Jean-David Jutras1, Keith Wachowicz1,2, and Nicola De Zanche1,2
1Oncology, University of Alberta, Edmonton, AB, Canada, 2Medical Physics, Cross Cancer Institute, Edmonton, AB, Canada
Synopsis
Multi-echo bipolar pulse sequences are
becoming increasingly popular in structural brain imaging applications that
require high SNR efficiency and minimal geometrical distortions or water-fat
shifts. However, various discrepant expressions for the SNR as a function of T2*
and the sampling bandwidth were previously reported, making it unclear as to how
multiple bipolar echoes should be combined to maximize SNR. In this study, we
compare some traditional single-echo (low-bandwidth) FLASH and MPRAGE sequences
with their multi-echo bipolar (high bandwidth) counterparts and validate the
SNR theory via phantom measurements. Bipolar MPRAGE/FLASH yield SNR gains of ~1.3/1.6-fold,
in good agreement with the theory.
- Purpose
High-bandwidth bipolar multi-echo gradient
echo pulse sequences have become increasingly popular for 3D structural brain
imaging due to advantages such as reduced susceptibility-induced geometrical
distortions and
water-fat shifts [1–3]. However, the SNR efficiency (SNR per square root of the
total scan duration) of these sequences has not been studied in detail, and
various inconsistent expressions for the SNR model as a function of T
2*
and the sampling bandwidth have been reported in the literature [4–6], making
it unclear as to how the multiple echoes should be combined to maximize the
SNR. In this study, we compare some traditional single-echo (low-bandwidth)
FLASH and MPRAGE sequences with their multi-echo bipolar (high bandwidth)
counterparts and validate the SNR theory via phantom measurements.
Theory
According to [4], the SNR of a full echo
(i.e. TE ≈ Tacq/2) as a function of acquisition time Tacq is
$$SNR\propto\sqrt{T_{acq}}e^{-T_{acq}/(2T_{2}^{*})}$$ Eq. 1
In the case of a multi-echo sequence, the root-sum-of-squares
(RSS) of the magnitude images (a.k.a. Multi-echo Recombined Gradient Echo or
MERGE) is the optimal combination that maximizes the SNR in the presence of T2* decay [7]. In
this case, Eq. 1 becomes
$$SNR\propto\sqrt{T_{acq}}\sqrt{\sum_{n=1}^Ne^{-2TE_{n}/T_{2}^{*}}}$$ Eq. 2
Incorporating
the effect of parallel imaging acceleration and the signal yield as a function
of all the scan parameters yields
$$SNR\propto\ \frac{\sqrt{T_{acq}}}{g\sqrt{R}}\cdot S_{0}\cdot D(TE;T_{2}^{*}), D= \begin{cases}e^{-TE/T_{2}^{*}}, & N_{echo}=1\\\sqrt{\sum_{n=1}^{N_{echo}}e^{-2\cdot TE_{n}/T_{2}^{*}},} & N_{echo}> 1\end{cases}$$ Eq.3
Here, S0
is the magnitude signal yield as a function of T1 and extrinsic scan
parameters [i.e. S0,FLASH=S(α,
TR; T1) , S0,MPRAGE=S(α,
TR, TI, TD, TRMP; T1)]. TI is
the inversion time, TD is the delay time, TRMP is the shot duration,
R is the parallel-imaging acceleration factor, g is the geometry factor and TEn
is the echo time(s). Figure 1 shows the predicted SNR (assuming g≈1 as obtained
using regularized SENSE reconstruction) for a single-echo versus a bipolar
multi-echo FLASH sequence as a function of Tacq
in (a) and TR in (b).
Methods
The
SNR theory of Eqs. 1 and 2 was verified experimentally using 4 MnCl2-doped
gelatin beakers (T2*=10, 21, 39, 76 ms). A single-slice
2D FLASH sequence (α/TR=50°/200ms) was tested with varying Tacq
(1–25ms) and enough echoes to fill the TR (1–32 echoes). To verify Eq. 3, a 3D
8-echo FLASH and a 6-echo MPRAGE sequences were optimized on a Philips 3T
Achieva scanner to yield identical scan time ( <9 min) and comparable
contrast to a low-bandwidth single-echo FLASH and MPRAGE (see the Table in Figure 2). All sequences were tested on a 5-layed agar
phantom doped with MnCl2 to mimic T1/T2*of
various tissue classes. An 8-channel head array operated in SENSE mode was
employed for signal reception. For each agar layer, the SNR was measured in 5
ROIs using the method of [8].
Results
As shown
in Fig. 3, the measured SNR agrees well with the theory (Eqs. 1 and 2). The RSS
combination of echoes causes the SNR to converge toward an asymptote. The
measured SNR gains for 3D FLASH and MPRAGE (Fig. 4a) agree with the theory
(Fig. 4b), except in the bottom layer (short T1/T2* mimicking fat) of the FLASH8 image,
probably owing to a higher B1 non-uniformity (or g>1) than in other
layers. SNR gains of 1.28 and 1.52 for MPR6 and FLASH8, over MPR1 and FLASH1,
respectively, are achieved in layer 3 (T1/T2*=1294/95
ms). Sagittal images of the phantom with ROI locations are displayed in Figure 5.
Note how the contrast in MPR6/ FLASH8 at TE1
is equivalent to that in MPR1/FLASH1, respectively, despite
significantly different scan parameters.Discussion
This
study compares the SNR efficiency of bipolar multi-echo gradient echo sequences
over their single-echo counterparts (FLASH and MPRAGE) and verifies the theoretical
expressions. At 3T, optimized multi-echo sequences can enable SNR gains of
1.3–1.6 with a mere 8-channel receiver array despite 3-fold higher bandwidths,
depending especially on the sequence parameters, and on the T2*.
These gains arise from a combination of increased signal yields (by using
longer TR and higher flip angles), and the combination of multiple
echoes. The measured SNR gains in the agar phantom agree well with the theory, as long as moderate regularized SENSE accelerations
are employed (≤3-fold with an 8-channel head array) which ensures that g ≈ 1.Conclusion
Multi-echo bipolar sequences are
recommended for structural brain imaging applications where high SNR and geometrical
fidelity are required, (e.g. radiation treatment planning and image-guided
surgery). However, since local SNR gains may depend considerably on hardware
and parallel imaging implementation, careful phantom validations (as performed
in this work) should be performed prior to routine clinical use. Acknowledgements
Funding from the Alberta Cancer Foundation,
Alberta Cancer Research Institute, and Natural Sciences and Engineering
Research Council (Canada) is gratefully acknowledged. We also thank Philips
Healthcare for technical support and Dr. Roger Luechinger for the PATI program
used for data transfer.References
1.
Helms G. and Dechent P.
Increased SNR and reduced distortions by averaging multiple gradient echo
signals in 3D FLASH imaging of the human brain at 3T. J Magn Reson Imaging
2009; 29:198-204.
2.
Fischl B, Salat DH, Van Der
Kouwe AJW et al. Sequence-independent segmentation of magnetic resonance
images. Neuroimage 2004; (23 Suppl 1): S69-S84.
3.
Van Der Kouwe AJW, Benner T,
Salat DH et al. Brain morphometry with multi-echo MPRAGE. Neuroimage 2008; 40:
559-569.
4. Fleysher L,
Fleysher R, Liu S et al. Optimizing the precision-per-unit-time of quantitative
MR metrics: examples for T1, T2, and DTI. Magn. Reson. Med. 2007;57:380–7.
5. Vinitski S,
Griffey R, Fuka M et al. Effect of sampling rate on magnetic resonance imaging.
Magn Reson Med 1987; 5:278-285.
6. Rahmer J, Bornert
P, Groen J et al. Three-dimensional radial ultra-short echo-time imaging with
T2 adapted sampling. Magn Reson Med 2006; 55:1075-1082.
7.
Jutras J-D, Wachowicz K, Gilbert
G and De Zanche N. SNR efficiency of combined bipolar gradient echoes: comparison
of 3D FLASH, MPRAGE and multiparameter mapping with VFA-FLASH and MP2RAGE. Magn
Reson Med 2016; (early view). DOI 10.1002/mrm.26306
8.
Preibisch C and Deichmann R. T1
mapping using spoiled FLASH-EPI hybrid sequences and varying flip angles. Magn
Reson Med 2009; 62: 240-246.