www.elsevier.com/locate/ynimg

NeuroImage 30 (2006) 726 – 734

The effect of global cerebral vasodilation on focal activation

hemodynamics

Bojana Stefanovic,* Jan M. Warnking, Karin M. Rylander, and G. Bruce Pike

McConnell Brain Imaging Centre, Montreal Neurological Institute, 3801 University St., Montreal, QC, Canada H3A 2B4

Received 4 July 2005; revised 19 October 2005; accepted 26 October 2005

Available online 5 December 2005

In view of the potential of global resting blood flow level to confound

the interpretation of blood oxygenation level-dependent (BOLD) fMRI

studies, we investigated the effect of pronounced elevation in baseline

cerebral blood flow (CBF) on BOLD and CBF responses to functional

activation. Twelve healthy volunteers performed bilateral finger

apposition while attending to a radial yellow/blue checkerboard. Three

levels of global CBF increase were achieved by inhaling 5, 7.5 or 10%

CO2. CBF and BOLD signals were simultaneously quantified using

interleaved multi-slice pulsed arterial spin labeling (PASL) and T2*-

weighted gradient echo sequences. Increasing basal CBF produced a

significant decrease in the activation-induced BOLD response, with the

slope of the optimal linear fit of activation versus basal BOLD signal

changes of �0.32 T 0.01%/% for motor and visual cortex regions of

interest (ROIs). While the modulation in basal flow level also produced

a statistically significant effect on the activation-induced CBF change,

the degree of relative attenuation of the flow response was slight, with a

slope of �0.18 T 0.02%/% in the motor and �0.13 T 0.01%/% in the

visual cortex ROI. The current findings describe a strong attenuation

of the BOLD response at significantly elevated basal flow levels and call

for independent quantification of resting CBF in BOLD fMRI studies

that involve subjects and/or conditions with markedly elevated global

perfusion.

D 2005 Elsevier Inc. All rights reserved.

Keywords: fMRI; BOLD; Hypercapnia; Perfusion; Basal vasodilation

Introduction

Functional magnetic resonance imaging (fMRI) using the blood

oxygenation level-dependent (BOLD) effect has become the

dominant tool for investigating human brain function. Nonetheless,

the dependence of the BOLD response on different physiological

parameters is still only partly understood. One parameter of

particular interest is the resting global cerebral blood flow which

may be perturbed in healthy volunteers by intake of, e.g., caffeine,

1053-8119/$ - see front matter D 2005 Elsevier Inc. All rights reserved.

doi:10.1016/j.neuroimage.2005.10.038

* Corresponding author.

E-mail address: bojana@bic.mni.mcgill.ca (B. Stefanovic).

Available online on ScienceDirect (www.sciencedirect.com).

alcohol or nicotine, due to increased anxiety (via the action of

epinephrine) or following therapeutic or experimental administra-

tion of various vasoactive drugs. In view of such common

modulations of global cerebral blood flow, the effect of basal

CBF on the magnitude and dynamics of the activation-induced

BOLD response has attracted much attention.

In spite of numerous investigations, no consensus has been

reached regarding the impact of baseline flow on either BOLD or

CBF response to neuronal activation in the healthy human brain.

Some reports, including those from our laboratory, described a

preserved stimulation-evoked BOLD response (expressed in

percents of normocapnic baseline) despite the hypercapnia-induced

increases in resting flow (Corfield et al., 2001; Hoge et al., 1999b)

or cocaine-induced decreases in global CBF (Gollub et al., 1998).

On the other hand, a steady decrease in BOLD response magnitude

with increasing basal flow (produced by either 5% inspired CO2 or

acetazolamide administration) has been observed (Brown et al.,

2003; Cohen et al., 2002; Bandettini and Wong, 1997; Bruhn et al.,

1994). Correspondingly, an increased BOLD response during

decreased basal flow, produced by either hyperventilation or

caffeine, was documented (Cohen et al., 2002; Mulderink et al.,

2002), though the effects of caffeine varied substantially across

subjects (Laurienti et al., 2003; Liu et al., 2004). Finally, a more

complicated dependence of the BOLD response on end-tidal CO2

was reported (Posse et al., 2001), with a diminished BOLD

response during hyperventilation-induced basal flow decreases

(Weckesser et al., 1999) and a maximal BOLD response magnitude

during mild hypercapnia (with end-tidal CO2 of 50 mm Hg) (Posse

et al., 2001). A decreased BOLD response following administra-

tion of the vasoconstrictive indomethacin has also been reported

(Bruhn et al., 2001).

The majority of both PET (Ramsay et al., 1993; Friston et al.,

1990; Maximillian et al., 1980) and MRI (Brown et al., 2003; Li et

al., 1999, 2000; Hoge et al., 1999a) human studies found a constant

absolute CBF change irrespective of the basal flow level. However,

an O-15 butanol PET study (Kemna et al., 2001) documented

increasing absolute CBF changes (but preserved fractional CBF

response) for end-tidal CO2 (ETCO2) elevation from 40 to 60 mm

Hg, with a severely attenuated (albeit extremely noisy) CBF

response at 70 mm Hg. Similarly, preserved fractional photic

Table 1

End-tidal CO2 (ETCO2) at various inspired CO2 (ICO2) levels (normo-

capnia: row 1; hypercapnia: rows 2, 3 and 4)

ICO2 (%) ETCO2 (mm Hg) DETCO2 (%)

0 38 T 1 0

5 46 T 1 21 T 2

7.5 51 T 1 34 T 2

10 59 T 1 55 T 3

The values correspond to the average across all subjects followed by the

standard error of the mean. The third column lists the mean percent increase

(Tstandard error) in the end-tidal CO2 during hypercapnia relative to the

normocapnic ETCO2.

B. Stefanovic et al. / NeuroImage 30 (2006) 726–734 727

stimulation-induced CBF changes were reported for eucapnia, a

single level of hypocapnia and a single level of hypercapnia

(Shimosegawa et al., 1995). While some variability in the results

may well have resulted from very different experimental paradigms

and possibly distinct CBF regulatory mechanisms elicited by

administration of various vasoactive agents, the lack of an effect of

resting CBF on the activation-induced BOLD response has often

been ascribed to a limited range of basal flow perturbations studied

(with steady-state inspired CO2 levels typically not exceeding 5%).

To address these issues, we sought to quantify the effect of a

wide range of steady-state basal flow increases on both BOLD

and CBF steady-state responses to neuronal activation in multiple

cortical regions. Following the prediction of the deoxyhemoglobin

dilution model of the BOLD fMRI response (Davis et al., 1998;

Hoge et al., 1999a), we hypothesized that the progressive decrease

of basal venous deoxyhemoglobin concentration at higher basal

flows will result in a narrowing of the dynamic range of BOLD

responses and decrease BOLD’s sensitivity to flow increases, so

that at higher basal CBF the same flow response to neuronal

activation produces a significantly diminished BOLD signal

change. Carbon dioxide was used to produce a robust and rapid

rise in global CBF (Poulin et al., 1996). BOLD and CBF signals

were simultaneously measured during normocapnia at rest, during

normocapnia with graded motor task (MT) and visual stimulation

(VS), during graded hypercapnia (GHC) at motor and visual

baseline and during graded hypercapnia with motor task and

visual stimulation. In contrast to earlier investigations, robust

estimation of the response magnitudes was achieved by explicitly

modeling the temporal variability in the global hypercapnic

response. We report here on the effect of pronounced global

vasodilatation on activation-induced BOLD and CBF changes in

visual and motor cortices. In addition to describing a confound in

the interpretation of BOLD fMRI studies, these data provide

insight into the bio-physical mechanism of the BOLD response as

well as the nature of focal CBF regulation following neuronal

activation.

Methods

Baseline CBF modulation

Mild-to-moderate hypercapnia was induced through adminis-

tration of mixtures of carbon dioxide and air through a non-

rebreathing face mask (Hudson RCI, Model 1069, Temecula, CA)

in 1 min off/3 min on/2 min off blocks. At baseline, the subjects

were inhaling medical air, supplied at 16 l/min. During

hypercapnic perturbations, a premixed preparation of 10% CO2,

21% O2 and balance N2 (BOC Canada Ltd., Montreal, Quebec,

Canada) was combined with medical air in a Y-connector. The

CO2 concentration in the mixture was set to 5, 7.5 or 10% and the

order of administration of different levels of CO2 randomized

across subjects. At each level, the total flow rate was maintained

at 16 l/min. End-tidal CO2 was measured via a nasal cannula with

monitoring aspirator (Normocap 200, Datex Inc., Plymouth,

Maine): the observed ETCO2 increases are listed in Table 1.

Subjects were asked to breathe at a constant rate. Prior to

scanning, subjects breathed 1 min of medical air followed by 3

min of 10% CO2 mixture to ensure they experienced no dyspnea

and to let them practice breathing the hypercapnic mixture at a

constant rate.

Motor task and visual stimulation

The subjects performed bilateral, sequential finger-to-thumb

apposition at low (1.5 Hz) or high (3 Hz) frequency (cued by a

metronome) while being presented with a radial yellow/blue

checkerboard at low (25%) or high (100%) contrast, reversing

contrast at 4 Hz. These Fon_ conditions alternated with rest and

uniform grey baseline in 0.5 min off/1.5 min on/1 min off blocks.

The low apposition frequency was accompanied by the low

contrast checkerboard, the high apposition frequency by the high

contrast checkerboard. Two functional blocks (one of low, one of

high level, in randomized order) preceded each of the three 6-min

hypercapnia blocks. In addition, a high level functional block was

applied so that its on-period coincided with either first or second

half of each hypercapnia-on interval. Fig. 1 displays the schematic

of the functional paradigm.

Data acquisition

Twelve healthy adults (7 females and 5 males; average age

27 T 1 years) participated in the study. The scanning protocol

consisted of a 3D RF-spoiled T1-weighted gradient echo (1 �1 � 2 mm3) sequence for anatomical reference, followed by

interleaved multi-slice PASL and T2*-weighted gradient echo

sequence for CBF and BOLD signal measurements. The high-

resolution gradient echo sequence employed a TR of 22 ms, a

TE of 10 ms and nonselective 30- RF-spoiled excitation. The

CBF and BOLD acquisitions covered 6 slices (4 � 4 � 5 mm3;

interslice gap of 1 mm) positioned to include the primary motor

and visual cortices. The CBF data were acquired using a

QUIPSS II sequence (Wong et al., 1997) with background

suppression (Ye et al., 2000). Two presaturation asymmetric

BASSI pulses (Warnking and Pike, 2004) in the imaging region

were followed by an adiabatic BASSI inversion pulse in the

labeling region (gap of 5 mm). In view of the large increases in

global CBF expected to result from the pronounced hypercapnic

challenges, we modified the common QUIPSS II parameteriza-

tion to minimize the risk of underestimation of CBF changes due

to a drop in the temporal width of the tag below the post-label

delay. We thus increased the spatial width of the labeling region

to 150 mm in addition to decreasing the QUIPSS II delay (TI1)

to 500 ms and the post-label delay (TI2) to 1100 ms. An EPI

readout (2232 Hz/pixel) was employed, with an echo time of 50

ms for BOLD and 22 ms for CBF. In both cases, the repetition

time was 1.5 s. Subjects were immobilized using a vacuum bag

and a head holder assembly. The RF body coil was used for

transmission and an 8-channel phased array head coil for signal

reception. All examinations were performed on a Siemens 1.5 T

Fig. 1. The schematic of a sample functional paradigm. The black bars

indicate the level of functional stimulation (low: 1.5-Hz apposition

frequency and 25% checkerboard contrast; high: 3-Hz apposition frequency

and 100% checkerboard contrast). The shaded regions represent hypercap-

nic periods (yellow: 5%, orange: 7.5%, and red: 10% ICO2).

B. Stefanovic et al. / NeuroImage 30 (2006) 726–734728

Magnetom Sonata system. The experimental protocol was

approved by the Research Ethics Board of the Montreal

Neurological Institute and informed consent obtained from each

subject prior to the scanning session.

Data analysis

Motion correction of the BOLD data set was performed using

AFNI’s 3dvolreg software (Cox, 1996). (Notably, none of the

estimated translations exceeded 2.5 mm, and the estimated

rotations were all below 0.9-.) The motion correction parameters

estimated for BOLD data were interpolated and the resulting

values applied to correct the CBF data for motion. The frames

with estimated translation exceeding 1 mm or rotation greater than

1- were excluded from the analysis. Both BOLD and CBF data

were spatially smoothed using a three-dimensional Gaussian filter

with full width half maximum of 6 mm. Drift was removed by

subtracting from each voxel’s time course the low-frequency

components of its discrete cosine transform, with a cutoff

frequency of one half of the stimulation paradigm frequency.

The generalized linear model (Worsley et al., 2002) was used to

identify areas of statistically significant task correlation at the

omnibus significance level of 0.05 (after correction for multiple

comparisons). A unique set of regions of interest, to be used in

the remainder of the analysis, was defined by thresholding the t

maps (at the t value corresponding to a of 0.05 after correction

for multiple comparisons) corresponding to the high-level

functional stimulation during normocapnia, with the motor and

visual cortex ROIs differentiated based on anatomical consider-

ations (namely with motor cortex ROI restricted to the precentral

gyrus and the visual cortex ROI, to the banks of the calcarine

sulcus). To allow for establishment of a physiological steady-state,

the data acquired within half a minute following a change in the

concentration of the inspired CO2 were excluded from the

analysis.

While most of the instability in the hypercapnic response

occurred in the initial 30 s of a hypercapnic period, as expected,

both BOLD and CBF signals exhibited considerable global

temporal variation over the following 2.5 min of hypercapnia.

Of particular concern for the present analysis were any

systematic differences in the breathing pattern between hyper-

capnia only periods and the neighboring intervals including both

hypercapnia and motor task/visual stimulation. It should be

noted that no hypercapnia-correlated motion was found in the

time courses of motion parameter estimates. Once the interac-

tion between hypercapnia- and stimulation-evoked response

magnitudes was established, robust estimation of BOLD and

CBF response amplitudes to functional activation during the

hypercapnic periods required explicit modeling of the temporal

instability in the global hypercapnia-induced response. To this

end, a reference grey matter region not participating in either

motor or visual processing was defined, and its average signal

time course used as a confound in the subsequent statistical

analysis to provide information about the variation of the global

grey matter hypercapnic response over the course of the

experiment.

Specifically, tissue segmentation of the anatomical data was

performed using a parametric Bayesian approach that assumes the

voxel intensities came from a mixed population of three statistical

distributions (grey matter (GM), white matter (WM) and

cerebrospinal fluid (CSF)) (Bezdek et al., 1993). The procedure

has been described in detail earlier (Stefanovic et al., 2003).

Briefly, manually selected training data set (100 voxels per class

per subject) allowed an explicit calculation of the parametric

estimates of the a posteriori probabilities as scaled likelihoods of

voxel intensities. The Bayes classification produced three nor-

malized a posteriori probability maps. The GM probability map

was then downsampled to the spatial resolution of the functional

acquisition and only the voxels having at least 60% normalized a

posteriori probability of being gray matter included in the

reference grey matter region. In selecting this threshold, both

spatial and temporal variability of the hypercapnic data in the

resulting region was considered, as to maximize the SNR of the

estimate of the instability in the GM response to hypercapnia. To

ensure that the reference region contained no voxels that were

detectably involved in either motor or visual processing, any

voxel with an absolute t value above 2 in the t maps

corresponding to the ‘‘high’’ functional activation condition was

excluded from this region. Finally, the selected region was

overlaid on the corresponding anatomical volume and careful

visual inspection done to ensure it included no voxels from higher

visual or motor areas.

From the time course of the average signal in the reference

grey matter region, we retained 3 intervals, corresponding to the

three blocks of hypercapnic perturbation, flanked by the

normocapnic baseline (7 min preceding and 2 min following the

hypercapnia block). These three time courses were zero padded

and used as confounds in the subsequent statistical analysis,

allowing robust estimation of activation-induced changes with the

effects of the temporal variation in global GM response to

hypercapnia removed. We thus exploited the spatial homogeneity

(to within a scaling factor) of the hypercapnic time course across

the grey matter (specifically, between the reference GM region

and MC/VC ROIs) to account for the temporal instability in the

hypercapnia-induced responses. It is important to note that the

primary objective of this analysis was the quantification of the

activation-elicited responses, not just detection of interaction

between hypercapnia and functional activation. In each subject,

both BOLD and CBF signals were averaged over the visual and

motor cortex ROIs (defined using the responses to high-level

functional challenges during normocapnia, as described above).

B. Stefanovic et al. / NeuroImage 30 (2006) 726–734 729

Temporal averaging (over each condition) was next performed

and the results expressed in percents of the average ROI signal

during the normocapnic baseline in a given subject. A straight line

was fit (Press et al., 1992) to the activation-induced responses as a

function of basal signal changes for both BOLD and CBF signals

in both motor and visual cortex ROIs of each subject. In addition,

the hypercapnia data were used to calibrate the deoxyhemoglobin

dilution model (Davis et al., 1998; Hoge et al., 1999a) and

estimate the CMRO2changes induced by functional activation: the

ensuing findings are described in the Appendix.

Fig. 3. The measured activation (motor cortex: red, visual cortex: green)-

and the estimated hypercapnia (blue)-induced changes corresponding to the

data of Fig. 2 (N = 1) during normocapnia (ICO2 = 0%) and three levels of

hypercapnia (ICO2 = 5, 7.5 or 10%). The two vertical bars during the

normocapnic condition correspond to low and high levels of stimulation.

The estimated HC responses are shown for the HC-only and the HC + VS/

MT conditions.

Results

A representative set of BOLD and CBF time courses, in MC

ROI, VC ROI and reference GM region of a subject, is shown in

Figs. 2a and b, with the corresponding estimates of hypercapnia-

and activation-induced responses in this subject displayed in Fig.

3. When controlling for intersubject variability, the effect of

hypercapnia on the activation-induced response was significant

for both BOLD (P < 10�6) and CBF (P < 10�4) signals. In

particular, DBOLD showed a small change (VC: by 0.5 T 1.8%,

with P ¨ 1.0; MC: by �9.8 T 1.8%, with P ¨ 0.01) during the

5% CO2 condition relative to that during normocapnia. In

contrast, there was a strong attenuation of the BOLD response

during the 7.5 and 10% CO2 challenges in both cortices, DBOLD

dropping by over 30% (P < 10�6) during the 7.5% CO2

condition and by over 50% (P < 10�6) during the 10% CO2

challenge.

The optimal straight line fit to the activation-induced responses

as a function of hypercapnia-induced changes, across all subjects,

is shown in Fig. 4. In view of the slopes of the optimal straight line

fits (�0.32 T 0.01%/% for BOLD MC; �0.32 T 0.01%/% for

BOLD VC; �0.18 T 0.02%/% for CBF MC and �0.13 T 0.01%/%

for CBF VC) and given that activation induced markedly smaller

percent signal changes than hypercapnia for BOLD, but not for

CBF, it is the effect of hypercapnia on the BOLD response to

functional activation (and not that on CBF changes elicited by

Fig. 2. Time courses of BOLD (a) and CBF (b) signal changes in a subject_s motor

not participating in either motor or visual processing. The horizontal black bars, i

The horizontal grey bars in the third row correspond to the intervals during which

ICO2 in this subject).

increased neuronal activity) that bears practical significance on this

range of baseline vasodilation. At the highest levels of HC-induced

vasodilation, the activation-induced BOLD response was reduced

by an across subject average of 57.0 T 1.1% in MC and 55.4 T1.1% in VC relative to the normocapnic condition. In contrast, the

corresponding CBF response at the maximum level of HC-induced

vasodilation was attenuated by only 13.9 T 2.0% in MC and 2.0 T1.9% in VC, relative to the CBF response in each of these cortices

during normocapnia. The across-subject average changes in BOLD

cortex ROI (red), visual cortex ROI (green), and reference GM region (blue)

n the first two rows, indicate the motor task/visual stimulation ON periods.

a hypercapnic mixture was administered (in the order of 7.5%, 10% and 5%

Fig. 4. The activation-induced percent signal changes as a function of the hypercapnia-induced percent signal changes in BOLD (motor cortex: a, visual cortex:

b) and CBF signals (motor cortex: c, visual cortex: d) for the high-level visual stimulation and motor task across all subjects and all normocapnic and

hypercapnic conditions. The percent values are expressed with respect to the normocapnic baseline.

B. Stefanovic et al. / NeuroImage 30 (2006) 726–734730

and CBF responses to activation in VC and MC cortices are

summarized in Table 2.

Discussion

The main finding of this work is the significant drop of

activation-induced BOLD response magnitude with increasing

Table 2

Across-subject average activation-induced changes in BOLD and CBF

signals in VC and MC cortices during normocapnia, NC (column 2) and

during themaximalHC-induced vasodilation condition,MAXHC (column3)

DSNCact (%) DSMAXHC

act (%) Change (%)

BOLDMC 0.99 T 0.01 0.42 T 0.01 �57.0 T 1.1

BOLDVC 1.15 T 0.01 0.51 T 0.01 �55.4 T 1.1

CBFMC 45.6 T 0.57 39.3 T 0.8 �13.9 T 2.0

CBFVC 40.0 T 0.43 39.2 T 0.6 �2.0 T 1.9

For reference, the percent changes between these two conditions are listed

in column 4.

global flow levels, concomitant to a small reduction in the

activation-evoked CBF response. This decreased sensitivity of

BOLD response to CBF changes is predicted by the deoxyhemo-

globin dilution model (Davis et al., 1998; Hoge et al., 1999a) and

likely results from the pronounced decrease in the basal venous

deoxyhemoglobin concentration due to the large global flow

elevation and hence a curbed dynamic range of the BOLD signal

increases brought about by increases in resting flow. Graphically,

this corresponds to the flattening of the iso-CMRO2 contours, at

high perfusion levels, in the BOLD-CBF plane such as that shown

in Fig. 1b of Hoge et al. (1999a).

The relatively slow CBF response to step changes in ETCO2

(estimated time constant of 45 s with delay of 6 s (Poulin et al.,

1996)) precludes ready comparison between the current study of

steady-state changes and earlier investigations of transient hyper-

capnic challenges. A slope of �0.14%/% was reported by Cohen et

al. (2002) at 7 T while employing transient (4 s) visual stimulation

and a 5% CO2 mixture inhalation. In addition to differences arising

from transient versus steady-state stimulation, we do expect a

lower slope to be estimated when employing a significantly

B. Stefanovic et al. / NeuroImage 30 (2006) 726–734 731

narrower range of global CBF perturbations. It is this limitation in

the estimation power that likely precluded a number of earlier

studies (Corfield et al., 2001; Hoge et al., 1999b) from detecting

any effect of baseline flow level on BOLD response magnitude.

A limited effect of the basal vasodilation on the CBF response

is consistent with the majority of current literature (Ramsay et al.,

1993; Friston et al., 1990; Maximillian et al., 1980; Brown et al.,

2003; Li et al., 1999, 2000; Hoge et al., 1999a). This finding

indirectly testifies to the nature of CBF regulation following

functional activation and, as noted earlier (Brown et al., 2003), is

not in agreement with the exhausted cerebrovascular reserve

hypothesis and the oxygen limitation model. It has thus motivated

the development of new models of CBF control (Buxton, 2004;

Behzadi and Liu, 2005). Despite a small effect of basal flow on

CBF response across subjects, it should be stressed that, as

evident from Fig. 4, there was a considerable intersubject

variability, presumably due to the differences in the normocapnic

vascular tone and/or cerebrovascular reactivity. Indeed, in some

subjects, a very pronounced attenuation of the percent CBF

changes with increasing basal flow levels was observed. Finally,

it is likely that a further increase in end-tidal CO2 would result in

a sharp drop in the perfusion response once the maximal

vasodilation has been reached, as suggested earlier (Kemna et

al., 2001). It is of note, however, that while we saw only a small

effect on CBF changes, the BOLD response to activation was not

only diminished across subjects but had reversed sign in at least

one subject. As expected, the attenuation of the BOLD response

was slight in the 5% CO2 condition but pronounced under the 7.5

and 10% CO2 challenges.

Considering the hypercapnia-induced responses, the average

increases in ETCO2 of ¨8/13/21 mm Hg were accompanied by

CBF increases of 20/50/70%, in agreement with 2–11% flow

increase in grey matter per mm Hg rise in end-tidal CO2 (Kemna et

al., 2001; Ramsay et al., 1993). The corresponding increases in

BOLD, of 1.5/2.25/3% in MC and 1.75/3/3.75% in VC, are also in

agreement with the average BOLD-ETCO2 reactivity of 0.1–0.4%

per mm Hg in grey matter (Posse et al., 2001; Rostrup et al., 2000),

though this reactivity is itself a function of ETCO2. Nonetheless, in

this work, as in our earlier studies, the hypercapnia data were rather

noisy.

Indeed, the temporal fluctuations in the hypercapnic responses

(cf. Fig. 2), possibly arising from dynamic changes in subjects’

breathing patterns, motivated us to introduce the reference grey

matter region time course as a confound in the statistical analysis.

It should be stressed that the presence of the interaction between

hypercapnia and functional activation was easily detectable just by

introduction of an interaction term in the analysis (data not shown)

but is not the main objective of the current analysis. Rather, the

reference GM region was introduced to allow robust quantification

of functional activation-induced responses at increasing basal flow

levels. The assumption that the time course of the hypercapnic

response is reasonably uniform (to within a weighting factor)

across GM, exploited in the present analysis, is supported by

hypercapnia reactivity studies (Rostrup et al., 2005) and may also

be observed in the sample data of Fig. 2. While the qualitative

nature of the results is not affected by the use of this confound

(BOLD responses are still markedly attenuated without it; CBF

responses are slightly diminished), the parameter estimates do

change: notably, the contribution of the hypercapnic response to

the total response during the GHC + VS/MT periods is typically

underestimated when the temporal instability in the hypercapnic

response is ignored. In other words, less pronounced decreases of

activation-induced BOLD and CBF changes with increasing basal

flow are observed when the temporal variation in the global

hypercapnic response is not accounted for.

A possible, albeit difficult, improvement on the present

methods could be the use of volume-controlled mechanical

ventilation. This would presumably reduce the temporal fluctua-

tions in the hypercapnia data and produce higher basal flow

changes with the same inspired CO2 mixtures, as it would preclude

an adjustment in the tidal volume during CO2 inhalation and

minimize respiratory dead space. However, extensive training of

the subjects would likely be required to avoid the confounds

associated with the perception of air hunger (Banzett et al., 2000;

Evans et al., 2002). While none of our subjects reported any

difficulty breathing, almost all of them could distinguish both 7.5

and 10% CO2 mixtures from medical air and 5% CO2 mixtures (the

last two being indiscernible by most subjects). It is thus not clear

whether basal flow may be upregulated far enough to null the

activation-induced CBF response without producing dyspnea.

Finally, it is important to stress that careful steps were taken in

the present work to minimize the bias in the ASL estimation of

CBF for high global flow values. In particular, the QUIPSS II

method is used to minimize errors due to spatial variation in transit

delays. A very wide labeling region (of 150 mm thickness) was

employed with short inversion times, TI1 and TI2 of 500 and 1100

ms, respectively, to ensure that the temporal width of the tag would

still be well defined (i.e., there would be no shortage of labeled

blood) even for highly elevated global CBF.

The present findings lend further support to the caveat that

the differences in global perfusion must be accounted for before

any conclusion regarding the activation- or region-specific

BOLD response differences may be drawn. Obviously, the

greater the global CBF perturbation or the smaller the BOLD

difference to be investigated, the more important this consider-

ation becomes. This is particularly significant for fMRI studies

of stroke recovery as well as pharmacological MRI studies

involving administration of vasoactive drugs that can have a

profound effect on the resting perfusion. Finally, the sharp

decrease in the BOLD response amplitude and its eventual

reversal at high global flow levels is also indicative of a

hemodynamic scenario that results in functionally active regions

being ‘‘silent’’ on BOLD fMRI maps.

Conclusion

We observed a significant drop of the activation-induced BOLD

fMRI signal change with increasing baseline global perfusion,

consistent with the prediction of BOLD signal models that a

significant decrease in the basal venous deoxyhemoglobin con-

centration at high global flow levels curbs the dynamic range of the

BOLD response, thus lowering its sensitivity to further CBF

increases. A limited effect of the basal vasodilation on the relative

CBF response was found, in accordance with the majority of earlier

reports. Such small changes in the absolute CBF response, in spite

of large baseline flow variation, indirectly testify to the nature of

CBF regulation following functional activation. The current

findings characterize the nonlinear regime of the deoxyhemoglobin

dilution model and call for quantification of basal CBF in BOLD

fMRI studies that involve subjects and/or conditions with markedly

elevated global perfusion.

B. Stefanovic et al. / NeuroImage 30 (2006) 726–734732

Acknowledgments

This work was supported by the Natural Sciences and

Engineering Research Council of Canada and the Canadian

Institutes of Health Research.

Appendix A

A.1. CMRO2estimation

The present hypercapnia data were also used to calibrate the

deoxyhemoglobin dilution model (Davis et al., 1998; Hoge et

al., 1999a) and estimate the CMRO2 changes induced by

functional stimulation, assuming no effect of the hypercapnic

challenges on the CMRO2 response (Grubb et al., 1974; Artru

and Michenfelder, 1980), as described earlier (Stefanovic et al.,

2004). The across-subject estimate of maximal BOLD signal

change were 0.061 T 0.011 (or DR2* of �1.2 T 0.2 s�1,

Fig. 5. The total percent changes in BOLD and CBF signals (MC ROI: a, VC RO

hypercapnic mixture in each subject. The across subject averages of hypercap

corresponding fit, representing the baseline iso-CMRO2 contour and providing

estimated M was substituted into the Eq. [13] of the deoxyhemoglobin dilution mod

intervals. The optimal linear fit to the activation-induced CMRO2responses versus

ROI: b, VC ROI: d). The shaded region represents the standard error in the linea

assuming mono-exponential decay) in the motor ROI and 0.076 T0.013 (DR2* of �1.5 T 0.2 s�1) in the visual cortex. The former

result is in excellent agreement with our previous findings in

primary motor cortices of healthy adult volunteers of 7.2 T 1.0%

or DR2* of �1.4 T 0.2 s�1 (Stefanovic et al., 2004). It is also in

reasonable agreement with the value of 9 T 3%, i.e., DR2* of

�2.2 T 0.7 s�1, for bilateral finger tapping in healthy volunteers

reported by Kastrup et al. (2002).

Figs. 5a, c displays the measured total (i.e., GHC + VS/MT-

induced) BOLD versus total CBF signal changes and the calculated

iso-CMRO2contours in the two ROIs. The estimated MT/VS-

induced CMRO2changes and the corresponding MT/VS-induced

CBF responses, for each subject, are shown in Figs. 5b, d. The

slope of the straight line fit to these data yielded a CMRO2/CBF

coupling ratio of 0.49 T 0.13 for MC ROI and 0.22 T 0.11 for VC

ROI. The former value agrees very well with the earlier estimates,

in the MC ROI, of 0.44 T 0.04 (Stefanovic et al., 2004) and 0.33 T0.06 (Kastrup et al., 2002). On the other hand, the slope presently

obtained in the visual cortex is significantly lower than these MC

I: c) induced by high level functional activation in combination with air or

nia-induced signal changes (black squares) are displayed along with the

the estimate of the maximum achievable BOLD signal change (M). The

el (Hoge et al., 1999a) to generate nonbaseline iso-CMRO2contours, at 10%

the activation-elicited CBF changes are displayed in the right column (MC

r fit.

B. Stefanovic et al. / NeuroImage 30 (2006) 726–734 733

values and is closer to the 0.34 VC ROI ratio observed in earlier

GHC + VS experiments in our laboratory (Hoge et al., 1999a).

It should be noted that the present hypercapnia data are rather

noisy (as in our earlier studies), precluding a robust per subject M

estimation. Had a higher SNR of hypercapnia data been achieved,

individual M estimates may well have shed light on some of the

intersubject variability observed. In particular, as M is proportional

to the maximal achievable BOLD response, a lower M may

correspond to a more pronounced decrease of BOLD’s sensitivity

to CBF changes for the same increase in the basal flow level. On

the other hand, a higher normocapnic venous blood volume may

correlate with a more pronounced effect of the given basal flow

increase on the relative CBF response.

References

Artru, A., Michenfelder, J., 1980. Effects of hypercarbia on canine cerebral

metabolism and blood flow with simultaneous direct and indirect

measurement of blood flow. Anesthesiology 52 (6), 466–469.

Bandettini, P.A., Wong, E.C., 1997. A hypercapnia-based normalization

method for improved spatial localization of human brain activation with

fMRI. NMR Biomed. 10 (4–5), 197–203.

Banzett, R., Mulnier, H., Murphy, K., Rosen, S., Wise, R., Adams, L., 2000.

Breathlessness in humans activates insular cortex. NeuroReport 11 (10),

2117–2120.

Behzadi, Y., Liu, T., 2005. An arteriolar compliance model of the

cerebral blood flow response to neural stimulus. NeuroImage 25 (4),

1100–1111.

Bezdek, J.C., Hall, L.O., Clarke, L.P., 1993. Review of MR image

segmentation techniques using pattern recognition. Med. Phys. 20 (4),

1033–1048.

Brown, G., Eyler-Zorrilla, L., Georgy, B., Kindermann, S., Wong, E.,

Buxton, R., 2003. BOLD and perfusion response to finger-thumb

apposition after acetazolamide administration: differential relationship

to global perfusion. J. Cereb. Blood Flow Metab. 23 (7), 829–837.

Bruhn, H., Kleinschmidt, A., Boecker, H., Merboldt, K., Hanicke, W.,

Frahm, J., 1994. The effect of acetazolamide on regional cerebral blood

oxygenation at rest and under stimulation as assessed by MRI. J. Cereb.

Blood Flow Metab. 14 (5), 742–748.

Bruhn, H., Fransson, P., Frahm, J., 2001. Modulation of cerebral blood

oxygenation by indomethacin: MRI at rest and functional brain

activation. J. Magn. Reson. Imaging 13 (3), 325–334.

Buxton, R., 2004. A hypothesis for cerebral blood flow regulation and the

origin of the BOLD effect. Proceedings of the Twelfth Annual Meeting

of the ISMRM. Kyoto, Japan, p. 273.

Cohen, E., Ugurbil, K., Kim, S.-G., 2002. Effect of basal conditions on the

magnitude and dynamics of the blood oxygenation level-dependent

fMRI response. J. Cereb. Blood Flow Metab. 22, 1042–1053.

Corfield, D., Murphy, K., Josephs, O., Adams, L., Turner, R., 2001. Does

hypercapnia-induced cerebral vasodilation modulate the hemodynamic

response to neural activation? NeuroImage 13 (6), 1207–1211.

Cox, R.J., 1996. AFNI: Software for analysis and visualization of functional

magnetic resonance neuroimages. Comput. Biomed. Res. 29, 162–173.

Davis, T., Kwong, K., Weisskoff, R., Rosen, B., 1998. Calibrated functional

MRI: Mapping the dynamics of oxidative metabolism. Proc. Natl. Acad.

Sci. U. S. A. 95, 1834–1839.

Evans, K., Banzett, R., Adams, L., McKay, L., Frackowiak, R., Corfield,

D., 2002. BOLD fMRI identifies limbic, paralimbic, and cerebellar

activation during air hunger. J. Neurophysiol. 88 (3), 1500–1511.

Friston, K., Frith, C., Liddle, P., Dolan, R., Lammertsma, A., Frackowiak,

R., 1990. The relationship between global and local changes in PET

scans. J. Cereb. Blood Flow Metab. 10 (4), 458–466.

Gollub, R., Breiter, H., Kantor, H., Kennedy, D., Gastfriend, D., Mathew,

R., Makris, N., Guimaraes, A., Riorden, J., Campbell, T., Foley, M.,

Hyman, S., Rosen, B., Weisskoff, R., 1998. Cocaine decreases cortical

cerebral blood flow but does not obscure regional activation in

functional magnetic resonance imaging in human subjects. J. Cereb.

Blood Flow Metab. 18 (7), 724–734.

Grubb, R., Phelps, M., Eichling, J., 1974. The effects of vascular changes in

PaCO2on cerebral blood volume, blood flow and vascular mean transit

time. Stroke 5, 630–639.

Hoge, R.D., Atkinson, J., Gill, B., Crelier, G., Marrett, S., Pike, G.,

1999a. Investigation of BOLD signal dependence on CBF and

CMRO2: the deoxyhemoglobin dilution model. Magn. Reson. Med.

42 (5), 849–863.

Hoge, R.D., Atkinson, J., Gill, B., Crelier, G.R., Marrett, S., Pike, G.B.,

1999b. Additive combination of perfusion responses to hypercapnia and

visual stimulation. Proceeding of the Fifth International Conference on

Functional Mapping of the Human Brain. Dusseldorf, FRGNeuroImage

vol. 9 (6), p. s306.

Kastrup, A., Krueger, G., Neumann-Haefelin, T., Glover, G., Moseley,

M.E., 2002. Changes of cerebral blood flow, oxygenation and oxidative

metabolism during graded motor activation. NeuroImage 15, 74–82.

Kemna, L., Posse, S., Tellmann, L., Schmitz, T., Herzog, H., 2001.

Interdependence of regional and global cerebral blood flow during

visual stimulation: an O-15-butanol positron emission tomography

study. J. Cereb. Blood Flow Metab. 21 (6), 664–670.

Laurienti, P., Field, A., Burdette, J., Maldjian, J., Yen, Y., Moody, D., 2003.

Relationship between caffeine-induced changes in resting cerebral

perfusion and blood oxygenation level-dependent signal. Am. J.

Neuroradiol. 24 (8), 1607–1611.

Li, T., Moseley, M., Glover, G., 1999. A FAIR study of motor cortex

activation under normo- and hypercapnia induced by breath challenge.

NeuroImage 10 (5), 562–569.

Li, T., Kastrup, A., Moseley, M., Glover, G., 2000. Changes in baseline

cerebral blood flow in humans do not influence regional cerebral blood

flow response to photic stimulation. J. Magn. Reson. Imaging 12 (5),

757–762.

Liu, T., Behzadi, Y., Restom, K., Uludag, K., Lu, K., Buracas, G.,

Dubowitz, D., Buxton, R., 2004. Caffeine alters the temporal dynamics

of the visual bold response. NeuroImage 23 (4), 1402–1413.

Maximillian, V., Prohovnik, I., Risberg, J., 1980. Cerebral hemodynamic

response to mental activation in normo- and hypercapnia. Stroke 11,

342–347.

Mulderink, T., Gitelman, D., Mesulam, M., Parrish, T., 2002. On the use of

caffeine as a contrast booster for BOLD fMRI studies. NeuroImage 15

(1), 37–44.

Posse, S., Kemna, L., Elghahwagi, B., Wiese, S., Kiselev, V., 2001. Effect

of graded hypo- and hypercapnia on fMRI contrast in visual cortex:

quantification of T2* changes by multiecho EPI. Magn. Reson. Med. 46

(2), 264–271.

Poulin, M., Liang, P., Robbins, P., 1996. Dynamics of the cerebral blood

flow response to step changes in end-tidal PCO2and PO2

in humans.

J. Appl. Physiol. 81 (3), 1084–1095.

Press, W.H., Teukolsky, S.A., Vetterling, W.T., Flannery, B.P., 1992.

Numerical Recipes in C: The Art of Scientific Computing, 2nd edRCambridge Univ. Press, New York, p. 666–670R Chap. 15.

Ramsay, S., Murphy, K., Shea, S., Friston, K., Lammertsma, A., Clark, J.,

Adams, L., Guz, A., Frackowiak, R., 1993. Changes in global cerebral

blood flow in humans: effect on regional cerebral blood flow during a

neural activation task. J. Physiol. 471, 521–534.

Rostrup, E., Law, I., Blinkenberg, M., Larsson, H., Born, A., Holm, S.,

Paulson, O., 2000. Regional differences in the CBF and BOLD

responses to hypercapnia: a combined PET and fMRI study. Neuro-

Image 11 (2), 87–97.

Rostrup, E., Knudsen, G., Law, I., Holm, S., Larsson, H., Paulson, O., 2005.

The relationship between cerebral blood flow and volume in humans.

NeuroImage 24 (1), 1–11.

Shimosegawa, E., Kanno, I., Hatazawa, J., Fujita, H., Ito, H., Shuichi, M.,

Matsutarou, M., Inuga-mi, A., Ogawa, T., Itoh, H., Okudera, T.,

Uemura, K., 1995. Photic stimulation study of changing the arterial

B. Stefanovic et al. / NeuroImage 30 (2006) 726–734734

partial pressure level of carbon dioxide. J. Cereb. Blood Flow Metab.

15, 111–114.

Stefanovic, B., Sled, J., Pike, G., 2003. Quantitative T2 in the occipital lobe:

the role of the CPMG refocusing rate. J. Magn. Reson. Imaging 18 (3),

302–309.

Stefanovic, B., Warnking, J.M., Pike, G.B., 2004. Hemodynamic and

metabolic responses to neuronal inhibition. NeuroImage 22 (2),

771–778.

Warnking, J.M., Pike, G.B., 2004. Bandwidth-modulated adiabatic RF

pulses for uniform selective saturation and inversion. Magn. Reson.

Med. 52 (5), 1190–1199.

Weckesser, M., Posse, S., Olthoff, U., Kemna, L., Dager, S., Muller-

Gartner, H., 1999. Functional imaging of the visual cortex with BOLD-

contrast MRI: hyperventilation decreases signal response. Magn. Reson.

Med. 41 (1), 213–216.

Wong, E., Buxton, R., Frank, L., 1997. Implementation of quantitative

perfusion imaging techniques for functional brain mapping using pulsed

arterial spin labeling. NMR Biomed. 10, 237–249.

Worsley, K., Liao, C., Aston, J., Petre, V., Duncan, G., Morales, F., Evans,

A., 2002. A general statistical analysis for fMRI data. NeuroImage 15,

1–15.

Ye, F., Frank, J., Weinberger, D., McLaughlin, A., 2000. Noise reduction in

3D perfusion imaging by attenuating the static signal in arterial spin

tagging (ASSIST). Magn. Reson. Med. 44 (1), 92–100.