In 2010, Sacolick et al. (1) introduced Bloch–Siegert (BS)-based B1+ mapping to MRI. Recent studies have shown that the portfolio of BS-based B1+ mapping sequences can be extended from gradient echo (GE) sequences (BS-FLASH) and standard spin-echo sequences (BS-SE; Ref. 1) to multi-SE and turbo-SE sequences (2–4).
As previously described (1–4), off-resonant pulses in BS-based B1+ mapping are applied after signal excitation to introduce a B1+-dependent phase into the signal. If certain requirements are met (1, 2), the introduced phase will depend on the parameters used for the off-resonant pulses (BS pulses). Thus, the BS pulse magnitude (B1), the angular off-resonance frequency (ωoff) of the BS pulse, and the BS pulse duration (tBS) all determine the phase shift introduced prior to signal acquisition.
Because of the time consuming BS pulses used in BS sequences, the echo time (TE) must be prolonged. Consequently, susceptibility artifacts that can interfere with B1+ mapping are worsened in GE-based BS methods such as BS-FLASH (3, 5). Unfortunately, susceptibility artifacts are especially problematic at high-field strengths (6, 7) in which B1+ mapping is of special interest as the B1+ field uniformity can be strongly hampered (8, 9). Because of their refocusing properties, SE and multi-SE-based BS methods minimize the problem of susceptibility artifacts. Unfortunately, those imaging methods involve high specific absorption rates (SARs) greater than those of GE-based BS methods. Consequently, BS-based SE/multi-SE/turbo-SE sequences have limited application with high-field systems. Thus, a SE-based BS B1+ mapping method with a low SAR would be desirable for high-field MRI.
To fill this gap, this study proposes a fast BS B1+ mapping sequence based on the SE-BURST sequence (10–12). Several advantages are combined in a BS-SE-BURST sequence.
(a)Unlike a comparable BS-SE sequence, the SAR is reduced by a factor equivalent to the number of excitation pulses.
(b)Fast B1+ mapping is possible as multiple excitation pulses are combined with acquisition of differently phase-encoded echoes in one repetition time (TR) cycle.
(c)Unlike a GE-based sequence, the BS-SE-BURST sequence intrinsically minimizes susceptibility artifacts due to its refocusing properties.
These advantages secure the BS-SE-BURST sequence as an excellent candidate for high-field BS-SE-based B1+ mapping.
The proposed BS-SE-BURST sequence was investigated in phantom and volunteer experiments at 3 T using a standard clinical MR scanner. In phantom experiments, the proposed sequence was compared with BS-SE sequences in regard to the signal-to-noise ratio (SNR), SAR, the standard deviation of the B1+ maps (δB1+), and the experiment time (Texp). Volunteer experiments further demonstrated the applicability of the proposed method in clinical situations.
MATERIALS AND METHODS
Figure 1 shows the two BS sequences compared in this study. In Fig. 1a, a diagram of the BS-SE sequence similar to the one introduced by Sacolick et al. (1) is displayed. The corresponding simplified phase graph of Fig. 1b shows the additional phase introduced by the BS pulses applied around the refocusing pulse. The proposed BS-SE-BURST sequence is shown in Fig. 1c. Unlike the BS-SE sequence, multiple excitation pulses were applied prior to the BS pulses and the refocusing pulse. Thus, the number of relevant phase configurations that were started equaled the number of applied excitation pulses (cf. Fig. 1d). As all excitation pulses were applied prior to the first BS pulse, the same B1+ dependent phase was introduced in all phase trains. Thus, BURST-based BS-B1+ mapping was enabled. The number of applied excitation pulses has been termed BURST factor (BF) in the following sections. Unlike a SE sequence with the same TR, using the proposed BS-SE-BURST sequence reduces the Texp by the BF. This is because a different phase-encoded echo is acquired for each excitation pulse. Please note that the SNR of the BS-SE-BURST sequence in this case is less than that of the BS-SE sequence. In the following sections, the BS-SE-BURST sequence has been termed BS-BURST.
To enable B1+ calculation, all BS experiments consisted of two scans (1, 2). With the second scan, the signs of ωoff were opposite to those of the first scan.
All BS sequences shown in Fig. 1 were implemented on a 3 T whole body human scanner (Magnetom Trio, Siemens AG, Erlangen, Germany). For excitation, the body coil of the scanner was combined with a 12-channel receive-only head matrix array.
BS Pulse Parameters
For all B1+ mapping experiments, the same gaussian-shaped pulses were used as off-resonant BS pulses. Furthermore, tBS was set to 5.12 ms, and the off-resonance frequency was set to ±5 kHz (ωoff = ±31.4 × 103 rad/s) for all experiments. For both BS-SE and BS-BURST sequences, the sign of ωoff for the BS pulse applied after the refocusing pulse was reversed to the ωoff of the initial BS pulse.
The phantom consisted of six 15-mL and two 50-mL tubes (Greiner Bio-One GmbH, Frickenhausen, Germany) placed inside a cylindrical tube filled with hydroxyethyl cellulose dissolved in tap water (3 g/100 mL). To obtain a variety of T1 and T2 relaxation times, the inner tubes were filled with different concentrations of customized cross-linked iron oxide (CLIO) particles in tap water (30 nm hydrodynamic diameter). The T1 values of the phantom at 3 T were between 2.3 and 3 s and the T2 values were between 0.07 and 2 s.
MR Sequence Parameters
For all experiments, a single slice of the phantom was imaged using a field-of-view of 250 × 250 mm2, a measurement matrix of 128 × 128 points, and a slice thickness of 5 mm. Sinc-shaped pulses lasting 0.8 ms for excitation and 2.56 ms for refocusing were used. Furthermore, all experiments were preceded by two dummy scans. In BURST-based imaging, the phase of the excitation pulses was varied following Ref. 13, which allowed adjustment of the excitation pulse flip angle α to:
This scheme led to a better SNR than using α = 90°/BF, which is necessary without phase cycling (14). To enable correction of BS-BURST phase errors, an additional phase correction scan was acquired without phase encoding (15) and with the BS pulse magnitude set to zero. The data were corrected before postprocessing using the information from these phase correction scans. All further sequence parameters for the phantom experiments are provided in Table 1. All experiments were repeated five times to allow quantitative comparison of the different setups.
Table 1. Sequence Parameters for the Performed Phantom Experiments
Flip angle (°)
The results from all experiments are shown in Figs. 2 and 3. TE indicates the echo time, BF is the BURST factor, and TR is the repetition time. Please note that the BS-SE sequence equaled BF = 1 for the BS-BURST sequence. A total of 36 different sequence setups were performed.
In accordance with institutional guidelines and after obtaining informed consent, a head scan of a healthy volunteer was performed to demonstrate the in vivo applicability of the proposed sequence at clinical field strength. Other than the additional parameters given in Table 2, the same parameters and setup as for the phantom experiments were used. Unlike the phantom experiments, the phase correction scans of both BS measurements were averaged for each experiment, and the subsequent average was used to correct both BS measurements. This was done to reduce errors in the correction scans that were more prominent in vivo. Furthermore, two independent repetitions of the volunteer experiments were performed.
Table 2. Sequence Parameters for the Performed Volunteer Experiments
Flip angle (°)
Results from the BS-SE and BS-BURST experiments are shown in Fig. 4. TE indicates the echo time, BF is the BURST factor, and TR is the repetition time. Unlike the phantom experiments, not all of the experiment setups were tested as the sequences with long TR and low BF would have lasted too long.
Postprocessing was performed in MATLAB (The MathWorks Inc., Natick, MA). The B1+ maps were calculated using the equations previously described (1, 2). To summarize, the difference of the BS phase shift (ΔφBS) was calculated from two BS scans with the second scan having opposite ωoff signs. As described in Refs. 1 and2, ΔφBS was converted to the magnitude of a hard pulse (B1_hp):
Here, γ is the gyromagnetic ratio, K = 114.3 (μT2/rad) is the pulse-specific constant of a hard pulse, and KBSpulse = 0.375 is the norm factor for the chosen normalized gaussian pulse (B1normalized), and as two BS pulses flanking the refocusing pulse were applied with opposite sign of ωoff, tBseff equaled two times tBs. N refers to the element number of the shaped pulse.
As previously described, Eq. 2 is only applicable if the following conditions are met (1, 2):
To allow a comparison between BS-SE and BS-BURST phantom experiments, different quantitative parameters were derived from the five independent experiments. For the different experiment setups, SNR calculation for each of the five experiments was performed as follows:
with M being the image pixel intensity and σ is the standard deviation of a noise-containing ROI. The mean SNR of each pixel was calculated by averaging the SNR data of the five experiments. The mean SNR of all pixels containing signal was calculated in a final step.
Similar to Ref. 2, the average standard deviation of the B1 value (δB1) was obtained pixelwise from the five independent experiments. This was done to obtain a quality measure for the B1+ maps:
with I being the number of performed experiments, i is the experiment index, B1i is the B1 value of the ith experiment of one specific pixel, and B1 is the average B1 value of this pixel. Furthermore, the mean δB1 of all pixels containing signal was calculated for each experiment setup in a final step.
The relative SAR was estimated for the used experiment setups following Refs. 2, 16–19:
with J being the number of pulses applied in one TR cycle, j is the pulse index, αj is the flip angle of pulse j, Pdurj is the duration of pulse j, and Kpj is the pulse shape-specific constant of pulse j. For SAR estimation, sinc-shaped excitation and refocusing pulses (Kpj = 5.25) and gaussian-shaped BS pulses (Kpj = 1.71) were used.
Using Eq. 9 allows the determination of SAR of a BS-BURST sequence relative to that of a BS-SE sequence. If BF excitation pulses are used in the BS-BURST sequence, the flip angle of each pulse is reduced by a factor of √BF (Eq. 1). Thus, according to Eq. 9, the complete deposited power of all BURST pulses is the same as for the 90° excitation pulse of the BS-SE sequence. Therefore, the same energy is absorbed in one TR cycle for both sequences. Compared with a BS-SE sequence with the same Texp, the BS-BURST sequence provides a reduced SAR. This is achieved by setting TRBURST = BF·TRSE, resulting in the following relative SAR:
Figure 2 shows the calculated B1+ maps obtained from exemplary BS-SE and BS-BURST experiments. An overview of Texp, TR, and the BF is provided in Fig. 2a and is given as a matrix indicating all performed experiments.
The result from the longest BS-SE B1+ mapping experiment performed (TR = 6.4 s and Texp = 26 min 40 s) is shown in Fig. 2b as reference. Figure 2c shows an exemplary BS-BURST experiment (TR = 0.8 s, BF = 16, and Texp = 13 s). Although the B1+ pattern was similar to that of the reference, greater noise corruption was visible in the BS-BURST experiment.
All experiments displayed in Fig. 2d had the same overall Texp = 51 s and are indicated as the antidiagonal elements of the Texp matrix in Fig. 2a. Although similar B1+ values as with the reference experiments were achieved, slight artifacts were visible in the phase encoding direction of experiments with short TR (cf. Fig. 2d, arrows).
Figure 3 provides a quantitative comparison of the different phantom experiments in terms of SNR, δB1, SAR, and Texp. All data are presented in the same matrix form as in Fig. 2a. Thus, all antidiagonals have the same Texp. Figure 3a provides the measured mean relative SNR values calculated from the five independent experiments for all investigated parameter settings. As expected, the SNR decreased with decreasing Texp. Thus, the longest experiment (BS-SE: TR = 6.4 s and Texp = 26 min 40 s) had the highest SNR and the fastest (BS-BURST: TR = 0.2 s and Texp = 1.6 s) had the lowest. Furthermore, as evident from Fig. 3a, experiments with the same Texp had similar SNR values. Figure 3b displays the measured mean δB1 calculated from the five independent measurements. Again, for experiments with the same Texp, similar δB1 values were achieved. Furthermore, δB1 increased with decreasing Texp. In Fig. 3c, the calculated relative SAR values for the different sequences are provided. Contrary to the other investigated parameters, experiments with the same TR had the same SAR values, whereas experiments with the same Texp had different SAR values. Thereby, the SAR increased with decreasing TR. If only sequences with the same TR were analyzed, the SAR values were independent of the BF.
Figure 4 shows exemplary images from the volunteer measurements. In Fig. 4a, the magnitude data of an exemplary BS-BURST experiment (TR = 0.8 s, BF = 4, and Texp = 51 s) are provided. Figure 4b provides the absolute value data from one exemplary BS-SE experiment (TR = 0.2 s and Texp = 51 s). In comparison to the experiments provided in Fig. 4a, a different contrast was given for the BS-SE experiment. Figure 4c provides the B1+ maps obtained from different BS-BURST experiments with BF = 4 and one with BF = 8 (TR = 0.2 s and Texp = 6 s). In Fig. 4d, the B1+ maps calculated from different BS-SE experiments are shown. For all experiments shown in Fig. 4c,d, a similar B1+ pattern was obtained. In support of this assertion, Fig. 4e displays a difference map of the brain showing that the mean difference between two B1+ maps (BS-SE: TR = 0.8 s and BS-BURST: TR = 0.8 s, BF = 4) was 1.5%. In general, the mean difference between the shown B1+ maps of the brain calculated from the BS-BURST sequences and the longest BS-SE experiment ranged from −1.9 to 2.9%. When the BS-SE experiments were compared, the mean difference between the calculated B1+ and the longest BS-SE experiment ranged from −2.1 to 2.4%. Because of SAR restrictions, the experiment with TR = 0.2 s (Texp = 51 s) was the fastest possible using the BS-SE sequence with the chosen parameters.
This study demonstrates that BS-BURST sequences enable SE-based B1+ mapping with significantly lower SAR than BS-SE sequences with the same Texp. Furthermore, for the same Texp, both approaches resulted in similar B1+ map qualities. Additionally, the BS-BURST sequence enables faster acquisition of B1+ information. This, however, comes at the cost of reduced B1+ map quality.
As previously mentioned in Ref. 1, different T1/T2 values had only a small influence on B1+ mapping regarding phantom measurements. However, the quality of the B1+ maps was influenced by the reached SNR (2). This can be seen when the absolute value data inlay in Fig. 2d and the corresponding B1+ map are correlated. Although no significant influence on the B1+ values was observed, different SNR values due to different T1/T2 weighting of the absolute value data resulted in different levels of local B1+ map quality (asterisks, Fig. 2d). As this issue was previously investigated in greater detail (2), only the mean SNR for the different parameter setups was investigated in this study and correlated to the δB1, SAR, and Texp (cf. Fig. 3). In general, experiments with the same Texp had similar SNR and δB1 values (cf. Fig. 3a,b). However, the SAR value of BS-BURST sequences was lower by a factor of BF than that of the used BS-SE experiment with the same Texp (cf. Fig. 3c). Thus, the BS-BURST sequences using this setup theoretically enabled a Texp of up to 32 times faster. For the fastest BS-BURST experiments (Texp = 1.6 s), however, no useful B1+ maps could be generated as the SNR was insufficient (data not shown).
Several different BS-BURST and BS-SE in vivo imaging experiments were performed. Similar B1+ maps were obtained for both BS-SE and BS-BURST experiments with BF = 2, 4, and 8 (cf. Fig. 4). Because of SAR restrictions, the minimum TR for the BS-SE sequence was 0.2 s (Texp = 51 s). For the BS-BURST volunteer experiments with BF = 2, 4, and 8, a reduction in scan time of up to factor eight could be achieved (cf. Fig. 4, TR = 0.2 s, BF = 8, and Texp = 6 s). No high-quality B1+ maps could be obtained using BF = 16/32 (data not shown). This could have various causes. For example, the imperfect phase compensation worsens with increasing BF. Furthermore, less signal was obtained when the BF was increased and Texp kept constant. The latter issue is most likely due to two effects: (a) the T2 decay and (b) diffusion effects (20). Increasing the BF factor increases TE. Hence, strong diffusion and T2 weighting can significantly decrease the SNR and, therefore, the B1+ map quality. These effects were probably masked in phantom experiments due to long T2 values in the majority of the phantom compartments.
As mentioned in the Results section, a similar variety of the different in vivo B1+ maps of both sequence types was observed (cf. Fig. 4). This variability in the in vivo B1+ maps is most likely due to effects that similarly influence both BS methods.
Different strategies for fast SE-BURST imaging have been proposed (14, 15, 20) and could be implemented in the future, e.g., to allow even faster acquisition of B1+ maps using BS-BURST. Furthermore, the TE should be optimized to allow in vivo BF factors higher than those of this proof-of-principle study. Moreover, centric encoding of the echo train should further improve the SNR of the BS-BURST sequence. In this case, however, blurring in the phase-encoding direction might influence the B1+ map quality (2) in the presence of short T2. Thus, organs other than the brain normally have a shorter T2 relaxation constant (21, 22), and T2 decreases with increasing field strength (23). An appropriately adapted BF factor, however, should allow for useful BS-BURST B1+ mapping in such situations.
In general, SAR reduction should allow BS-BURST application at high-field strengths within an acceptable measurement time. For example, for application at 7 T, the minimal TR for the used BS-SE sequence would have to be increased by a factor of ∼5.4 (TR ∼ 1080 ms) as SAR α (B0)2. Thus, with the parameters of this study, the fastest BS-SE sequence possible would last ∼4 min 35 s. When the implemented BS-SE-BURST sequence with the same TR and BF = 8 was used, Texp could be reduced to ∼34 s.
In phantom experiments with a short TR, artifacts in the phase direction were visible in B1+ (cf. Fig. 3) and magnitude data (data not shown). These artifacts were not visible in BS-BURST sequences with higher TR and might be due to insufficient spoiling. These artifacts were intrinsically avoided due to T2 relaxation in BS-BURST sequences with the same Texp but a longer TR.
Besides SAR constraints, BS-BURST sequences can potentially overcome T2* problems found with GE-based BS sequences at high-field strengths. The long BS pulses used to encode the B1+-dependent phase into the signal lead to an increased TE for BS-based sequences. Thus, T2* effects become more profound for gradient-based sequences, making application at high-field strengths challenging. Alternatively, the combination of SE and fast GE-based echo-planar imaging (EPI) readout techniques such as SE-EPI (24) or gradient and spin-echo (GRASE) (25), yield reduced T2* effects when compared against solely GE-based techniques. These combined sequences achieve a SAR reduction similar to corresponding BURST sequence types when compared against standard SE/turbo-SE techniques. Even though SE-EPI-based sequences provide a higher SNR than comparable BURST sequences (15), short T2* times and field inhomogeneities can hamper the image quality (26) depending on the applied echo-planar imaging acceleration factor. This is especially problematic at high-field strengths. Because of the SE properties of BS-BURST, such artifacts are reduced (8, 9, 15). Thus, BS-BURST sequences provide a promising alternative for fast B1+ mapping, especially at high- and very high-field strengths.
The proposed BS-BURST sequence is a straightforward extension of the BS-SE sequence (1) for fast B1+ mapping. The BS-BURST sequence is especially interesting at high-field strengths as it combines a low SAR with stability against B0 inhomogeneities. Thus, this sequence could provide a fast B1+ mapping sequence for human scanners at even high experimental field strengths.
The authors would like to thank Ashley Basse-Lüsebrink for her help with this manuscript and Reiner Beringer for providing the CLIO particles. T.C.B.-L, T.K, A.F, and V.J.F.S contributed equally to this article.