated nominal flip angles that refer to quadrature excitation overestimate the actual flip angles in the calibration scan by roughly a factor of two because, in actuality, only one single channel, instead of dual channel transmission, is used. The selected AFI and SDAM calibration scan protocols had been optimized and approved by the manufacturer and hence were applied without further modification. The DREAM protocol was tailored and optimized in a pre-study to this work. The delay of 2 s between the acquisition of the two DREAM B₁ ⁺ maps was introduced to avoid saturation effects potentially degrading the accuracy of the method [24].

For the acquisition of the RF shimmed and quadrature B₁ ⁺ maps the calibration scan protocols were used as a template and adapted accordingly. The nominal flip angles were set to 50° (AFI), 65° (SDAM) and 15°/65° (DREAM), individually adjusting to the proper working ranges of the methods. All scans were performed in transversal orientation approximately intersecting the centre of the liver (Fig. 1b). For the shimmed acquisitions a user-selected B₁ ⁺ shim volume was positioned tightly around the torso to exclude the arms from the RF shim ROI used by the RF shimming routine, illustrated also in Fig. 1b.

For complementary assessment of the RF shimming performance in addition to the B₁ ⁺ mapping scans a 3D dual echo Dixon FFE scan (FOV = 520 × 359 × 120 mm³, measured voxel size = 1.9 × 1.9 × 5 mm³, TE₁/TE₂/TR = 1.1/2.0/3.2 ms, flip angle = 5°, whole body SAR level = 0.2 W/kg, WFS = 0.24 pixel, scan time = 16 s, breath-holds = 1) was performed for each shim set. This low-tip angle scan was chosen as an additional and independent visual test for the B₁ ⁺ homogeneity because it is sensitive to B₁ ⁺-induced tip angle variations that can lead to significant contrast changes and loss of image quality in such a diagnostic protocol. For each subject the water image of the central slice was analyzed and scored with respect to the B₁ homogeneity [accordingly, 1 (excellent) to 5 (very bad)] by two different experienced readers.

The net scan time of the overall examination was 5:40 min (cf. Fig. 1a). The measured B₁ ⁺ maps were exported from the scanner database after the examination and stored along with the corresponding RF shim sets for subsequent evaluation using in-house image processing tools written in Java.

Results

Experimental issues

For all volunteers the complete examination could be carried out in about 10–15 min. The long 15-sec breath-holds, required for the AFI and SDAM calibration scans, were perceived as stressful by about 20 % of the volunteers. For one volunteer (#1) the AFI scans (calibration maps and shimmed maps as well) were degraded by respiratory motion (see detailed results below). Moreover, about 30 % of the volunteers mentioned a noticeable warming by the SDAM sequence, which can be explained by the higher, but still well in-spec, SAR level of this TSE-based scan.

B₁ ⁺ calibration scans and RF shimming

Figure 2 shows the transmit channel sensitivities measured using the three different B₁ ⁺ calibration methods and the employed shim mask for two exemplary volunteers (#13 and #1). All calibration methods reveal the well-known diagonal shading pattern for Tx channel #1, typical for this coil configuration in abdominal MRI at 3 T [25]. The AFI maps show a higher background inhomogeneity from noise and artefacts than the SDAM and DREAM maps which was observed for most of the volunteers in this study. Particularly, the AFI maps of volunteer #1 are degraded by respiratory motion artifacts. The intensity-based masks that exclude B₁ ⁺ measures degraded by insufficient signal or vanishing B₁ ⁺ are similar for all three methods. The user-selected shim volume additionally excludes the arms from the shim ROI.

Fig. 2

B₁ ⁺ calibration scans of selected volunteers. Transmit sensitivities of the two transmit channels (top: channel #1; centre: channel #2; bottom: shim mask) as measured by the three calibration sequences (left: AFI; centre: SDAM; right: DREAM) are shown for two volunteers (#13 and #1). The colored frames (top) indicate the user-selected shim volume excluding the arms from the shim mask (bottom). For volunteer #1 ghosting artifacts from respiratory motion are noticeable (arrow). Note that the AFI and SDAM scan protocols employ interpolation to increase resolution in the reconstructed maps

Figure  3a, b show the B₁ ⁺ inhomogeneity predicted from the three calibration scans for the same volunteers (#13 and #1). The 2D plots show the expected CV maps as a function of the gain and phase between the two channels [cf. Eq. (3)]. The centre of the plots corresponds to quadrature shim settings (i.e., zero gain and phase between channels). All plots show a maximum in the left centre and a minimum in the upper centre of the 2D parameter space. The minimum corresponds to the optimum shim setting leading to the most uniform B₁ ⁺ field and will be in the focus of the analysis in the following. Minor deviations between the different mapping methods are visible leading to slightly different predictions for the optimal RF shim setting. Despite the motion artifacts in the AFI calibration maps of volunteer #1 (cf. Fig. 2), the resulting CV map is rather smooth and regular. Comparison between the two subjects shows systematic differences in the optimal shim settings affecting mainly the phase offset between the two Tx channels.

Fig. 3

CV maps. The expected CV of the torso B₁ ⁺ is shown as a function of gain and phase between the two Tx channels for two selected volunteers. CV was predicted from the measured calibration scan (left: AFI; centre: SDAM; right: DREAM) according to Eq. (3). The crosses indicate the optimum RF shim settings yielding the minimum CV

Figure 4 summarizes the optimal shim settings determined from the different calibration scans for all volunteers. For most volunteers the optimum gain and phase between the two channels is about 7–10 dB and 30°–50°, respectively. For two volunteers (#1, #19) a significantly smaller phase between 0°–25° was determined by all calibration methods.

Fig. 4

RF shim settings. Optimal gain and phase between the two transmit channels is shown for the different volunteers and calibration scans

RF shimmed B₁ ⁺ mapping scans

Figure 5 shows the B₁ ⁺ maps measured under different shimming conditions for one volunteer (#13). Each of the four RF shim settings (quadrature, AFI, SDAM, DREAM) was evaluated by each of the three B₁ ⁺ mapping techniques (AFI, SDAM, DREAM) resulting in twelve B₁ ⁺ maps.

Fig. 5

RF Shimmed B₁ ⁺ maps. B₁ ⁺ maps measured with different techniques (top to bottom: AFI, SDAM, DREAM) and with RF shims derived from different calibration scans (left to right: quadrature, AFI, SDAM, DREAM) are shown. CV and the mean B₁ ⁺ (in  % of the nominal B₁ ⁺) in the torso are shown as insets in the images. In addition, low-angle, water-only FFE images are shown for the different shim sets (bottom row) revealing the impact of B₁ ⁺ shimming

For quadrature RF shims the typical shading artifacts are observed in the anterior and posterior torso [5]. A noticeable reduction of the B₁ ⁺ inhomogeneity is observed for all calibration scans and all mapping scans resulting in a strongly reduced CV. Moreover, the deviation of the mean B₁ ⁺ from the nominal B₁ ⁺ is reduced. The AFI maps show again an increased background inhomogeneity from noise resulting in larger CV values. Similar to the B₁ ⁺ maps the quadrature low-angle FFE images confirm these findings showing strong shading that is strongly reduced after RF shimming.

Figure 6 summarizes CV values measured by the different B₁ ⁺ maps for the different RF shim settings for all 19 volunteers. For fixed shims the CV varies between 20 and 35 % and is reduced to 10–25 % after RF shimming. Table 1 summarizes the corresponding results averaged over the volunteer group. The AFI B₁ ⁺ maps show a higher average CV value than the SDAM and DREAM B₁ ⁺ maps which is probably due to the higher noise-related artifact level observed for the employed AFI scan protocol. For one volunteer (#1) ghosting artifacts in the shimmed AFI maps resulted in an apparent increase of the CV after shimming. These obviously erratic data were excluded from further evaluation and, hence, are not shown in the figure. The CV averaged over all subjects and over the three B₁ ⁺ mappers is about 27 % for quadrature excitation and about 17 % for RF shimmed excitation regardless of the employed calibration method; this indicates similar RF shimming performance of the three methods. This is supported by scoring of the low-angle FFE images with an average score of 3.2 for quadrature excitation and about 1.5 for RF shimmed excitation (cf. Table 1). The plots show also the CV values predicted from the B₁ ⁺ calibration scans for comparison. The DREAM method shows the best accordance between predicted and measured CV values with a mean deviation of 1.1 % over the whole volunteer group (Table 2). In comparison, for AFI and SDAM a mean deviation of 2.3 and 3.3 %, respectively, is observed. This indicates that the DREAM data show the best self-consistency.

Fig. 6

B₁ ⁺ homogeneity. CV of the torso B₁ ⁺ maps is plotted for the different volunteers. Four different shim settings (quadrature: top left; AFI: top right; SDAM: bottom left; and, DREAM: bottom right) are evaluated by three B₁ ⁺ mapping techniques (AFI: solid squares; SDAM: solid circles; DREAM: solid triangles). In addition, CV values predicted by the different calibration scans are shown as open symbols. The lines are guides for the eyes. For all calibration methods the CV is significantly reduced compared to quadrature excitation

Table 1 RF shimming performance averaged over volunteers

Table 2 Self-consistency of B₁ ⁺ mappers

Figure 7 shows the improvement in B₁ ⁺ homogeneity, ζ, for the different calibration techniques [cf. Eq. (6)]. For clarity of presentation improvement was averaged over the three B₁ ⁺ mappers. Dependent on the volunteer, an improvement between 20 and 55 % is achieved. For the individual subjects differences between the calibration methods are observed but without a strong and clear overall trend favoring one of the methods. Averaged over all volunteers almost a 40 % improvement is achieved for all methods (AFI: 38.7 %; SDAM: 38.1 %; DREAM: 39.1 %; cf. Table 1).

Fig. 7

B₁ ⁺ homogeneity improvement. The improvement in B₁ ⁺ homogeneity, ζ, is compared for the different shim sets (AFI, SDAM, DREAM). The lines are guides for the eyes. Deviations between the calibration methods are small compared to deviations between volunteers. On average a similar RF shimming performance is observed for all calibration methods (cf. Table 1)

Figure 8 shows the mean torso B₁ ⁺, μ, achieved for the different calibration schemes. For fixed shim settings the mean B₁ ⁺ is between 55 and 80 %, dependent on the volunteer. B₁ ⁺ shimming yields a mean B₁ ⁺ between 80 and 100 %. On average RF shimming increases the mean B₁ ⁺ from about 65 % to about 90 % (AFI: 88.0 %; SDAM: 90.5 %; DREAM: 89.6 %; cf. Table 1). Deviation from the expected 100 % may be mainly attributable to the overtipping protection of the shimming routine, which poses a limit on the maximum B₁ ⁺ also outside the shim ROI. Thus, for the chosen anatomy the overall B₁ ⁺ is scaled down to reduce the relative strong fields in the arms (cf. Figs. 2, 5).

Fig. 8

B₁ ⁺ accuracy. The mean torso B₁ ⁺ is shown for the different volunteers and RF shim settings (given in  % of the nominal B₁ ⁺). The lines are guides for the eyes. All calibration methods increase the mean B₁ ⁺ and achieve on average about 90 % of the prescribed B₁ ⁺ (cf. Table 1)

Discussion

This comprehensive in vivo study shows that the DREAM method fully matches the RF shimming performance of two established B₁ ⁺ calibration methods (i.e., AFI and SDAM) used for RF shimming of the liver on the employed 3 T MRI system. All calibration methods achieved about 40 % homogeneity improvement and about 90 % of the nominal B₁ ⁺ on average. Regarding the individual volunteers slight differences were observed between the methods, however, without a clear trend favoring one of the approaches. Furthermore, differences were hardly noticeable in the low-angle FFE images. Nevertheless, the DREAM approach showed the best consistency between calibration scans and RF shimmed B₁ ⁺ mapping scans. Deviations between the methods may be attributable to the known limitations of the individual approaches (e.g., low dynamic range, SNR, sensitivity to motion, flow, T₁, T₂, etc.), protocol trade-offs, variability in the automatically generated shim masks or physiological effects like the limited reproducibility of breath hold levels or peristaltic motion. It is difficult to analyze the impact of these different aspects separately in an in vivo study and, hence, the attained B₁ ⁺ homogeneity was used as an integral measure for the performance of the methods. Interestingly, the generally lower B₁ ⁺ map quality of the employed AFI calibration scan observed in this study did not lead to a noticeable degradation of RF shimming performance. A potential explanation is the considered two-channel RF shimming problem represents a highly overdetermined optimization problem where one complex number has to be determined from many measured B₁ ⁺ values, making it robust against, for example, noise or motion artifacts in the measured data.

RF shim settings obtained in this volunteer study varied in a range of 10°–60° phase offset and 6–13 dB gain between the two transmit channels. For real patients with certain pathologic findings (e.g., ascites [5]) an even stronger variability of RF shading was reported, further underlining the need of patient-specific RF shimming. The DREAM approach allows these issues to be addressed without adding significant overhead to the examination. It is an order of magnitude faster and has a much lower SAR level than the existing techniques. Thus, patient discomfort and image quality problems related to breath holding can be reduced, which represents a strong improvement in workflow over the existing techniques. For this very first comparative study a quite conservative single-slice DREAM scan protocol has been applied with a comparatively “long” delay of 2 s between the two shots, resulting in a total scan time of 2.4 s performed in a short breath hold. This was done to facilitate a fair comparison of the methods, thus making the study design simple and establishing a baseline for the performance of the DREAM approach. Nevertheless, shorter delays are conceivable and the single-shot characteristic of the DREAM approach enables free-breathing B₁ ⁺ mapping [27] that both could further improve the workflow towards short and hidden preparation phases.

In all volunteers of this study significant improvements in B₁ ⁺ homogeneity were achieved after shimming. The CV maps (cf. Fig. 3) and their good accordance to the measured B₁ ⁺ maps indicate that the calibration scans exploit the RF shimming capabilities of the employed transmit coil to a large extent.

Transmit coils with a higher channel count can further improve B₁ ⁺ homogeneity at 3 T [28]. However, the increased complexity of such a system with respect to hardware demands, B₁ ⁺ interferometry [29–31], RF shimming algorithms and SAR prediction has to be carefully balanced against the potential gain in RF shimming performance possible at 3 T. On the other hand, at ultra high field strength transmit field inhomogeneities will be even more pronounced [24] underlining the need for more than two transmit channels. In this respect DREAM has been shown to be applicable for multi-channel B₁ ⁺ mapping at ultra high field strength [32].

In conclusion, for body imaging at 3 T the DREAM method is as robust as, but much faster, than current RF shim technology. Hence, the RF shimming capabilities of the employed MRI system may be fully exploited using patient-specific calibration without compromising work flow.