Technical ReportPDF Available

Improving Sensitivity and Specificity in BOLD fMRI Using Simultaneous Multi-Slice Acquisition

Authors:

Abstract and Figures

Functional MRI techniques, which involve rapid serial imaging of the brain to detect activation-induced changes, have always placed high demands on the speed, precision, and stability of MRI systems. This is particularly true of studies requiring high spatial resolution, due to the dramatic reduction in signal amplitude with decreasing voxel volume. Recent developments in simultaneous multi-slice (SMS) encoding promise to have a major impact on functional MRI. Current trends in MRI system hardware will help maximize this impact, and expand the range of fMRI applications that are feasible in clinical practice and basic research. In this article, the authors discuss the advantages of highly accelerated fMRI and show example images from a visual activation paradigm. The future benefits of this technology include the ability to perform pre-surgical mapping with high reliability and detail, with clinically feasible exam times.
No caption available
… 
No caption available
… 
Content may be subject to copyright.
Improving Sensitivity and
Specificity in BOLD fMRI Using
Simultaneous Multi-Slice Acquisition
Richard D. Hoge; AmanPreet Badhwar; Julien Doyon; David Ostry,
McConnell Brain Imaging Centre, Montréal Neurological Institute, Deptartment of Neurology & Neurosurgery,
McGill University, Montreal, Quebec, Canada
Unité de Neuroimagerie Fonctionnelle, Institut Universitaire de Gériatrie de Montréal, Université de Montréal, Montreal,
Quebec, Canada
Deptartment of Psychology, McGill University, Montreal, Quebec, Canada
Haskins Laboratories, New Haven, CT, USA
Abstract
Functional MRI techniques, which
involve rapid serial imaging of the
brain to detect activation-induced
changes, have always placed high
demands on the speed, precision,
and stability of MRI systems. This is
particularly true of studies requiring
high spatial resolution, due to the
dramatic reduction in signal amplitude
with decreasing voxel volume. Recent
developments in simultaneous multi-
slice (SMS) encoding promise to have
a major impact on functional MRI.
Current trends in MRI system hardware
will help maximize this impact, and
expand the range of fMRI applications
that are feasible in clinical practice
and basic research. In this article,
the authors discuss the advantages
of highly accelerated fMRI and show
example images from a visual activa-
tion paradigm. The future benefits
of this technology include the ability
to perform pre-surgical mapping
with high reliability and detail,
with clinically feasible exam times.
Introduction
Conventional echo-planar imaging
(EPI), which has been widely adopted
for functional imaging of the brain
over the last decade, has typically
acquired multiple 2D slices in a rapid
sequence whose minimum duration is
limited by the echo-time (TE) required
for sensitivity to blood oxygenation,
the EPI readout length, and the
number of slices required. This has
generally meant that, despite
advances in parallel imaging that
can shorten single-slice readouts,
the minimum time needed to image
the entire brain with typical slice
thicknesses has been on the order of
two seconds or more. Simultaneous
multi-slice imaging removes this limi-
tation, reducing the number of TE
delays required to image the entire
brain while preserving the necessary
T2*-weighting.
To understand the benefits of slice-
accelerated fMRI, it is helpful to
recall that sensitivity in functional
MRI depends ultimately on the ratio
between a given functional effect
size and the degree of unrelated
measurement variability. This
measurement variability, or noise,
arises from three main sources:
1) so-called ‘thermal’ noise;
2) physiological fluctuations in the
subject; and 3) instrumental instabil-
ity. Thermal noise is physically
unavoidable in electronic measure-
ments conducted above absolute
zero, but as uncorrelated Gaussian
noise its impact can be systematically
reduced through signal averaging or
increasing the size of image voxels.
Physiological fluctuations in the sub-
ject, such as cardiac or respiratory
cycles, are evidently difficult to elimi-
nate but it is possible to control for
their effects through physiological
noise modelling methods [1]. Instru-
mental instability can contribute low
level signal fluctuations that, while
not necessarily dominant in standard
BOLD fMRI acquisitions, can limit the
Stimulation paradigm with high and low-intensity visual stimulation.
All condition blocks were 30 seconds long, with a total run duration of ten minutes.
1
1
Simultaneous Multi-Slice BOLD Clinical
MAGNETOM Flash | (63) 3/2015 | www.siemens.com/magnetom-world 65
SNR gains achievable by managing
the other noise sources. Optimizing
sensitivity in fMRI generally involves
identifying the dominant noise
source in a particular application
and reducing it until other types of
noise become significant.
Acquisition speed in fMRI
In typical BOLD fMRI experiments
(3 Tesla, ~3 mm resolution, 2-3 sec-
ond repetition time), physiological
noise is by far the dominant noise
source [2]. Because of this, it has
been noted that BOLD fMRI time
series commonly exhibit a high
degree of temporal autocorrelation
[3]. This would suggest that there
is little statistical power (and hence
sensitivity) to be gained through
increases in imaging rate, since faster
sampling of an autocorrelated signal
does not necessarily increase the
number of independent samples.
However, this reasoning applies
only in the case where autocorrelated
physiological noise is dominant.
Indeed, the popularity of 3T fMRI
at an isotropic resolution of 3 mm
is likely due to the fact that further
reductions in voxel size lead to a
relatively precipitous drop in SNR.
This drop is due to the third power
loss of signal with voxel volume,
which results in a shift to thermal
noise as the predominant source
of signal fluctuation. Conversely,
increases in voxel size above 3 mm
yield little improvement in signal
stability, because uncontrolled physi-
ological fluctuations are already
the dominant noise source. With this
in mind, it becomes apparent that
acquisitions in which uncorrelated
thermal noise is dominant may still
stand to benefit significantly from
increases in imaging rate. At 3T,
this is true for situations including
high spatial-resolution BOLD fMRI
(e.g. ≤2 mm) and arterial spin-
labeling, in which the labeling
signal is generally much smaller
than typical BOLD effect sizes.
Simultaneous multi-slice
acquisition
While imaging rates attainable in
fMRI applications have been fairly
constant for many years, a recent
resurgence in simultaneous multi-
slice (SMS) encoding techniques
has created exciting possibilities for
highly accelerated functional imag-
ing. The general concept was pro-
posed as early as 2001 [4] but has
become practical only more recently
with the emergence of techniques
for managing issues such as voxel
tilting and aliasing [5]. SMS encoding
allows substantial increases in
imaging rate compared with standard
sequential multi-slice methods.
It is based on the parallel imaging
approach [6], in which the spatial
information provided by an array
of localized RF coil elements is used
to reduce the number of 2D Fourier
samples required to generate an
unaliased image. While parallel
imaging has been in use for nearly
a decade to reduce the time needed
for standard multi-excitation images,
its impact on the single-excitation
readouts used for BOLD fMRI has
been limited due to the need for a
relatively long post-excitation echo-
time (TE) in each slice, in order to
achieve the T2*-weighting required
for sensitivity to blood oxygenation.
The advance enabled by SMS encod-
ing has been preservation of the
long TE needed for BOLD fMRI while
reducing the impact of this delay on
the total time needed to cover the
entire brain. The result is that whole-
brain acquisitions formerly requiring
up to three seconds can now be
performed in less than half a second.
This corresponds to a factor of six
or more increase in the number of
images acquired per unit time, which
can have a substantial impact on
statistical power particularly when
thermal noise is dominant. A well-
known application of this has been
the use of an 8x slice acceleration
factor to achieve 2 mm isotropic
resolution with excellent sensitivity
in the Human Connectome Project
(HCP) [7].
Methods
To better understand the perfor-
mance gains achievable using SMS
in task activation studies, the authors
have performed a systematic assess-
ment of BOLD fMRI sensitivity using
a range of acceleration factors and
functional effect sizes, in both thermal
and physiological noise-dominated
regimes. In these tests healthy human
subjects underwent a visual activation
study including blocks of intense
visual stimulation using a high lumi-
nance-contrast radial checkerboard
modulated in a temporal squarewave,
as well as much weaker stimulation
using low-contrast sinusoidal gratings
drifting slowly across the visual field
(Fig. 1). This allowed sensitivity and
specificity to be assessed under condi-
tions of both high and low-amplitude
effects. The acquisitions were repeated
with larger (3 mm) and smaller
(2 mm) isotropic voxels sizes, to create
conditions under which physiological
and thermal noise are respectively
dominant.
Experiments were conducted on a
3T Siemens scanner (MAGNETOM Trio,
A Tim System), running software
release syngo MR B17, and the
32-channel head coil. For comparison
against accelerated acquisitions, a
standard gradient-echo EPI/BOLD
sequence with TR 3 s, TE 30 ms, α 90°
was used as a reference. Resolutions
of 3 mm isotropic and 2 mm isotropic
were acquired, with respective matrix
sizes of 64 x 64 and 100 x 100. The
3 mm isotropic scans had 42 slices,
while the 2 mm scans had 30 slices
due to the longer readout required at
the higher resolution. In addition to
these standard scans, both resolutions
were also repeated with a slice
acceleration factor of 6x using a WIP
sequence. The number of time
points acquired over the ten minute
stimulation protocol was 200 for the
non-accelerated sequence, with
1,200 time points for the 6x SMS scan.
The TR values for the accelerated scan
was 0.5 s, and the flip angle was
adjusted to the Ernst angle assuming
the T1 of grey matter, corresponding
to a flip angle of 45°.
The product is still under development
and not commercially available yet.
Its future availability cannot be ensured.
Clinical Simultaneous Multi-Slice BOLD
66 MAGNETOM Flash | (63) 3/2015 | www.siemens.com/magnetom-world
All image time-series were processed
using in-house software used for
quantitative image analysis. Motion
correction was performed, followed by
spatial smoothing with 3D Gaussian
kernels with FWHM values of 6 mm
for the 3 mm scan and 4 mm for the
2 mm acquisition. Smoothing was per-
formed at these widths because they
were found to provide an optimal com-
promise between sensitivity, specific-
ity, and spatial resolution. Following
preprocessing, the time series data
were fit using a General Linear Model
(GLM) including separate regressors
for the high and low-amplitude stimu-
lation. T statistics were computed by
dividing estimated effect sizes by the
residual standard error assuming
uncorrelated residuals. The resultant
T values were converted to the nega-
tive logarithm of the p value, -log(p),
based on the nominal degrees of free-
dom in the time series. Neglecting
autocorrelations in the data can result
in exaggerated significance levels, but
the errors are modest in the standard
fMRI scans, particularly for the 2 mm
acquisitions. The importance of auto-
correlation may be further reduced
due to the attenuation of steady-state
magnetization at the shorter TR val-
ues. The maps of -log(p) are referred
to as significance maps in the
remainder of this article.
Results
Figure 2 shows unthresholded signifi-
cance maps for 3 mm scans overlaid
in false color on the grayscale EPI
scans for reference. Maps are shown
for standard and accelerated scans,
and for high and low-intensity stimu-
lation. The color legend has been
matched in all maps to allow compar-
ison of apparent significant levels,
although autocorrelated noise might
artificially boost the values in the
accelerated scans. Nonetheless,
the region of visual activation is
extremely well delineated in the
accelerated scans, for both levels
of stimulus intensity. Extents of
activation appeared less clearly
in the maps computed from non-
accelerated data, although this might
simply have reflected the dynamic
range chosen.
Because of the potential for exagger-
ation of significance levels in the
accelerated scans, and the clear
difference in dynamic range, we
also assessed activation maps with
the color legends adapted to the
respective maps. These alternatively
displayed maps (also for 3 mm scans)
are shown in Figure 3. By ‘zooming’
the dynamic range for the non-
accelerated maps, the location and
extent of active areas is indeed more
apparent. However, this also ampli-
fies the background noise (random
fluctuations in the significance
level following its distribution under
the null hypothesis) and it can be
seen that the sensitivity and spatial
specificity of the accelerated maps
remains superior.
Although the spatial specificity
afforded by slice acceleration in
the unthresholded maps shown in
Figures 2 and 3 is striking, fMRI has
customarily involved thresholding of
activation maps after correction of
significance values for the multiple
comparisons inherent in image data.
To assess the performance gains of
SMS acquisition in this setting, the
significance maps for 3 mm scans
were thresholded using the false dis-
covery rate (FDR) approach, with a
threshold of 0.001. The thresholded
maps are shown in Figure 4, which
reveals that the non-accelerated scan
failed to detect the weak activation
associated with the low-contrast
grating. It should be noted that
relaxing the threshold sufficiently to
Significance maps for high and low-intensity stimulation, with 1x and 6x acceleration at 3 mm isotropic resolution
(before smoothing). False colour significance maps are overlaid on greyscale EPI scans. Dynamic range of colour legend
is equalized for both acquisitions.
2
No acceleration: TR = 3 s 6x acceleration: TR = 0.5 s
high contrast checkerboard low contrast grating high contrast checkerboard low contrast grating
2
Simultaneous Multi-Slice BOLD Clinical
MAGNETOM Flash | (63) 3/2015 | www.siemens.com/magnetom-world 67
Significance maps for high and low-intensity stimulation from 3 mm scans, with colour legend adapted to dynamic range of
respective acquisitions.
3
reveal the patch of weak activation
in the low-contrast, non-accelerated
scan also resulted in the appearance
of substantial swaths of non-visual
artifact.
In addition to apparent improve-
ments in the sensitivity and specific-
ity of 3 mm scans, we also noted
significant improvements in the
higher resolution 2 mm scans.
Figure 5 shows significance maps
for the high-contrast stimulus,
acquired at 2 mm resolution with
and without acceleration. While the
occipital visual response is readily
detected in both cases, careful
inspection of the maps reveals
that the accelerated maps provide
considerably improved delineation
of cortical activation. In the maps
acquired with acceleration, much
of the cortical ribbon can be clearly
discerned in the activation pattern,
with robust specificity against
background fluctuations.
Discussion and conclusions
The results shown above demonstrate
that slice acceleration can substan-
tially improve the sensitivity, specific-
ity, and spatial detail of BOLD func-
tional MRI. Despite concerns that gains
might be limited due to autocorrelated
noise, the unthresholded maps sug-
gest that the apparent improvements
reflect more than simple boosting of
significance caused by inflated degrees
No acceleration: TR = 3 s 6x acceleration: TR = 0.5 s
high contrast checkerboard low contrast grating high contrast checkerboard low contrast grating
3
No acceleration: TR = 3 s 6x acceleration: TR = 0.5 s
high contrast checkerboard low contrast grating high contrast checkerboard low contrast grating
4
Significance maps for high and low-intensity stimulation during 3 mm scans, thresholded at FDR 0.001.
14
Clinical Simultaneous Multi-Slice BOLD
68 MAGNETOM Flash | (63) 3/2015 | www.siemens.com/magnetom-world
Contact
Richard Hoge
Room WB316
McConnell Brain Imaging Centre
Montreal Neurological Institute
3801 University Street
Montreal, Quebec H3A 2B4, Canada
Phone: +1 (514) 398-1929
Fax: +1 (514) 398-2975
rick.hoge@mcgill.ca
Significance maps high-intensity stimulation, based on 2 mm isotropic data.
The orange arrows indicate a section of cortical ribbon that can be readily
delineated in the 6x accelerated scan. Top row shows false color significance
maps overlaid on greyscale T1-weighted scan. Bottom row shows unthresholded
activation map alone.
5
of freedom. For clinical applications
in which averaging over repeated fMRI
runs is not feasible, these capabilities
offer significant advantages. An obvi-
ous example is pre-surgical mapping,
in which functional images of high
reliability must be acquired in a
relatively short time.
During the planning and optimization
of protocols, considerable effort was
devoted to finding suitable readout
characteristics such as bandwidth,
echo spacing, and matrix size.
Improvements in gradient technology
should increase the flexibility with
which SMS techniques can be applied,
while preserving and even improving
image quality. A second challenge
met during these experiments was
related to the long reconstruction
times required to generate images
from the SMS raw data. Here too,
improvements in scanner technology
to cope with the greatly increased
data throughput and complexity will
play an important role.
Acknowledgements
This work was supported by the
Canadian Foundation for Innovation
(Leaders Opportunity Fund 17380),
Canadian Institutes of Health
Research (MOP-273379), Natural
Sciences and Engineering Council of
Canada (355583-2010) and MITACS
(scholarship held by A.B.).
References
1 R. H. Tijssen, M. Jenkinson, J. C. Brooks,
P. Jezzard, and K. L. Miller. Optimizing
RetroICor and RetroKCor corrections for
multi-shot 3D FMRI acquisitions. Neuro-
image, 84:394–405, Jan 2014.
2 C. Triantafyllou, R. D. Hoge, G. Krueger,
C. J. Wiggins, A. Potthast, G. C. Wiggins,
and L. L. Wald. Comparison of physio-
logical noise at 1.5 T, 3 T and 7 T and
optimization of fMRI acquisition param-
eters. Neuroimage, 26(1):243–250, May
2005.
3 P. L. Purdon and R. M. Weisskoff. Effect
of temporal autocorrelation due to physi-
ological noise and stimulus paradigm on
voxel-level false-positive rates in fMRI.
Hum Brain Mapp, 6(4):239–249, 1998.
4 D. J. Larkman, J. V. Hajnal, A. H. Herlihy,
G. A. Coutts, I. R. Young, and G.
Ehnholm. Use of multicoil arrays for
separation of signal from multiple slices
simultaneously excited. J Magn Reson
Imaging, 13(2):313–317, Feb 2001.
5 K. Setsompop, B. A. Gagoski, J. R.
Polimeni, T. Witzel, V. J. Wedeen, and L.
L. Wald. Blipped- controlled aliasing in
parallel imaging for simultaneous
multislice echo planar imaging with
reduced g-factor penalty. Magn Reson
Med, 67(5):1210–1224, May 2012.
6 K. P. Pruessmann, M. Weiger, M. B. Schei-
degger, and P. Boesiger. SENSE: sensi-
tivity encod- ing for fast MRI. Magn
Reson Med, 42(5):952–962, Nov 1999.
7 S. M. Smith, C. F. Beckmann, J.
Andersson, E. J. Auerbach, J. Bijster-
bosch, G. Douaud, E. Duff, D. A.
Feinberg, L. Griffanti, M. P. Harms, M.
Kelly, T. Laumann, K. L. Miller, S. Moeller,
S. Petersen, J. Power, G. Salimi-Khorshidi,
A. Z. Snyder, A. T. Vu, M. W. Woolrich, J.
Xu, E. Yacoub, K. U?urbil, D. C. Van
Essen, and M. F. Glasser. Resting-state
fMRI in the Human Connectome Project.
Neuroimage, 80:144–168, Oct 2013.
5
1x 6x
Simultaneous Multi-Slice BOLD Clinical
MAGNETOM Flash | (63) 3/2015 | www.siemens.com/magnetom-world 69
ResearchGate has not been able to resolve any citations for this publication.
Article
New theoretical and practical concepts are presented for considerably enhancing the performance of magnetic resonance imaging (MRI) by means of arrays of multiple receiver coils. Sensitivity encoding (SENSE) is based on the fact that receiver sensitivity generally has an encoding effect complementary to Fourier preparation by linear field gradients. Thus, by using multiple receiver coils in parallel scan time in Fourier imaging can be considerably reduced. The problem of image reconstruction from sensitivity encoded data is formulated in a general fashion and solved for arbitrary coil configurations and k-space sampling patterns. Special attention is given to the currently most practical case, namely, sampling a common Cartesian grid with reduced density. For this case the feasibility of the proposed methods was verified both in vitro and in vivo. Scan time was reduced to one-half using a two-coil array in brain imaging. With an array of five coils double-oblique heart images were obtained in one-third of conventional scan time. Magn Reson Med 42:952-962, 1999.
Article
Resting-state functional magnetic resonance imaging (rfMRI) allows one to study functional connectivity in the brain by acquiring fMRI data while subjects lie inactive in the MRI scanner, and taking advantage of the fact that functionally related brain regions spontaneously co-activate. rfMRI is one of the two primary data modalities being acquired for the Human Connectome Project (the other being diffusion MRI). A key objective is to generate a detailed in vivo mapping of functional connectivity in a large cohort of healthy adults (over 1000 subjects), and to make these datasets freely available for use by the neuroimaging community. In each subject we acquire a total of 1h of whole-brain rfMRI data at 3T, with a spatial resolution of 2×2×2mm and a temporal resolution of 0.7s, capitalizing on recent developments in slice-accelerated echo-planar imaging. We will also scan a subset of the cohort at higher field strength and resolution. In this paper we outline the work behind, and rationale for, decisions taken regarding the rfMRI data acquisition protocol and pre-processing pipelines, and present some initial results showing data quality and example functional connectivity analyses.
Article
Increased acquisition efficiency has been achieved by exciting several slices simultaneously. The mixed data were unfolded to produce separate slices using the spatial encoding information inherent in a multicoil receiver system. Each coil yields a linear combination of signals from all excited slices weighted by the sensitivity of each coil. A matrix inversion provides a solution to unfold these images. J. Magn. Reson. Imaging 2001;13:313–317. © 2001 Wiley-Liss, Inc.
Article
Simultaneous multislice Echo Planar Imaging (EPI) acquisition using parallel imaging can decrease the acquisition time for diffusion imaging and allow full-brain, high-resolution functional MRI (fMRI) acquisitions at a reduced repetition time (TR). However, the unaliasing of simultaneously acquired, closely spaced slices can be difficult, leading to a high g-factor penalty. We introduce a method to create interslice image shifts in the phase encoding direction to increase the distance between aliasing pixels. The shift between the slices is induced using sign- and amplitude-modulated slice-select gradient blips simultaneous with the EPI phase encoding blips. This achieves the desired shifts but avoids an undesired "tilted voxel" blurring artifact associated with previous methods. We validate the method in 3× slice-accelerated spin-echo and gradient-echo EPI at 3 T and 7 T using 32-channel radio frequency (RF) coil brain arrays. The Monte-Carlo simulated average g-factor penalty of the 3-fold slice-accelerated acquisition with interslice shifts is <1% at 3 T (compared with 32% without slice shift). Combining 3× slice acceleration with 2× inplane acceleration, the g-factor penalty becomes 19% at 3 T and 10% at 7 T (compared with 41% and 23% without slice shift). We demonstrate the potential of the method for accelerating diffusion imaging by comparing the fiber orientation uncertainty, where the 3-fold faster acquisition showed no noticeable degradation.
Article
Statistical mapping within a binary hypothesis testing framework is the most widely used analytical method in functional MRI of the brain. A common assumption in this kind of analysis is that the fMRI time series are independent and identically distributed in time, yet we know that fMRI data can have significant temporal correlation due to low-frequency physiological fluctuation (Weisskoff et al. [1993]; Proc Soc Magn Reson Med 9:7; Biswal et al. [1995]: Mag Reson Med 34:537-541). Furthermore, since the signal-to-noise ratio will vary with imaging rate, we should expect that the degree of correlation will vary with imaging rate. In this paper, we investigate the effect of temporal correlation and experimental paradigm on false-positive rates (type I error rates), using data synthesized through a simple autoregressive plus white-noise model whose parameters were estimated from real data over a range of imaging rates. We demonstrate that actual false-positive rates can be biased far above or below the assumed significance level alpha when temporal autocorrelation is ignored in a way that depends on both the degree of correlation as well as the paradigm frequency. Furthermore, we present a simple method, based on the noise model described above, for correcting such distortions, and relate this method to the extended general linear model of Worsley and Friston ([1995]: Neuroimage 2:173-181).
Article
Previous studies have shown that under some conditions, noise fluctuations in an fMRI time-course are dominated by physiological modulations of the image intensity with secondary contributions from thermal image noise and that these two sources scale differently with signal intensity, susceptibility weighting (TE) and field strength. The SNR of the fMRI time-course was found to be near its asymptotic limit for moderate spatial resolution measurements at 3 T with only marginal gains expected from acquisition at higher field strengths. In this study, we investigate the amplitude of image intensity fluctuations in the fMRI time-course at magnetic field strengths of 1.5 T, 3 T, and 7 T as a function of image resolution, flip angle and TE. The time-course SNR was a similar function of the image SNR regardless of whether the image SNR was modulated by flip angle, image resolution, or field strength. For spatial resolutions typical of those currently used in fMRI (e.g., 3 x 3 x 3 mm(3)), increases in image SNR obtained from 7 T acquisition produced only modest increases in time-course SNR. At this spatial resolution, the ratio of physiological noise to thermal image noise was 0.61, 0.89, and 2.23 for 1.5 T, 3 T, and 7 T. At a resolution of 1 x 1 x 3 mm(3), however, the physiological to thermal noise ratio was 0.34, 0.57, and 0.91 for 1.5 T, 3 T and 7 T for TE near T2*. Thus, by reducing the signal strength using higher image resolution, the ratio of physiologic to image noise could be reduced to a regime where increased sensitivity afforded by higher field strength still translated to improved SNR in the fMRI time-series.
  • M Smith
  • C F Beckmann
  • J Andersson
  • E J Auerbach
  • J Bijsterbosch
  • G Douaud
  • E Duff
  • D A Feinberg
  • L Griffanti
  • M P Harms
  • M Kelly
  • T Laumann
  • K L Miller
  • S Moeller
  • S Petersen
  • J Power
  • G Salimi-Khorshidi
  • A Z Snyder
  • A T Vu
  • M W Woolrich
  • J Xu
  • E Yacoub
  • D C Van Essen
  • M F Glasser
M. Smith, C. F. Beckmann, J. Andersson, E. J. Auerbach, J. Bijsterbosch, G. Douaud, E. Duff, D. A. Feinberg, L. Griffanti, M. P. Harms, M. Kelly, T. Laumann, K. L. Miller, S. Moeller, S. Petersen, J. Power, G. Salimi-Khorshidi, A. Z. Snyder, A. T. Vu, M. W. Woolrich, J. Xu, E. Yacoub, K. U?urbil, D. C. Van Essen, and M. F. Glasser. Resting-state fMRI in the Human Connectome Project. Neuroimage, 80:144–168, Oct 2013. 5 1x 6x Simultaneous Multi-Slice BOLD Clinical MAGNETOM Flash | (63) 3/2015 | www.siemens.com/magnetom-world 69