Sriranga Kashyap1, Dimo Ivanov1, Martin Havlíček1, Benedikt A Poser1, and Kâmil Uludağ1
1Department of Cognitive Neuroscience, Maastricht University, Maastricht, Netherlands
Synopsis
Cortical-depth dependent fMRI usually relies on
the definition of depths on an anatomical image (eg. MPRAGE). The geometric dissimilarities
of the functional compared to the anatomical data require further spatial
processing of the functional data to ensure good co-registration. We propose an
alternative approach that uses an optimised inversion-recovery EPI derived T1
image, whose resolution and readout, hence distortions, are identical to that
of the functional data, in order to delineate cortical depths. As a result, the
cortical-depth specific fMRI data can be analysed in the native space without
any spatial confounds stemming from distortion correction and inaccurate
registration.Introduction
Recently, there has been growing interest in
cortical depth-dependent fMRI, particularly due to increased accessibility of
ultra-high field scanners. Since the cortical depths are usually determined on an
anatomical image, good co-registration with the distorted functional
data is crucial in high-resolution studies. The standard approach has been to correct
the EPI for distortions using algorithms such as FSL-TOPUP
[1] and co-registering
the undistorted EPI to the anatomical data with the depth definitions. However,
distortion correction might reduce laminar information in the fMRI data due to smoothing
and interpolation, and sub-optimal registration might lead to erroneous
classification of depths. The present study builds upon previous work
[2][3]
to acquire, for the first time, sub-millimetre distortion-matched quantitative
T
1 image using inversion-recovery EPI (IR-EPI) at 7T and demonstrates
the feasibility of using the IR-EPI T
1 map to define cortical-depths
for fMRI analyses. In this study, cortical depth-dependent haemodynamic
responses at 7T were investigated using a full-contrast static and reduced-contrast
flickering hemi-annulus, a stimulus combination that is well-suited to study
the laminar distribution of the positive BOLD amplitude and the post-stimulus
undershoot
[4].
Methods
Images were acquired with a Siemens 7T scanner
(Erlangen, Germany) using a 32-channel head coil (Nova Medical, USA). Two subjects
participated in this study after written informed consent. Sequence parameters (Table
1) were optimised in pilot studies.
IR-EPI: Thirty inversion times (TI) were
acquired by placing EPI readouts following an optimized tr-FOCI
[5] inversion
pulse and permuting the slice order
[6] each TR
[2][3]. Ten
repetitions of each inversion time were averaged and T
1 and M
0
maps were obtained from the mean IR-EPI by modelling each voxel with $$$M(TI) =
M_{0}*(1-2*E*e^{(-TI/T_{1})}+e^{(-TR/T_{1})})$$$, where E is the inversion
efficiency, using a non-linear least squares algorithm in MATLAB (MathWorks,
USA).
Experiment: Static and flickering hemi-annuli
were created using PsychoPy
[7] and presented to the left hemi-field
in a block design for 20s followed by 40s of isoluminant grey background. Three
runs of ten blocks each were acquired for both conditions.
fMRI analyses: 3D gradient-echo (GE) EPI runs were motion-corrected and registered using SPM12. The
statistical analyses were done using FSL FEAT with standard haemodynamic
predictors. The IR-EPI T
1 image was co-registered to the 3D GE-EPI. The
region-of-interest (ROI) was manually delineated on the IR-EPI T
1
map using ITK-SNAP
[8] and ten cortical-depths were defined within
the ROI using an equidistant-layering algorithm in MATLAB.
Results
Figure 1 shows the enhanced grey-white matter contrast in the
distortion-matched IR-EPI T
1 map compared to the mean EPI image. The
co-registration is qualitatively presented by the statistical map overlaid on
both the images (Fig. 1 a, b). Ten cortical depths were defined in the ROI between
the CSF-GM and GM-WM boundaries as illustrated in Figure 2. It can be seen that
the depths are accurately positioned within the grey matter. The cortical
depth-dependent haemodynamic responses for both conditions are shown in Figure 3. Both
the flickering and static stimuli evoked similar positive BOLD responses as a
function of depth followed by post-stimulus timecourses depending on the depth
and stimulus. Cortical-depth profiles were then plotted for positive BOLD (Fig. 4a)
and the post-stimulus undershoot (PSU) (Fig. 4b), both of which, approximately
decrease linearly in amplitude as a function of depth. Since the PSU has similar
depth profile as the positive BOLD signal, the ratio of the PSU to the peak
BOLD amplitude is shown in Figure 4c. The ratio has its maximum magnitude in the
middle GM depth.
Discussion
The present study demonstrates, for the first
time, the feasibility of acquiring sub-millimetre resolution distortion-matched
anatomical images using IR-EPI at 7T. The advantage of the IR-EPI lies in the
fact that minimal spatial processing of the functional data is required and its
enhanced grey-matter contrast can be used to define cortical depths. Therefore,
the anatomical reference obtained by IR-EPI approach is a viable alternative to
the standard approaches in high-resolution fMRI studies.
Using this approach, we demonstrated that even
with gradient-echo (GE) fMRI laminar specificity could be achieved. GE signal
is largely macrovasculature weighted
[9] and this influences the
spatial specificity of the haemodynamic response across cortical depths. Thus,
the positive BOLD response alone may not reveal cortical-depth dependent differences
in neuronal activity. However, deconvolving the standard cortical-depth profile
of the response may be necessary to obtain information about laminar neuronal
activity from GE functional data. In this study, we normalized the
post-stimulus undershoot with the positive BOLD response. This analysis
revealed the cortical-depth dependent modulation of the PSU by removing the
standard cortical response profile of the GE BOLD response.
Acknowledgements
This work was financially supported by the VIDI grant (452-11-002).References
1. Andersson JL, Skare S,
Ashburner J. (2003), Neuroimage, 20(2):870-88.
2. Renvall V et al. (2014),
Proc. ISMRM, v. 22, p. 4282.
3. Ivanov D et al. (2015),
Proc. OHBM, p. 56, 1564.
4. Sadaghiani S, Ugurbil K,
Uludag K. (2009), Magn Reson Imaging, 27(8):1030-8.
5. Hurley AC et al. (2010),
Magn Reson Med, 63(1):51-8.
6. Clare S, Jezzard P. (2001),
Magn Reson Med, 45(4):630-4.
7. Peirce JW (2007), J
Neurosci Methods, 162(1-2): 8–13.
8. Yushkevich PA et al.
(2006), Neuroimage, 31(3):1116-28.
9. Uludag K, Müller-Bierl
B, Ugurbil K. (2009), Neuroimage, 48(1):150-65.