Download Report

NMR IN BIOMEDICINE
NMR Biomed. 2001;14:408–412
DOI:10.1002/nbm.715
Investigation of the initial dip in fMRI at 7 Tesla
Essa Yacoub, Amir Shmuel, Josef Pfeuffer, Pierre-Francois Van De Moortele, Gregor Adriany, Kamil
Ugurbil and Xiaoping Hu*
Center for Magnetic Resonance Research and Department of Radiology, University of Minnesota, Minneapolis, MN, USA
Received 5 February 2001; Revised 10 May 2001; Accepted 16 May 2001
ABSTRACT: In agreement with optical imaging studies, previous fMRI studies have reported an initial decrease (i.e.
the initial dip) in the BOLD response, which is believed to arise from an increase in oxygen consumption and to be
mostly microvascular. To date, experimental studies of the initial dip in humans have been performed at fields up to 4
T, with relatively low spatial resolution. Because the sensitivity to microvascular contribution is increased at high
magnetic fields, the present study investigated the initial dip at 7 T. In addition, to reduce the partial volume effect, the
study is conducted at a high spatial resolution. The initial dip was detected in all subjects studied and was found to
reside mostly in the gray matter. The relative amplitude of the early response was found to be 0.6, higher than that at
4 T (0.3) and 1.5 T (0.11). In addition, based on the assumption that the initial dip is a result of increased oxygen
utilization, the fractional change in oxygen utilization was estimated to be 40% of that of the fractional change in
cerebral blood flow. These results are in agreement with the notion that the initial dip arises from an increase in
oxygen consumption. Copyright  2001 John Wiley & Sons, Ltd.
KEYWORDS: BOLD; fMRI; initial dip; high-resolution; CMRO2
INTRODUCTION
The detection of the hemodynamic response secondary to
brain activation, using blood oxygenation level dependent (BOLD)1–3 contrast in magnetic resonance imaging,
has become a routine technique for mapping brain
function. However, how specific this hemodynamic
response is as a marker of neural activity is still under
debate. Several studies, conducted at fields up to 4 T,
have shown that BOLD contrast is associated primarily
with large vessels, particularly draining veins,4–7 which
may be distant from the actual site of neuronal activity. In
addition, it has been suggested that the hemodynamic
response itself may not be specific to the site of neuronal
activity. Intrinsic signal optical imaging studies,8 which
have the ability to assess the oxygenation state of
hemoglobin with high spatial and temporal resolutions,
have provided some details regarding the characteristics
of physiological responses to neural activity. In parti*Correspondence to: X. Hu, Center for Magnetic Resonance Research,
2021 Sixth Street SE, University of Minnesota, Minneapolis, MN
55455, USA.
Email: xiaoping@cmrr.umn.edu
Contract/grant sponsor: National Institutes of Health; contract grant
number: P41RR08079; contract grant number: RO1MH55346.
Contract/grant sponsor: W. M. Keck Foundation.
Contract/grant sponsor: National Foundation for Functional Brain
Imaging.
Abbreviations used: CBF, cerebral blood flow; CMRO2, cerebral
metabolic rate of oxygen; OVS, outer volume suppression; PVE,
partial volume effect.
Copyright  2001 John Wiley & Sons, Ltd.
cular, these studies have reported a biphasic response9,10
in deoxyhemoglobin concentration, with an early increase, which began shortly after stimulus onset and
lasted about 4 s, followed by a subsequent decrease that
remained for several seconds. Furthermore, it was also
demonstrated that the early response was localized to
columnar structures, while the late response extended
beyond the active columns, suggesting a 2–3 mm underlying limitation of the spatial specificity of the positive
BOLD response in fMRI10 using single condition
mapping. While this conjecture seems to contradict fMRI
data that exhibited patterns corresponding to ocular
dominance columns,11 those patterns were obtained with
differential mapping and their correspondence to columnar activity remains to be verified. Furthermore, recent
data in a cat model12 demonstrated that the initial dip
provided more reliable orientation columns than the
delayed response.
Subsequent to the optical imaging studies, the initial
increase in deoxyhemoglobin concentration, which should
lead to a decrease in the BOLD signal, was observed with
both functional MRS13,14 and fMRI.15,16 The initial fMRI
studies agreed remarkably well with the optical imaging
data. To further understand the early response, additional
experimental data were obtained, addressing the issues of
the echo-time dependence of the early response,17 its
detectability at low field and thereby its field dependence,18 its presence in regions other than the visual
cortex17 and potential confounds in previous studies.17
Theoretical analyses suggest that the BOLD contrast
NMR Biomed. 2001;14:408–412
INITIAL DIP AT 7 TESLA
can be characterized by two components: one that is
proportional to the static magnetic field and another that
increases quadratically with the static magnetic field.20
The former is believed to arise from macroscopic vessels
and the latter from the microvasculature. If the early
response reflects the initial increase in oxygen utilization,
it is expected to be mainly microvascular and to increase
with the static magnetic field quadratically. On the other
hand, the late response in the voxels exhibiting the dip
may be a combination of both micro- and macro-vascular
contributions, particularly at low resolution, and is
expected to increase with the static field strength slower
than quadratically. As a result, the ratio of the initial dip
amplitude to that of the hyperemic response should
increase with field strength. At 1.5 T,18 the ratio of the
early response to the delayed positive response was found
to be 0.11, much less than 0.3 for 4 T.16 In this work, the
initial dip is studied at 7 T to provide additional evidence
for its field dependence.
Although recent studies in a cat model have utilized
the dip to map columnar structure at high spatial
resolution,12 previous human studies of the initial dip
have utilized a relatively low spatial resolution (1.5 1.5
5 mm or 3 3 5 mm) and were affected more by the
partial volume effect (PVE). The PVE may reduce the
relative amplitude of the observed dip signal if the dip
arises mainly from microvasculature, as the PVE mixes
microvascular and macrovascular contributions to the
late response. With the increase in signal-to-noise ratio
(SNR) and the BOLD contrast at 7 T,21 high-resolution
studies of the initial dip in humans become feasible and
were conducted to provide a detailed examination of the
initial dip.
There have been several MR studies aimed at assessing
the change of oxygenation consumption (i.e. cerebral
metabolic rate of oxygen or CMRO2) associated with
neuronal stimulation.22–25 Nonetheless, the exact value of
the neural activity-induced change in CMRO2 is still
under debate. Based on the assumptions that the initial
dip arises from the increase in CMRO2 and that PVE is
negligible at the spatial resolution used in this study, an
estimation of the fractional CMRO2 change relative to
the fractional change of cerebral blood flow (CBF) was
obtained, providing an alternative estimation of the
CMRO2 change that is in agreement with the published
data.
409
The experiments were performed on a 90 cm bore 7 T
human system, controlled by a Varian console (Varian
Inc., Palo Alto, CA). A 6 cm quadrature surface coil was
used for transmission and detection. Inversion recovery
images were obtained to localize the slices of interest.
Based on the scout image, two to three sagittal slices were
positioned a few millimeters off the midline, cutting
through the calcarine sulcus. T2*-weighted functional
images were acquired with a FOV of 3.2 cm in the phaseencoding direction, 12.8 cm in the readout direction and a
2 mm slice thickness. Outer volume suppression (OVS)
using a B1 insensitive technique26 was used in the phase
encoding direction. Single-shot GE (gradient echo) EPI
images, with a total readout time of 30 ms, were acquired
using a matrix size of 32 128, yielding an inplane
resolution of 1.0 1.0 mm. Two (or three) EPI slices
were acquired (150 ms/slice) with a TR of 300 ms (or
450 ms), a TE of 20 ms (T2*gm = 25 ms21), and a flip
angle of 35° (or 40°). A single adiabatic RF pulse was
applied along with appropriate gradients to saturate the
posterior and anterior sides of the desired field of view.
The SAR was calculated to be 0.9 W/kg which is well
within FDA limits.
Visual stimulation was presented via a mirror placed
over the subject’s eyes and a rear projection setup that
transmitted the stimulus onto a screen placed behind the
subject. The subject was presented with flickering black
and white checkerboard patterns and asked to fixate on a
fixation point in the center of the screen. Each epoch of
the stimulus presentation consisted of a 4 s on period and
32 s off period. The epoch was repeated 12 times in a
single run. To minimize head motion, a bite bar was
employed during the study. In addition, the subject’s
breathing and heart rate were monitored.
Data processing and analysis
The measured k-space data were preprocessed to remove
physiological fluctuations associated with respiration and
heart beat using a retrospective technique.27 Subsequently, the data were Fourier transformed into the image
space for further analysis.
METHODS
Data acquisition
Five normal subjects (three female, two male, with ages
between 20 and 25) participated in this study. All subjects
provided informed consent to the experimental protocol,
which was approved by the institutional review board at
the University of Minnesota.
Copyright  2001 John Wiley & Sons, Ltd.
Figure 1. Correlation template used for detecting the initial
dip signal (solid line) and the positive response (dashed line).
The rectangular box depicts the stimulus presentation
NMR Biomed. 2001;14:408–412
410
E. YACOUB ET AL.
Figure 2. The time course of pixels activated in the initial dip
map shown in Plate 1(a) is represented by the solid line. The
dashed line represents the results of ®tting the hemodynamic
response to the dip pixels. The region between the vertical
lines is that used for the ®t. The rectangular box depicts the
stimulus presentation
The image time series were processed with correlation
analysis28 using two templates, one for the detection of
initial dip and the other for detecting the pixels exhibiting
the late positive response (Fig. 1), regardless of its initial
behavior. The template for the positive response was
generated by convolving the stimulation paradigm with a
hemodynamic response function given by Friston et al.30
The negative response model was derived from a
representative timecourse of the previously published
data.15–19 An initial dip map was generated with a cross
correlation threshold of 0.52, corresponding to a pixelwise statistical significance of p < 0.02, and spatial
cluster size threshold of four pixels, leading to a
significance of p < 0.0001 according to the results of
Forman et al.29 The connectivity was defined in two
dimensions with pixels having one common edge defined
as connected. Note that the cluster size threshold was
used to eliminate isolated pixels and does not degrade the
spatial resolution in the resultant map. The correlation
coefficient for the late positive response was thresholded
at two levels, one at the same significance level as that for
the initial dip map (p < 0.0001) and the other achieved by
varying the correlation threshold until the number of
activated pixels matched that of the activated pixels in the
dip map. The latter was obtained to ascertain if the
difference between the dip map and the positive response
map was due to sensitivity differences. The spatial
overlap between the initial dip and the positive maps was
determined for all subjects.
1(a)] exhibits more detailed structure and appears
localized to the gray matter (91 5%). The composite
map in Plate 1(c) shows that most dip pixels exhibit a
significant positive response (hence overlapping with the
positive map) and pixels showing positive response alone
do not simply surround the dip pixels. Contrasting the dip
map to the positive map thresholded to produce the same
number of activated pixels [Plate 1(d)], there are
substantial differences, indicating that the difference in
the spatial pattern of Plate 1(a) and Plate 1(b) is not a
result of the sensitivity differences. In fact, the overlap
between these maps averaged over all subjects was only
32 4%.
The solid line in Fig. 2 represents the time course of
pixels showing the initial dip in the map in Plate 1(a). The
amplitude of the dip response is almost comparable to
that of the positive response. Quantitative analysis shows
that the average ratio between the amplitudes of the
negative response and the positive response is 0.6 0.1.
Compared with 0.11 at 1.5 T18 and 0.3 at 4 T,16 the ratio
of 0.6 at 7 T indicates that the relative contribution of the
initial response increases with the field strength.
Qualitatively, this increase is consistent with the notion
that the early response is more of a microvascular origin,
while the positive response is less so and the microvascular response is supposed to increase with B0 more
rapidly. However, the dependence of the hyperemic
response on the magnetic field is more complicated. First,
the relative contributions of macro- and micro-vascular
components change with B0, with the latter becoming
more prominent; in fact, recent experimental results21
suggest that the microvascular contribution becomes
dominant at 7 T. Second, if we assume that the dip signal
is solely microvascular, the contribution of the macrovascular signal to the hyperemic response in the dip
pixels also depends on the amount of PVE, decreasing
with spatial resolution. Therefore, the increased ratio of
the dip to the hyperemic response can be attributed to all
of the above factors. Nonetheless, the dip signal is more
prominent at 7 T, making its detection and mapping
easier.
RESULTS AND DISCUSSION
A significant initial decrease in the BOLD signal was
detected in all five subjects. The functional maps
corresponding to the initial dip and those for the positive
response are shown in Fig. 2 for one subject. Compared to
the positive response map thresholded at the same
statistical significance [Plate 1(b)], the dip map [Plate
Copyright  2001 John Wiley & Sons, Ltd.
Figure 3. The difference between the dip time course (solid
line in Fig. 2) and the ®tted positive response (dotted line in
Fig. 2). This curve provides an estimate of the signal arising
from the oxygen consumption increase
NMR Biomed. 2001;14:408–412
Plate 1. (a) The initial dip map for one slice of one subject overlaid on the corresponding T1-weighted anatomic image. (b) The corresponding positive response map
obtained at the same statistical con®dence. (c) A composite map contrasting the dip map with the positive map. The overlap is shown in yellow, the pixels with dip alone are
shown in red and the pixels with positive response only are shown in green. (d) The positive response map thresholded to match the number of pixels in the initial dip map.
Color bars in panels (a), (b), and (d) indicate the color coding for the correlation coef®cients
INITIAL DIP AT 7 TESLA
Copyright  2001 John Wiley & Sons, Ltd.
NMR Biomed. 2001;14
INITIAL DIP AT 7 TESLA
While the high spatial resolution used in this study is
not necessary for the observation of the dip signal, it
enhanced the detected amplitude by reducing the PVE. In
fact, when lower-resolution images were simulated by
truncating the measured k-space (data not shown), the dip
signal detected showed a similar spatial extent but with a
reduced amplitude. This observation is also qualitatively
consistent with the much lower amplitude of the dip
observed in MRS studies where large voxels were
used.13,14 On the other hand, the use of high resolution
degraded the SNR in the raw data, slightly compromising
the detection sensitivity. Thus, the dependence on spatial
resolution of the dip requires further investigation.
While the initial dip is clearly evident in the 7 T data, it
is still a small signal that requires careful experimental
design and effort to detect. For example, the reproducibility of the detected dip pixels, calculated by comparing
the results of the two halves of each run, was found to be
53%, a number below that for the positive response
(59%). In addition, a cluster size threshold was used to
remove isolated pixels in order to improve the reliability.
A semi-quantitative interpretation of the measured
data can be obtained with the following simplified model.
The relaxivity arising from deoxygenated hemoglobin
can be written as
R 2 / CBV 1
Y
1
where Y is the blood oxygen content and CBV is the
cerebral blood volume. The relative change of the BOLD
signal, (DS/S), is proportional to DR 2* which is given
by20
R 2 Y =1
Y
CBV/CBV
2
Using the conservation of matter and assuming that the
relative changes in CMRO2 and CBF are small enough
that high order terms can be ignored,
Y =1
Y / CBF/CBF
CMRO2 =CMRO2
3
Assuming that Grubb’s relationship31 holds, DCBV/CBV
0.38 DCBF/CBF. Consequently,
S=S1
R 2 CMRO2 =CMRO2
0:62CBF/CBF
4
Therefore, the response of pixels exhibiting the initial dip
can be described as a summation of two components: one
[first term in eqn (4)] reflecting the CMRO2 change and
the other [second term in eqn (4)] corresponding to the
BOLD response as a result of the change of CBF (i.e. the
hemodynamic response). It is assumed that the former
follows the latter by a few seconds.32 Assuming that the
hemodynamic response of the dip pixels takes the same
form as that of pixels exhibiting the positive response
only, the shape of the hemodynamic response is
estimated from the timecourse of pixels exhibiting the
Copyright  2001 John Wiley & Sons, Ltd.
411
positive response only. This response form was then
fitted to a 10 s portion of the time course of the dip pixels,
starting at 2 s after the stimulus cessation, to obtain the
hemodynamic response in these pixels. Subsequently, a
subtraction of the fitted response from the actual dip time
course provided an estimate of the CMRO2 response [see
eqn (4)]. It is also interesting to note that taking the
amplitudes of the dip and the hyperemic response directly
from the time course underestimates the ratio of the
CMRO2-related change over CBF-related change because they tend to cancel each other in the region where
they overlap. The result of modeling the response of
pixels exhibiting the dip for one of the subjects is shown
in Fig. 2. In this figure, the dotted line shows the fit of
scaling the time course of the pixels exhibiting the
positive response alone to the time course of dip pixels
(solid line). The difference time course shown in Fig. 3 is
taken as the MR response to CMRO2 changes. The ratio
between this amplitude and that of the positive amplitude
was calculated for all subjects. When averaged over all
five subjects, the ratio was 0.72 0.15. Thus,
CMRO2 =CMRO2 0:72 0:62CBF/CBF
0:44 CBF/CBF
5
In other words, DCMRO2/CMRO2 is about 40% of
DCBF/CBF. For example, a 50% change in CBF means a
22% change in CMRO2. This value falls within the range
of published values.22–25
Of course, the CMRO2 change arrived at above is only
an estimate because it depended on a number of
assumptions. In addition to assuming that the dip signal
arises from an increase in CMRO2, it was assumed that
the CBF-related BOLD change in the dip pixels had the
same form as the BOLD response in pixels not exhibiting
the dip. While this assumption may not be fully satisfied,
particularly when large vessel contributions are substantial, it is a reasonable assumption for the present data
because at 7 T the large vessel contributions are
diminished20 and the PVE is negligible with the spatial
resolution used. If large vessel contributions are significant in the time course of pixels not exhibiting the dip,
this time course may be somewhat delayed relative to the
CBF related BOLD effects in the dip pixels, which are
presumably microvascular. As a result, this may lead to
an underestimation of CMRO2.
CONCLUSION
In this work, the initial dip in the BOLD response to
visual stimulation was investigated at 7 T with high
spatial resolution. Experimental data indicated that the
dip could be readily detected at this ultrahigh magnetic
field. The dip was found to reside in a spatial pattern that
appeared to be more localized to the gray matter and
differed from that of the positive response, suggesting
NMR Biomed. 2001;14:408–412
412
E. YACOUB ET AL.
that it might be more specific. The amplitude of the dip
relative to the hyperemic response was found be
substantially higher than that at 1.5 and 4 T. In addition,
based on fitting the time course of the dip pixels, the
fractional increase of CMRO2 was estimated to be 40%
of that of the fractional increase in CBF. These results are
in agreement with the notation that the initial dip arises
from an increase in oxygen consumption.
15.
16.
17.
REFERENCES
1. Ogawa S, Lee T-M. Magnetic resonance imaging of blood vessels
at high fields: in vivo and in vitro measurements and image
simulation. Magn. Reson. Med. 1990; 16: 9–18.
2. Ogawa S, Lee T-M, Kay AR, Tank DW. Brain magnetic resonance
imaging with contrast dependent on blood oxygenation. Proc. Natl
Acad. Sci. USA 1990; 87: 9868–9872.
3. Ogawa S, Lee T-M, Nayak AS, Glynn P. Oxygenation-sensitive
contrast in magnetic resonance image of rodent brain at high
magnetic fields. Magn. Reson. Med. 1990; 14: 68–78.
4. Duyn JH, Moonen CTW, Yperen GH, Boer RW, Luyten PR.
Inflow versus deoxyhemoglobin effects in BOLD functional MRI
using gradient echoes at 1.5T. NMR Biomed. 1994; 7: 83–88.
5. Frahm J, Merboldt KD, Hanicke W, Kleinschmidt A, Boecker H.
Brain or vein-oxygenation or flow? On signal physiology in
functional MRI of human brain activation. NMR Biomed. 1994; 7:
45–53.
6. Lai S, Hopkins AL, Haacke EM, Li D, Wasserman BA, Buckley P,
Friedman L, Meltzer H, Hedera P, Friedland R. Identification of
vascular structures as a major source of signal contrast in high
resolution 2D and 3D functional activation imaging of the motor
cortex at 1.5 T: preliminary results. Magn. Reson. Med. 1993; 30:
387–392.
7. Kim S-G, Hendrich K, Hu X, Merkle H, Ugurbil K. Potential
pitfalls of functional MRI using conventional gradient-recalled
echo techniques. NMR Biomed. 1994; 7: 69–74.
8. Frostig RD, Lieke EE, Ts’o DY, Grinvald A. Cortical functional
architecture and local coupling between neuronal activity and the
microcirculation revealed by in vivo high-resolution optical
imaging of intrinsic signals. Proc. Natl Acad. Sci. USA 1990; 87:
6082–6086.
9. Grinvald A, Frostig RD, Siegel RM, Bartfeld RM. High-resolution
optical imaging of functional brain architecture in the awake
monkey. Proc. Natl Acad. Sci. USA 1991; 88: 11559–11563.
10. Malonek D, Grinvald A. The imaging spectroscopy reveals the
interaction between electrical activity and cortical microcirculation: implication for optical, PET, and MR functional brain
imaging. Science 1996; 272: 551–554.
11. Menon RS, Ogawa S, Strupp JP, Ugurbil K. Ocular dominance in
human V1 demonstrated by functional magnetic resonance
imaging. J. Neurophysiol. 1997; 77: 2780–2787.
12. Kim DS, Duong TQ, Kim S-G. High-resolution mapping of isoorientation columns by fMRI. Nature Neurosci. 2000; 3: 164–169.
13. Ernst T, Hennig J. Observation of a fast response in functional MR.
Magn. Reson. Med. 1994; 32: 146–149.
14. Hennig J, Janz C, Speck O, Ernst T. Functional spectroscopy of
Copyright  2001 John Wiley & Sons, Ltd.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
brain activation following a single light pulse: examination of the
mechanism of the fast initial response. Int. J. Imag. Systems
Technol. 1995; 6: 203–208.
Menon RS, Ogawa S, Hu X, Strupp JS, Andersen P, Ugurbil K.
BOLD based functional MRI at 4 Tesla includes a capillary bed
contribution: echo-planar imaging mirrors previous optical
imaging using intrinsic signals. Magn. Reson. Med. 1995; 33:
453–459.
Hu X, Le TH, Ugurbil K. Evaluation of the early response in fMRI
in individual subjects using short stimulus duration. Magn. Reson.
Med. 1997; 37: 877–884.
Yacoub E, Le TH, Ugurbil K, Hu X. Further evaluation of the
initial negative response in functional magnetic resonance
imaging. Magn. Reson. Med. 1999; 41: 436–441.
Yacoub E, Hu X. Detection of the early negative response in fMRI
at 1.5 Tesla. Magn. Reson. Med. 1999; 41: 1088–1092.
Yacoub E, Hu X. Detection of the early decrease in fMRI signal in
the motor area. Magn. Reson. Med. 2001; 45: 184–190.
Ugurbil K, Hu X, Chen W, Zhu XH, Kim SG, Georgopoulos A.
Functional mapping in the human brain using high magnetic fields.
Phil. R. Soc. Lond. B 1999; 354: 1195–1123.
Yacoub E, Shmuel A, Pfeuffer J, Van de Moortele P-F, Adriany G,
Andersen P, Vanghan JT, Merkle H, Ugurbil K, Hu X. Imaging
brain function in humans at 7 T. Magn. Reson. Med. 2001; 45:
588–594.
Kim S-G, Rostrup E, Larsson HBW, Ogawa S, Paulson OB.
Determination of relative CMRO2 from CBF and BOLD changes:
significant increase of oxygen consumption rate during visual
stimulation. Magn. Reson. Med. 1999; 41: 1152–1161.
Hoge RD, Atkinson J, Gill B, Crelier GR, Marrett S, Pike GB.
Linear coupling between cerebral blood flow and oxygen
consumption in activated human cortex. Proc. Natl Acad. Sci.
USA 1999; 96(16): 9403–9408.
Kida I, Kennan RP, Rothman DL, Behar KL, Hyder F. Highresolution CMR(O2) mapping in rat cortex: a multiparametric
approach to calibration of BOLD image contrast at 7 Tesla. J.
Cereb. Metab. Blood Flow 2000; 20(5): 847–860.
Davis TL, Kwong KK, Weisskoff RM, Rosen BR. Calibrated
functional MRI: mapping the dynamics of oxidative metabolis.
Proc. Natl Acad. Sci. USA 1998; 95: 1834–1839.
Luo Y, DeGraaf RA, LelaBarre L, Tannus A, Garwood M. An
outer-volume suppression method that tolerates RF field inhomogeneity, BISTRO. Magn. Reson. Med. 2001; 45: 1095–1102.
Hu X, Le TH, Parrish T, Erhard P. Retrospective estimation and
correction of physiological fluctuation in functional MRI. Magn.
Reson. Med. 1995; 34: 201–212.
Bandettini PA, Jesmanowicz A, Wong EC, Hyde JS, Processing
strategies for time-course data sets in functional MRI of the human
brain. Magn. Reson. Med. 1993; 30(2): 161–173.
Forman S, Cohen J, Fitzgerald M, Eddy W, Mintun M, Noll D.
Improved assessment of significant activation in functional
magnetic resonance imaging (fMRI): use of cluster-size threshold.
Magn. Reson. Med. 1995; 33: 636–647.
Friston KJ, Jezzard P, Turner R. Analysis of functional MRI timeseries. Human Brain Mapping 1994; 1: 153–171.
Grubb J, Raichle ME, Eichling JO, Ter-Pogossian MM. The effects
of changes in PaCO2 on cerebral blood volume, blood flow, and
vascular mean transit time. Stroke 1974; 5: 630–639.
Grinvald A, Slovin H, Vanzetta I. Nature Neurosci. 2000; 3: 105–
107.
NMR Biomed. 2001;14:408–412