STEAM-Burst: a single-shot, multi-slice imaging sequence without
rapid gradient switching
Y. Crémillieux1, C. Wheeler-Kingshott2, A. Briguet1 and S.J. Doran2
1Laboratoire
de RMN, CNRS UPRESA 5012, UCB Lyon 1CPE, 69622 Villeurbanne Cedex, France
2Department
of Physics, University of Surrey, Guildford, Surrey, GU2 5XH, England
Running Head: Single-shot, multi-slice imaging using STEAM-Burst
Proofs to:
Dr. S.J. Doran
Department of Physics,
University of Surrey,
Guildford, GU2 5XH,
England
Tel. :
Fax:
e-mail:
+44 1483 259413
+44 1483 259501
S.Doran@surrey.ac.uk
1
Abstract
The STEAM-Burst sequence is a single-shot, multi-slice imaging technique which
does not involve rapid gradient switching. A Burst excitation pulse train is followed by
a 90 hard pulse and, after a mixing time, by a 90 slice-selective pulse. A read
gradient refocuses a set of stimulated echoes, which can be phase-encoded to form
an image. By repeating the selective pulse N times, each time with the carrier
frequency offset differently, it is possible to sample N slices in a single-shot.
A comparison is made of the sequence with other 3-D single-shot methods.
Experiments implementing the technique on a 3 T whole-body imaging system and a
2 T, 31 cm bore animal imager are described. Both phantom and brain images are
presented. The principal advantages of the new sequence are its speed, the
absence of rapid gradient switching and corresponding freedom from artifacts, its
insensitivity to static magnetic field inhomogeneities and its low acoustic noise. The
main disadvantages are the low signal-to-noise ratio of the images produced and the
concomitant limitation in resolution.
Key words: 3D imaging, ultra-fast imaging, Burst, STEAM.
2
Introduction
As MRI tackles ever more challenging applications, the role of ultra-rapid imaging
sequences is becoming more prominent. The undoubted success of fast single-slice
techniques has turned attention towards the development of new 3-D and multi-slice
sequences, in order to meet the demands of imaging extended 3-D regions of the
body in times as short as a few hundred milliseconds.
Ultra-rapid multi-slice imaging will be a key development in the study of the thorax
and abdomen, where cardiac and respiratory motion cause artifacts in the images
acquired with conventional sequences. Both the diagnosis of heart disease and the
study of the gastro-intestinal system will benefit. Another area of research
demanding fast multi-slice sequences is the study of the perfusion of pathological
tissues (for example, in the brain). To perform dynamic-contrast bolus-tracking
experiments, it is necessary to match one’s imaging speed to the transit times,
typically a few seconds, which characterise the perfusion of contrast agent into the
tissues. Finally, functional MRI is an immense domain of investigation. The activated
areas of the brain do not necessarily lie in the 2-D planes which are examined by
today’s ultra-rapid sequences and a 3-D, single-shot, multi-slice technique would
prove invaluable.
Burst (1) is the generic name for a class of pulse sequences involving the
application of multiple r.f. pulses under a constant gradient and the subsequent
refocusing of a set of echoes. The possibility of using Burst to obtain ultra-fast
images without the need for rapidly switched gradients is a very exciting one. Whilst
Burst sequences are well known to produce images with poor signal-to-noise, they
have a number of advantages over the standard fast imaging methods such as EPI
3
(2), Turbo-FLASH (3), RARE (4) and GRASE (5): they are free from artifacts due to
inhomogeneity of the main magnetic field B0; they do not involve a high r.f. power
deposition; they can be made acoustically very quiet.
Previous ultra-fast 3-D imaging sequences
We define as “single-shot” techniques only those which are able to acquire the
whole set of data “in one repetition time”, without the need to wait for magnetisation
recovery by T1 relaxation between the sampling of successive lines or planes of kspace.
For 3-D imaging, only three pulse sequences satisfying this criterion have been
found in the literature:
(i) Echo-Volumar Imaging (EVI) (6) is an EPI-based experiment, which has recently
been applied to the study of gastric filling and emptying (7). Single-shot abdominal
images have been obtained (64328 voxels in 102 ms, with a resolution per plane
of 6410 mm3). However, the large number of gradient switches involved in this
technique make it suitable only for specially designed systems with resonant or ultrashort ramp-time gradients.
(ii) URGEVI is a recently introduced hybrid of the URGE (Ultra Rapid Gradient Echo)
variant of Burst (8) with switched, EVI-like gradients. Heid describes, in abstract form
(9), the acquisition of 16 partial Fourier images of 5664 in 132 ms, with a resolution
of 5.75 mm2. As in EVI, the images produced are T2* weighted and the method is
suitable for bolus-tracking, abdominal scanning and kinematic joint movement.
4
(iii) Frequency-shifted Burst (10) combines a gradient-refocused Burst experiment
with a standard 3-D phase-encoding. Whilst the acquisition speed is an order of
magnitude slower than the above methods (484848 voxels in 3.1 s, with a
resolution of 4.54.54.7 mm3) and whilst the sequence is repeated for each phaseencode step in the third dimension, the frequency shifting principle means that no T1
relaxation delay is required between repetitions. This technique, too, has been used
successfully for bolus-tracking experiments (11).
Against this background, the STEAM-Burst (12) imaging sequence is a promising,
alternative technique, giving single-shot, multi-slice images. It is the combination of
the STimulated Echo Acquisition Mode (13) and the Burst excitation pulse train. Here,
we present phantom and human brain images to demonstrate that this new sequence
is well suited for ultra-fast multi-slice acquisition. Results are shown from both a highfield whole-body scanner and a conventional small-bore animal imaging system.
5
Theory
The sequence on which STEAM-Burst is based is the DUFIS (DANTE Ultra Fast
Imaging Sequence) variant of Burst (14). A DANTE (15) excitation pulse train of n
low flip angle -pulses is applied in the presence of a constant gradient in the read
direction. We modify DUFIS as shown in Fig. 1 by replacing the slice-selective 180
refocusing pulse with a non-selective 90. N slice-selective 90 pulses are then used
to read out N separate echo trains, with each train of n stimulated echoes
corresponding to one image. We have also changed the location of the phaseencoding gradient; by making it coincide with the excitation instead of with the
repeated acquisition, we reduce the number of gradient switches. In addition, kspace is now sampled parallel to kread and not at an angle, as in DUFIS.
In Fig. 1 the time between the last pulse and the centre of the 90° hard pulse is
indicated by T0. The magnetisation stays along the longitudinal axis for the mixing
time, TMj, until a 90° soft pulse selects a slice and tips its magnetisation back into the
transverse plane. The length of TMj is different for each slice, j. During this delay, a
gradient along the slice direction dephases any residual transverse magnetisation.
The 90° soft pulse is applied together with a slice-selective gradient followed
immediately by a slice-refocusing lobe. (This change from the original STEAM-Burst
proposal (12), in which the slice-refocussing lobe was placed after the excitation,
reduces the signal attenuation due to diffusion, but it increses the number of gradient
switches.) The readout gradient refocuses one stimulated echo from each excitation
pulse. Sequence timings are arranged such that the time between the (n-i+1)th pulse and the centre of the 90° hard pulse is equal to the time from the centre of the
selective 90° pulse to the i th echo. Each echo corresponds to a different line of the kspace. The dotted box of Fig. 1 represents the multi-slice loop which is repeated N
6
times to sample N images. Each time the program goes through this loop, the
transmit frequency is shifted according to the position of the slice to be acquired.
In order to avoid disruptive interference of higher order echoes with the primary
echoes,  must be small. This leads to a low S/N ratio, which is particularly significant
when acquiring stimulated echoes which give only half the signal of the corresponding
DUFIS spin echoes. This problem can be reduced by phase-modulating the Burst
pulses, as suggested by previous authors (16, 17) and, here, we have used Zha and
Lowe’s “two-phase OUFIS” modulation scheme.
The signal obtained is affected by decay processes. The major problem arises
from the diffusion of spins in the presence of the strong excitation and readout
gradients (18). In addition, T1 and T2 relaxation affect the signal decay with
consequent influence on the weighting of each image. The amplitude Aij of the i
th
echo in the j th slice is reduced by these decays as described in the following form:
 2T 0  i  
 TM j 
 

2

Aij  exp D G read   i2   ij  i   exp 
exp 
,


3 
T2
 T1 



[1]
where the D is the self-diffusion coefficient, Gread the amplitude of the excitation and
readout gradients,  ij and  ij the Stejskal-Tanner (19) parameters as applied to the
STEAM-Burst experiment (see Fig. 1) and  the distance between the centres of two
successive excitation pulses or echoes. T0 is determined by the echo time for the first
echo, chosen by the user.  i depends on the echo number, i, and TMj depends on
the slice number, j, while  ij depends both on the echo position and on the slice
number as follows:
7
TM j  TM1  ( j  1)Ts
i 
 1
  i  
2
 2
t ramp
[2]
 ij  2T0  TM j   i   1
.
TM0 is the mixing chosen by the user for the first slice, tramp is the gradient ramp time
and Ts is the slice repetition time, indicated in Fig. 1, which is the interval between any
two corresponding points of two successive slice acquisitions.
Equation [1]
represents the application of Tanner’s standard formula (20) to the case of Burst.
Equation 1 is stated for a uniform sample, for which D, T1 and T2 do not vary with
position. However, for a more complex sample, we may simply regard Aij as being the
contribution to the echo amplitude from an isochromat at position (x,y,z). The
resultant amplitude is found by integrating Eq. [1] over the slice considered.
All the images have the same T2 weighting, but a different diffusion and T1
weighting . Moreover, each of the echoes forming an image has a different diffusion
and T2 attenuation, but the same T1 attenuation. This gives rise to an image contrast
and blurring which is a complicated function of T1, T2 and D, particularly where these
parameters are spatially dependent.
The acquisition bandwidth, BW, has a role in determining the decay behavior of the
signal, since it determines the values of Gread,  i and  ij and it is closely related to the
final imaging speed. Figure 2 plots the decay of the signal amplitude, Aij, described by
Eq. [1], as a function of the echo number; the joined lines correspond to the
theoretical curves of slices 1, 3 and 5, determined using an acquisition bandwidth of
200 kHz (sample period = 5 s per complex point), while the dotted curves are the
8
corresponding decays for a bandwidth of 100 kHz (sample period = 10 s per
complex point). Note the gain in using the higher BW, though against this must be set
the fact that S/N  1/ BW.
9
Method and Results
The STEAM-Burst imaging sequence has been implemented on two different MRI
scanners: a whole-body 3 T system (Surrey Medical Imaging Systems (SMIS),
Guildford, England) and a 2 T, small-animal imager (Oxford Magnet, SMIS console).
Hardware and experimental details have been summarised in Table 1.
The predictions of Eq. [1] were tested by acquiring STEAM-Burst data without
phase-encoding. In Fig. 3, we have plotted the normalised values of the signal
amplitude from slices 1 () and 5 () out of a set of 6, together with the corresponding
decays (solid lines), as predicted from Eq. [1]. The data were acquired on a water
phantom, using the 3 T system with the phase encode gradient turned off. The 64
echoes were Fourier transformed in the read direction and, then, for a given point on
the resulting profiles, Eq. [1] was fitted to the decay data using the LevenbergMarquardt least-squares algorithm. For the curves shown, we obtained values of D =
1.99×10-5 cm2 s-1 () and D = 2.16×10-5 cm2 s-1 (). It is important to note that the
head gradient set used for these measurements gives rise to a field gradient whose
magnitude varies by up to 5% over the imaging volume. Calculation of D using a
nominal value of G can lead to errors. In addition, the equipment does not currently
include facilities for monitoring or stabilising the phantom temperature. For these
reasons we did not attempt a rigorous comparison with literature values of diffusion
coefficient. We are, as yet, unclear why the data for the first slice () appear noisier
then for the subsequent ones () and why the earlier echoes appear noisier than
subsequent ones.
In all the imaging studies presented below, image acquisition was truly single-shot;
no averaging was performed. Figure 4 shows a set of six full k-space 64  64 STEAM-
10
Burst images of the water phantom used to test the sequence on the 3 T system,
whilst Fig. 5 shows the results of using the same sequence on a human head. In both
cases, all six images were acquired in a total time 180 ms (i.e., 20 ms per image, with
the remaining time being taken by the various slice-selective pulses, gradient ramps,
etc.). The image resolution is 4  4 mm2 for the phantom images and 5  5 mm2 for
the head images. Figure 6 shows six half k-space images of a lemon, on the 2 T
instrument; the nominal resolution of these images is 1.25  1.6 mm2. All the other
imaging parameters are given in Table 1.
Discussion
We have described the implementation of the STEAM-Burst sequence on two
different MR scanners, demonstrating its performance as a single-shot multi-slice
technique.
Good, distortion-free and artefact-free images are obtained both on a water
phantom, where diffusion is heavily present, and on the human brain, where diffusion
is reduced, but T2 decay is increased. The images on the 2 T animal system
demonstrate that the STEAM-Burst sequence can be used for higher resolution
studies, but they highlight a number of problems, which will be discussed further
below.
All the images presented here show a good definition of the samples consistent
with the nominal pixel resolutions given in Table 1. White and gray matter are
distinguishable in the human brain images, where it is possible to recognise internal
structures. The only major image artifact is a vertical line, in the head data, arising
from local radio frequency interference, which will, in future, be screened out.
11
The STEAM-Burst sequence has a number of attractive features:
(i) Speed — We acquire multi-slice data very rapidly (66464 full k-space head
images in an acquisition period of 180 ms — 210 ms including excitation period).
(ii) Gradient Performance — As a general rule, the higher the data sampling rate,
the stronger will be the gradient needed. This is true for most fast imaging
sequences and Burst is no exception. Here, γGread = BW / FOVread and so for our
ultra-fast experiments ( 20 ms / image) we do need strong gradients (14.6 mT m-1
for our head images). However, the fact that the gradient does not need to be
switched rapidly means that Burst can be implemented on systems without
specialised “EPI” gradient hardware and that the images do not suffer from artifacts
due to eddy currents. Notice in Table 1 that the head images were acquired with a
gradient ramp time of 1 ms — i.e. only 16 T m-1 s-1.
(iii) Inhomogeneities in B0 — Since each echo is independently r.f. refocussed
(i.e., a true stimulated echo), images do not exhibit distortion or signal “drop-out” in
regions of inhomogeneous B0. STEAM-Burst is particularly robust in regions of poor
shimming and/or rapid changes in susceptibility across the sample. In Fig. 5, we
obtain good images from low down in the brain, where EPI images would show
considerable signal loss around the air spaces in the sinuses. Notice that this
advantage would render the implementation of STEAM-Burst described here
unsuitable for functional imaging, which uses T2* contrast. However, by adding a
negative gradient lobe in the read-direction immediately after the acquisition, one
can obtain a T2*-weighted image in addition to the standard T2-weighted one, albeit
with a further reduced S/N.
12
(iv) Acoustic Noise — Preliminary data indicate that Burst can be made very quiet.
In certain frequency bands, we have observed Burst to be 15 dB quieter than an
equivalent EPI sequence.
However, two major problems exist with Burst sequences:
(i) Low Signal-to-Noise — To overcome the low S/N typical of all Burst
experiments, we find that a phase-modulation scheme for the -pulses is essential.
With the implementation of Zha and Lowe’s optimised two phase OUFIS scheme
(17), our 3T head images have a S/N which varies between 15:1 to 20:1, whilst the
lemon images at 2 T have S/N of approximately 6:1.
(ii) Signal Decay — Equation [1] shows that D, T1 and T2 are all potentially important
factors governing the viability of the technique. Due to the acquisition of stimulated
echoes, T2* does not contribute to the observed signal decay. Table 2 shows the
signal attenuation caused by each of these three factors, calculated with Eq. [1], on
the middle echo of successive slices for brain tissue and water respectively, assuming
an ideal experiment. In the brain, T2 is the most significant factor and limits the
maximum length of the echo train to approximately 80 ms; by contrast, T1 does not
have a major effect over the images, due to its long value (more than 1.5 s at 3 T)
compared with the total acquisition time.
13
Limits to resolution in STEAM-Burst
Ultimately, the most severe limit on resolution is low signal-to-noise. Without
taking any other complicating factors into account, simply reducing the voxel size
from 5510 mm3 as in the current head images to 2.52.55 mm3, would entail a
reduction in S/N by a factor of 8 (i.e., images with S/N  2). In addition, we must
consider the signal decay and the need to add more r.f. pulses to the Burst train.
The resolution of Burst images is limited both along the read and the phaseencode direction. To achieve a better resolution along kread, a higher number of
sample points per echo should be acquired, while to improve the resolution along
kphase it would be necessary to increase the number of excitation pulses. In both the
cases, according to Eq. [1], this would mean longer acquisition times, with a more
evident signal decay within each slice and, also, a quicker degradation of the signal
as a function of the slice number. (Increasing the number of pulses also has the
undesirable consequence that one must reduce the elementary flip angle , which will
further reduce the signal-to-noise of the image.)
The signal decay due to diffusion is particularly significant for small-bore systems
(e.g., in animal imaging), because of the larger gradient values used, and may well
prove to be a fundamental limit to the resolution attainable using the technique. The
2 T lemon images were acquired using 25 r.f. pulses to shorten the sequence and so
lessen the signal decay, increasing S/N by using a larger flip angle . To overcome
the resulting problem of a poor resolution along the phase encode direction it is
possible to use asymmetric k-space scanning strategies. The lemon images
demonstrate this; a matrix of 64  50 has been reconstructed from the 25 echoes of a
half k-space acquisition.
14
The number of pulses for the 3T experiments (Fig. 4 and 5) has been set at 64 and
k-space has been sampled symmetrically throughout, so the images have a matrix
size of 6464. An asymmetric sampling could increase the size to 64128. At 3 T it
should be possible to extend the number of sample points per echo to 96, without
paying too much in S/N and signal decay, so that our best matrix size would be about
96 128. Experiments to achieve this are still in progress.
The signal decay process limits the number of slices which can be acquired. At 3 T
it is possible to scan 8–10 slices before late echoes start to be unacceptably
attenuated in the final slice. Unfortunately, there is little that can be done to overcome
this problem, apart from trying to optimise all the timings to make the sequence as
short as possible. In practice, the large slice width and the finite active volume of the
head coil can also limit the number of slices.
Conclusions
In this study, we propose a novel method combining stimulated echo acquisition
and BURST excitation. This new sequence has been applied to ultra-fast multi-slice
acquisition and has been demonstrated on phantom as well as in vivo experiments.
Multi-slice STEAM-Burst represents one approach to optimise volume acquisition in
a single-shot with BURST-type sequences. It has proved to be suitable for ultra-fast
imaging on small bore as well as on whole-body systems.
The technique has two main limitations: low signal-to-noise and signal decay due to
T2 and diffusion. At present it seems that 8–10 is the maximum number of slices
which may be acquired at 3T whilst maintaining adequate S/N in the later images.
Image weighting can be introduced with preparation pulses, prior to the STEAMBurst experiment. The diffusion decay, which is a problem for this method, can be
15
exploited in experiments to measure self-diffusion coefficients, as previous work has
demonstrated (21).
The STEAM-Burst sequence offers the possibility for a wide range of experiments
which are still under investigation. Preliminary results (not shown) indicate that by
using an acquisition bandwidth of 400 kHz, we can reduce the time for 6 images of
6464 pixels to 118 ms.
References
1.
J. Hennig, M. Hodapp, Burst imaging. MAGMA 1, 39-48 (1993).
2.
P. Mansfield, Multi-planar image formation using NMR spin echoes. J. Phys.
C: Solid State 10, L55-L58 (1977).
3.
A. Haase, J. Frahm, D. Matthaei, W. Hänicke, K.-D. Merboldt, FLASH
imaging . Rapid NMR imaging using low flip-angle pulses. J. Magn. Reson.
67, 258-266 (1986).
4.
J. Hennig, A. Nauerth, H. Freidburg, RARE imaging: a fast imaging method
for clinical MR. Magn. Reson. Med. 3, 823-833 (1986).
5.
K. Oshio, D. A. Feinberg, GRASE (gradient- and spin-echo) imaging: a novel
fast MRI technique. Magn. Reson. Med. 20, 344-349 (1991).
6.
P. Mansfield, P. R. Harvey, M. K. Stehling, Echo Volumar Imaging. MAGMA
2, 291-294 (1994).
7.
P.R. Harvey, P. Mansfield, Echo-Volumar Imaging (EVI) at 0.5 T: First wholebody volunteer studies. Magn. Reson. Med. 35, 80-88 (1996).
8.
O. Heid, M. Deimling, W. J. Huk, Ultra-Rapid Gradient Echo Imaging.
Magn. Reson. Med. 33, 143-149 (1995).
16
9.
O. Heid, Ultra Rapid Gradient Echo Planar/Volumar Imaging, in “Proc.,
SMRM, 3rd Annual Meeting, Nice, 1995,” p. 98.
10.
J.H. Duyn, P. van Gelderen, G. Liu, C.T.W. Moonen, Fast volume scanning
with frequency-shifted Burst MRI. Magn. Reson. Med. 32, 429-432 (1994).
11.
J.H. Duyn, P. van Gelderen, P. Barker, J.A. Frank, V.S. Mattay, C.T.W.
Moonen, 3D bolus tracking with frequency-shifted BURST MRI. J. Comput.
Asst. Tomog. 18 (5), 680 (1994).
12.
Y. Crémillieux, B. Sola, O. Beuf, A. Briguet, STEAM-Burst: Application to
sub-second multi-slice acquisition, in “Proc., SMRM, 4th Annual Meeting,
New York, 1996,” p. 1500.
13.
J. Frahm, K. Merboldt, W. Hänicke, A. Haase, Stimulated echo imaging, J.
Magn. Reson., 64, 81–93 (1985)
14.
I. J. Lowe, R. E. Wysong, DANTE Ultrafast Imaging Sequence (DUFIS). J.
Magn. Reson. Ser. B 101, 106-109 (1993).
15.
G.A. Morris, R. Freeman, Selective excitation in Fourier transform nuclear
magnetic resonance. J. Magn. Reson. 29, 433-462 (1978).
16.
P. van Gelderen, J.H. Duyn, C.T.W. Moonen, Analytical solution for
phase modulation in Burst imaging with optimum sensitivity. J. Magn.
Reson. Ser. B 107, 78-82 (1995 ).
17.
L. Zha, I.J. Lowe, Optimized Ultra-Fast Imaging Sequence (OUFIS).
Magn. Reson. Med. 33, 377-395 (1995).
18.
S.J. Doran, M. Décorps, A robust, single-shot method for measuring
diffusion coefficients using the “Burst” sequence. J. Magn. Reson. Ser. A
117, 311-316 (1995).
17
19.
E.O. Stejskal, J.E. Tanner, Spin diffusion measurements: Spin echoes in the
presence of a time-dependent field gradient. J. Chem. Phys. 42 (1), 288-292
(1965).
20.
J. E. Tanner, Use of the stimulated echo in NMR diffusion studies. J. Chem.
Phys. 52 (5), 2523-2526 (1970)
21.
K.-D. Merboldt, W. Hänicke, J. Frahm, Diffusion imaging using stimulated
echoes. Magn. Reson. Med. 19, 233-239 (1991).
18
T0
T0
TMj
90°
90°
RF and
Signal
n -pulses
n STE
Gslice
i
ij
Gread
Gphase
Ts
N slices
Fig. 1. The STEAM-Burst sequence diagram. Multi-slice acquisition is based on the
repetition of the right part of the sequence (rectangular box). The timing parameters
are explained further in the text with the corresponding values summarised in Table 1.
19
Fig. 2. Theoretical signal decay in a single-shot multi-slice STEAM-Burst
experiment. The calculated attenuation of the signal takes into account diffusion, T1
and T2 decays as predicted by Eq. [1], assuming the parameters of the experiment
at 3T. Dotted line 100 kHz acquisition bandwidth; solid line 200 kHz acquisition
bandwidth. The curves are numbered according to the slice they refer to. All curves
are normalised to the amplitude of echo 1 of slice 1 with a 200 kHz bandwidth.
20
Fig. 3. Plot showing the signal decay versus echo number in a STEAM-Burst
experiment. Data were acquired without phase encoding and each echo was Fourier
transformed separately. The “amplitude” on the y-axis is the magnitude of a given
point on the resulting profile. Similar results would be obtained if the magnitude of
the corresponding echo were used.  = data from the slice 1,  = data from the slice
5. The solid curves represent the fit of Eq. [1] to the data. Both curves are
normalised to the theoretical (fit) value at echo 1 for slice 1. The reduced initial value
for echo 1 of slice 5 is approximately that predicted by Eq. [1]. An exact comparison
is not possible because of the (slight) spatial inhomogeneity of the phantom, which
meant that slice 5 contained a different amount of material from slice 1.
21
Fig. 4. Single-shot, multi-slice STEAM-Burst images of a water phantom at 3T. The
sample diameter is 200 mm and the field of view 250 mm. 6 slices of 5 mm slice
thickness have been acquired in a total acquisition time of 180 ms. Other parameters
are as in Table 1. The original 64  64 images have been interpolated to 256  256.
22
Fig. 5. Single-shot, multi-slice STEAM-Burst images of a human brain at 3T. 6
slices of 10 mm slice thickness have been acquired in a total acquisition time of 180
ms. Other parameters are as in Table 1. The original 64  64 images have been
interpolated to 256  256.
23
Fig. 6. Single-shot, multi-slice STEAM-Burst images of a lemon at 2 T. 6 slices of 5
mm slice thickness have been acquired in a total acquisition time of 267 ms. Other
parameters are as in Table 1. The original 64  50 images have been interpolated to
128 128.
24
Table1. Hardware details and experimental parameters
System 1
System 2ystem 2
Hardware
80-cm, whole-body, 3T
(127 MHz) SMIS imager
30 cm horizontal bore, 2 T
(85.1 MHz) magnet, SMIS console
Gradient strength (Gread)
14.6 mT m-1 (human brain)
12 mT m-1 (phantom images)
15 mT m-1
Ramp time (tramp)
1 ms
1 ms
No of Burst pulses
(= number of echoes)
64
25
(half k-space)
Burst pulse length
10 s (phantom images)
17 s (human brain)
27 s
Phase scheme
two-phase OUFIS
two-phase OUFIS
Flip Angle ()
8
15
Inter-pulse spacing (
320 s
1280 s
Total excitation time
= acq. time per image
20.5 ms
32 ms
T0, TM1, Ts
7.7 ms, 3 ms, 32 ms
7.6 ms, 4.1 ms, 44 ms
Acquisition bandwidth
200 kHz
50 kHz
Sample points / echo
64
64
Processed image matrix Size 64646
64506
Nominal Image Resolution
44 mm2 (phantom images)
55 mm2 (head images)
1.251.6 mm2
Slice thickness
5 mm (phantom images)
10 mm (head images)
5 mm
Number of slices
6
6
25
Total acquisition time
180 ms (210 ms including
excitation period)
267 ms (311 including
excitation period)
Number of averages
1
1
Signal to noise ratio
19:1 to 24:1 (phantom images) 6:1
15:1 to 20:1 (head images)
26
Table 2. Attenuation of the Burst signal caused by T1, T2 and diffusion. Results are
presented in terms of the amplitude of the ith echo compared with the amplitude of
echo 1 (i.e., the echo attenuation). The intra-slice attenuation is given for echo 32
during acquisition of slice 1 to assess the signal at the point corresponding to the
“zero phase-encode step” and for echo 64 of slice 1 to show the amplitude at the
end of the echo train. The inter-slice attenuation is given between echoes 1 of slice
1 and slice 5 to illustrate the reduction in S/N of successive images. The separate
evaluation of the contributions of T1, T2 and diffusion decay is based on Eq. [1],
using the data values D = 2.2  10-5 cm2 s-1, T1 = 1.5 s, T2 = 1.3 s for water and D =
0.85  10-5 cm2 s-1, T1 = 1.35 s, T2 = 0.1 s for brain tissue.
Water Phantom
Decay
process
Brain Tissue
Attenuation
Echo 32
Attenuation
Echo 64
Inter-slice
attenuation
Attenuation Attenuation
Echo 32
Echo 64
Inter-slice
attenuation
T1
1
1
0.91
1
1
0.90
T2
0.98
0.97
1
0.81
0.66
1
D
0.94
0.72
0.73
0.97
0.88
0.88
Overall
0.92
0.70
0.66
0.79
0.58
0.79
27
Figure and Table Legends
Fig. 1. The STEAM-Burst sequence diagram. Multi-slice acquisition is based on the
repetition of the right part of the sequence (rectangular box). The timing parameters
are explained further in the text with the corresponding values summarised in Table 1.
Fig. 2. Theoretical signal decay in a single-shot multi-slice STEAM-Burst
experiment. The calculated attenuation of the signal takes into account diffusion, T1
and T2 decays as predicted by Eq. [1], assuming the parameters of the experiment
at 3T. Dotted line 100 kHz acquisition bandwidth; solid line 200 kHz acquisition
bandwidth. The curves are numbered according to the slice they refer to. All curves
are normalised to the amplitude of echo 1 of slice 1 with a 200 kHz bandwidth.
Fig. 3. Plot showing the signal decay versus echo number in a STEAM-Burst
experiment. Data were acquired without phase encoding and each echo was Fourier
transformed separately. The “amplitude” on the y-axis is the magnitude of a given
point on the resulting profile. Similar results would be obtained if the magnitude of
the corresponding echo were used.  = data from the slice 1,  = data from the slice
5. The solid curves represent the fit of Eq. [1] to the data. Both curves are
normalised to the theoretical (fit) value at echo 1 for slice 1. The reduced initial value
for echo 1 of slice 5 is approximately that predicted by Eq. [1]. An exact comparison
is not possible because of the (slight) spatial inhomogeneity of the phantom, which
meant that slice 5 contained a different amount of material from slice 1.
Fig. 4. Single-shot, multi-slice STEAM-Burst images of a water phantom at 3T. The
sample diameter is 200 mm and the field of view 250 mm. 6 slices of 5 mm slice
thickness have been acquired in a total acquisition time of 180 ms. Other parameters
are as in Table 1. The original 64  64 images have been interpolated to 256  256.
28
Fig. 5. Single-shot, multi-slice STEAM-Burst images of a human brain at 3T. 6
slices of 10 mm slice thickness have been acquired in a total acquisition time of 180
ms. Other parameters are as in Table 1. The original 64  64 images have been
interpolated to 256  256.
Fig. 6. Single-shot, multi-slice STEAM-Burst images of a lemon at 2 T. 6 slices of 5
mm slice thickness have been acquired in a total acquisition time of 267 ms. Other
parameters are as in Table 1. The original 64  50 images have been interpolated to
128 128.
Table1. Hardware details and experimental parameters
Table 2. Attenuation of the Burst signal caused by T1, T2 and diffusion. Results are
presented in terms of the amplitude of the ith echo compared with the amplitude of
echo 1 (i.e., the echo attenuation). The intra-slice attenuation is given for echo 32
during acquisition of slice 1 to assess the signal at the point corresponding to the
“zero phase-encode step” and for echo 64 of slice 1 to show the amplitude at the
end of the echo train. The inter-slice attenuation is given between echoes 1 of slice
1 and slice 5 to illustrate the reduction in S/N of successive images. The separate
evaluation of the contributions of T1, T2 and diffusion decay is based on Eq. [1],
using the data values D = 2.2  10-5 cm2 s-1, T1 = 1.5 s, T2 = 1.3 s for water and D =
0.85  10-5 cm2 s-1, T1 = 1.35 s, T2 = 0.1 s for brain tissue.
29
List of symbols
TM0
italic , zero subscript
Tmj
italic , j subscript
N
italic , caps
T1
italic , 1 subscript
T2
italic , 2 subscript
T2*
italic , 2 subscript, * superscript
n
italic , small caps
 
greek alpha
kread
italic
kphase
italic
T0
italic , zero subscript
Ts
italic , s subscript
Aij
italic , ij subscript
ith
italic , th superscript
jth
italic , th superscript
D
italic, caps
Gread
italic , read subscript
i
italic , i subscript, greek delta, small caps
ij
italic , ij subscript, greek delta, caps

italic , greek tau

BW


italic , greek gamma
italic , caps
30