manipulation of cortical gray matter oxygenation by hyperoxic respiratory challenge: field...

Manipulation of cortical gray matteroxygenation by hyperoxic respiratorychallenge: field dependence of R2* and MRsignal responseCristina Rossia*, Andreas Bossa, Olivio F. Donatia, Roger Luechingerb,Spyridon S. Kolliasc, Antonios Valavanisc, Juerg Hodlera and Daniel Nanza

The aim of this study was to quantitatively assess the field strength dependence of the transverse relaxation rate (R2*)change in cortical gray matter induced by hyperoxia and hyperoxic hypercapnia versus normoxia in an intra-individualcomparison of young healthy volunteers. Medical air (21% O2), pure oxygen and carbogen (95% O2, 5% CO2) were al-ternatively administered in a block-design temporal pattern to induce normoxia, hyperoxia and hyperoxic hypercapnia,respectively. Local R2* values were determined from three-dimensional, multiple, radiofrequency-spoiled, fast fieldecho data acquired at 1.5, 3 and 7T. Image quality was good at all field strengths. Under normoxia, the mean graymat-ter R2* values were 13.3� 2.7 s–1 (1.5T), 16.9�0.9 s–1 (3 T) and 29.0�2.6 s–1 (7 T). Both hyperoxic gases induced relax-ation rate decreases ΔR2*, whosemagnitudes increased quadratically with the field strength [carbogen: –0.69� 0.20 s–1

(1.5T), –1.49�0.49 s–1 (3 T), –5.64� 0.67 s–1 (7 T); oxygen: –0.39� 0.20 s–1 (1.5T), –0.78�0.48 s–1 (3 T), –3.86�1.00 s–1

(7T)]. Carbogen produced larger R2* changes than oxygen at all field strengths. The relative change ΔR2*/R2* also in-creasedwith the field strengthwith a power between 1 and 2 for both carbogen and oxygen. The statistical significanceof the R2* response improvedwith increasing B0 andwas higher for carbogen than for oxygen. For a sequence with pureT2* weighting of the signal response to respiratory challenge, the results suggested a maximum carbogen-inducedsignal difference of 19.3% of the baseline signal at 7 T and TE=38ms, but amaximum oxygen-induced signal differenceof only 3.0% at 1.5 T and TE=76ms. For 3T, maximum signal changes of 4.7% (oxygen) and 8.9% (carbogen) were com-puted. In conclusion, the R2* response to hyperoxic respiratory challenge was stronger for carbogen than for oxygen,and increased quadratically with the staticmagnetic field strength for both challenges, which highlights the importanceof high field strengths for future studies aimed at probing oxygen physiology in clinical settings. Copyright © 2012 JohnWiley & Sons, Ltd.

Keywords: brain oxygenation; R2* quantification; hyperoxia; hyperoxic hypercapnia; carbogen; high-field MRI

INTRODUCTION

The tissue response to changes in oxygen supply may provideinformation on the mechanisms underlying oxygen deliveryand consumption. A quantitative assessment may shed newlight on the pathophysiology of cerebrovascular diseases (1,2)and support the interpretation of functional neuroimaging sig-nals (3). The oxygen supply can be manipulated and controlledby respiratory challenges, e.g. inhalation of hyperoxic gases.One option for measuring the response to such a challenge isthe quantification of changes in the relaxation rate (R2* = 1/T2*)of the transverse magnetization of brain tissue hydrogen nuclei(4). As a result of its paramagnetic nature, changes in the amountof deoxyhemoglobin (dHb) in the capillary bed affect the localmagnetic susceptibility of the tissue, which alters R2* and, in turn,both the signal intensity and contrast of MR images acquiredduring the challenge [blood oxygenation level-dependent(BOLD) effect] (5).Monitoring of the response of the MR signal to respiratory

challenges may have a clinical impact on the development ofnew strategies for the treatment of radiotherapy-resistanttumors (6). Moreover, as alterations of cerebral tissue reactivity

may occur in the case of severe brain injuries (7), the monitor-ing of the response of the tissue to a vasoactive stimulus(e.g. CO2) may provide new insights into the pathologicalmechanisms of these diseases and support the clinical manage-ment of patients.

* Correspondence to: C. Rossi, University Hospital of Zurich, Department of Diag-nostic and Interventional Radiology, Rämistrasse 100, CH-8091 Zurich, Switzerland.E-mail: [email protected]

a C. Rossi, A. Boss, O. F. Donati, J. Hodler, D. NanzDepartment of Diagnostic and Interventional Radiology, University Hospital ofZurich, Zurich, Switzerland

b R. LuechingerInstitute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland

c S. S. Kollias, A. ValavanisDepartment of Neuroradiology, University Hospital of Zurich, Zurich, Switzerland

Abbreviations used: BOLD, blood oxygenation level-dependent; CB, carbogen;CBF, cerebral blood flow; dHb, deoxyhemoglobin; FA, flip angle; R2*, effectivetransverse relaxation rate constant; ROI, region of interest; T1-FFE, T1-weightedfast field echo; T2*, effective transverse relaxation time constant.

Research Article

Received: 28 September 2011, Revised: 14 November 2011, Accepted: 12 December 2011, Published online in Wiley Online Library: 2012

(wileyonlinelibrary.com) DOI: 10.1002/nbm.2775

NMR Biomed. (2012) Copyright © 2012 John Wiley & Sons, Ltd.

The attractive potential of the MR-based monitoring of tissueoxygenation is limited, among other things, by the small size(on the order of 3% in gray matter) of the effect induced bythe breathing of hyperoxic gas mixtures on BOLD signals mea-sured at 1.5 T (8–11). The BOLD contrast has been reported tobenefit from increasing static magnetic field, because of signalgain and the enhancement of susceptibility-related effects (12).High-field MRI may help to increase the magnitude of the BOLDeffect and may allow the investigation of healthy and impairedoxygen physiology using a highly resolved and noninvasivetechnique.

This study aims to quantitatively assess the field strengthdependence of BOLD signal transverse relaxation rate changesin cortical gray matter induced by hyperoxia and hyperoxichypercapnia versus normoxia in young healthy volunteers.

MATERIALS AND METHODS

Subjects

Ten young healthy volunteers (mean age, 25� 4 years; six men)without a history of respiratory, cardiovascular or neurologicaldisease were enrolled in the study. Subjects gave written in-formed consent to the MR examination and the scientific evalu-ation of the datasets. The study was approved by the local ethicscommittee. One volunteer (man, 20 years) completed only oneof the three measurement sessions and, for this reason, wasexcluded from the study.

MR protocol

Data were acquired on three MR scanners operating at fieldstrengths of 1.5, 3 and 7 T (Philips Achieva; Philips MedicalSystems, Best, the Netherlands). At 1.5 and 3 T, the signals werereceived via eight-channel head coils (Philips Healthcare, Best,the Netherlands). At 7 T, a 16-channel receive–volume transmitcoil (Nova Medical, Wilmington, MA, USA) was used. To minimizemotion, foam padding was positioned between the coil formerand the subjects’ heads. In each of the three measurementsessions, the MR protocol started with a radiofrequency-spoiled,gradient recalled echo sequence to provide localizer images andan axially oriented, T1-weighted, turbo spin echo sequence for an-atomical reference. Subsequently, a three-dimensional, multi-echo,radiofrequency-spoiled, gradient echo sequence (T1-weighted fastfield echo, T1-FFE) sequence for R2* quantification was repeatedlyscanned before, during and after the respiratory challenges. Ateach of the three field strengths, the flip angle (FA) and TR wereadjusted in a trade-off between T1 weighting and MR signal(1.5 T: FA=60� , TR= 113ms; 3 T: FA=50� , TR= 93ms; 7 T: FA=45�,TR= 46ms). The range of TEs acquired for the computation of thetransverse relaxation rate was adjusted in a range around theexpected gray matter T2* (1.5 T: TE= 12, 30, 48, 66, 84, 102ms;3 T: TE= 8, 24, 40, 56, 72, 88ms; 7 T: TE= 3, 11, 19, 27, 35, 43ms).Readout bandwidth values were 170Hz (1.5 T), 215Hz (3 T) and400Hz (7 T). The voxel size was 0.5� 0.5� 1.0mm3 at all fieldstrengths. Flow sensitization via T1 contrast was chosen to helpwith the identification of the macrovasculature (13), whereas de-tection of smaller venous vessels was possible on images acquiredwith long TEs. Transverse slabs were placed superior to the lateralventricle and oriented parallel to the nasal cavity (Fig. 1a). The totalacquisition time and positioning of the subject resulted in a totalexamination time of c. 35min per individual and field strength.

Breathing system and gas administration protocol

R2* quantification was performed during the inhalation of medicalair (i.e. 21% O2), 100% oxygen (O2) and carbogen (5% CO2+95%O2). We refer to the situations induced by medical air, oxygenand carbogen by the terms ‘normoxia’, ‘hyperoxia’ and ‘hyperoxichypercapnia’, respectively. The gases were administered in blocksof 6min each in the sequence: medical air, oxygen, carbogen andmedical air. Breathing gas from a tank was administered through amask with a one-way valve and a 0.5-L reservoir bag. The maskwas adjusted as tight as possible, and the subjects were requestedto breathe normally. The gas flow rate was set to 8 L/min.

MR data processing

MR images were processed off-line using routines written inMatlab (Matlab; The MathWorks, Natick, MA, USA). In order to ac-count for the response of the MR signal to the respiratory chal-lenge (14), the following datasets were selected: (i) the first datasetacquired during the inhalation of medical air (baseline); (ii) the lastdataset acquired during the administration of 100% O2 (O2); and(iii) the last dataset acquired during carbogen breathing (CB).For each subject and each measurement session, a four-step

scheme was applied for data processing and evaluation.

Co-registration

At each of the three field strengths, the images acquired using theminimum TE during hyperoxia and hyperoxic hypercapnia wereco-registered to the corresponding image of the baseline dataset.Rigid body, two-dimensional, image co-registrationwas performedusing functions in the Matlab Image Processing Toolbox. Initially, aset of reference points was defined in both images using the‘cpselect’ command. Successively, the ‘cp2tform’ function was ap-plied to estimate the parameters of the transformation neededfor optimal image co-registration. A ‘nonreflective similarity’ trans-formation was selected to allow image rotation and translation,but not image reshaping. Finally, the ‘imtransform’ function wasused to apply the spatial transformation to the whole image. Thetransformation matrices computed for the image acquired usingthe minimum TE were then applied to all O2 and CB images.

R2* mapping

Parametric R2* mapswere computed for each dataset on a pixel-by-pixel basis by fitting the MR signal to the expression:

S tð Þ ¼ S0 exp �R�2�t� �

[1]

The fitting was performed using an algorithm based on theinterior-reflective Newton method (15). Fitting variables were S0and R2*.

Region of interest (ROI) analysis

Local R2* values were assessed in ROIs drawn over the gray matterfor each of the three field strengths. For the definition of commonregions and the segmentation of gray and white matter, ROIs weredrawn over the T1-FFE image acquired with minimum TE duringmedical air breathing and overlaid onto the multi-echo imagesacquired during different stages of the breathing paradigm. LocalR2* values resulted from the mono-exponential fit of the meansignal measured over the ROI. For each dataset, the cortical gray

C. ROSSI ET AL.

wileyonlinelibrary.com/journal/nbm Copyright © 2012 John Wiley & Sons, Ltd. NMR Biomed. (2012)

matter R2* value was obtained as the average of values measuredin six ROIs. Small ROIs (minimum size, 9 voxels) were carefullydrawn to avoid large vessels and regions of obvious susceptibilitygradients (Fig. 1b).

Statistical analysis

In each subject, R2* values obtained under respiratory challengewere statistically compared with the R2* values measured undernormoxia using a paired, single-tailed, Student’s t-test (p< 0.05).

MR signal response to hyperoxic respiratory challenge

To estimate the magnitude of the optimal TE that would lead to amaximum MR signal response in a single-echo experiment, the MRsignal change induced by hyperoxic respiratory challenge was com-puted for a sequence with pure T2* weighting as a function of TE (t).For each of the three field strengths, MR signal changes were

computed from the expression (16):

ΔS ¼ S0�e�R�2;0�t e� R�2;G�R�2;0ð Þ�t � 1� �

[2]

where S0 is the signal at t= 0, and R�2;0 and R�2;G indicate the meanrelaxation rates assessed in the cohort of volunteers undernormoxia and during inhalation of the hyperoxic gas mixture,respectively.

RESULTS

The results are summarized in Figs 2–5 and Tables 1 and 2. Goodimage quality was obtained at 1.5, 3 and 7 T (Fig. 2). ParametricR2* maps did not seem to be affected detrimentally by geomet-rical distortions or signal voids associated with local magneticfield inhomogeneities, even at 7 T (Fig. 3). In one volunteer(man, 22 years), images acquired at both 1.5 and 3 T could notbe adequately corrected for motion. For this reason, none ofthe datasets acquired from this volunteer (including the mea-surements performed at 7 T) were considered in the evaluation.Under normoxia, mean R2* values of 13.3� 2.7 s–1 (1.5 T),

16.9� 0.9 s–1 (3 T) and 29.0� 2.6 s–1 (7 T) were measured in corti-cal gray matter. Breathing of hyperoxic gas mixtures resulted in adecrease in the relaxation rate that was larger at higher fieldstrengths (Table 1).

At each of the three field strengths, carbogen breathingproduced larger changes in R2* (ΔR2*) than the inhalation ofpure oxygen (hyperoxic hypercapnia: –0.69� 0.20 s–1 at 1.5 T,–1.49� 0.49 s–1 at 3 T, –5.64� 0.67 s–1 at 7 T; hyperoxia: –0.39� 0.20 s–1 at 1.5 T, –0.78� 0.48 s–1 at 3 T, –3.86� 1.00 s–1

at 7 T). The magnitude of ΔR2* increased quadratically withB0 for both hyperoxic challenges, but more rapidly underhyperoxic hypercapnia than under hyperoxia (Fig. 4a).

Under hyperoxic hypercapnia, the relative R2* response(ΔR2*/R2*) showed a fairly linear dependence on the strengthof the static magnetic field (B0), whereas breathing of pureoxygen seemed to lead to a hyperlinear increase in the R2*response with B0 (hyperoxic hypercapnia: –5.3� 1.7% at 1.5 T,–8.7� 2.8% at 3 T, –19.7� 2.2% at 7 T; hyperoxia: –3.2� 1.9%at 1.5 T, –4.6� 2.9% at 3 T, –13.6� 4.0% at 7 T). A linear regres-sion was applied to the data to place the slopes of –ΔR2*/R2*versus field strength in relation. A slope ratio of 1.34: 1.00(hyperoxic hypercapnia: hyperoxia) was found (Fig. 4b).

The statistical significance of the R2* response improved withincreasing B0 and was higher under hyperoxic hypercapnia thanunder hyperoxia (Table 1).

To gauge the influence of inflow effects on the gray matter sig-nal changes, the intercept at TE= 0ms of the MR signal (S0) wascomputed by exponential fitting of the multi-echo data acquiredfor each subject in all three measurement sessions (16). Signaldeviations on the order of a few per cent were computed duringrespiratory challenges (hyperoxic hypercapnia: 0.12� 6.70% at1.5 T, –4.56� 3.41% at 3 T, –3.15� 3.45% at 7 T; hyperoxia:1.04� 5.03% at 1.5 T, –1.77� 3.92% at 3 T, 2.73� 3.57% at 7 T) withno clear trends.

The measured ΔR2* values imply that the TEs at which a max-imum challenge-induced MR signal difference is expected for asequence with pure T2* weighting of the image contrast de-crease with increasing field strength, and are marginally longerfor carbogen than for oxygen breathing (Table 2 and Fig. 5).

DISCUSSION

Cortical gray matter R2* values under normoxia (baseline)

The R2* relaxation rates computed in this study over the corticalgray matter under normoxia (i.e. during medical air breathing)present an optimal accordance with the values reported by

Figure 1. (a) Position and orientation of the oblique transverse slab acquired for R2* quantification. (b) Regions of interest (ROIs) manually drawn overthe T1-weighted fast field echo (T1-FFE) image acquired with minimum TE (TE= 3ms; TR = 46ms; flip angle, 45�) during medical air breathing for onehealthy volunteer at 7 T.

FIELD DEPENDENCE OF R2* RESPONSE OF CORTICAL GRAY MATTER TO HYPEROXIA

NMR Biomed. (2012) Copyright © 2012 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/nbm

Peters et al. (18) after correction for through-slice, intravoxel,dephasing field gradients. Uludag et al. (19) proposed a linear ex-pression for the dependence of the R2* values of cerebral tissueon the strength of the main magnetic field. In comparison withthe relaxivity rates predicted by this expression for 1.5, 3 and7 T, the R2* values reported in this work are slightly, but consis-tently, lower. Although, in our study, no correction for through-slice, intravoxel, dephasing gradients was implemented, boththe accordance with the values reported by Peters et al. (18)and the slightly lower magnitude of the values in comparisonwith the rates of relaxation reported in the literature (19) suggestthat no systematic errors that would increase with the fieldstrength should be expected over our uncorrected R2* values.The good agreement of our rates of relaxation with the cor-rected R2* values of the study by Peters et al. (18) may be

mainly a result of the significantly smaller voxel size in ourstudy (0.25mm3 versus 13.35mm3), and thus a significantlysmaller intravoxel frequency dispersion than in their studyand others (20,21). The volume-targeted homogenization ofthe static field applied before data acquisition in this study, aswell as the careful selection of cortical gray matter ROIs, withminimal apparent susceptibility gradients and the absence ofany indications of arterial or venous vessels in the immediatevicinity, may have also contributed to the decrease in intravoxelfrequency dispersion. The good agreement with the datareported by Peters et al. (18) suggests that the use of highlyresolved three-dimensional sequences for the quantificationof R2* values may help to overcome the problem of thethrough-slice dephasing which more strongly affects two-dimensional datasets.

Figure 2. Parametric R2* maps (to scale) from data of a single individual acquired at 1.5, 3 and 7 T. R2* increased with increasing field strength. Lowerrelaxation rates and blood oxygenation level-dependent (BOLD) signals were measured in gray matter relative to white matter. There was a relativelylarge regional heterogeneity in the relaxivity of white matter, which is well known and is a topic of active research (17).

Figure 3. Brain R2* relaxation rate maps obtained from a healthy volunteer at 7 T under normoxia (a), hyperoxia (b) and hyperoxic hypercapnia (c),together with a three-dimensional T1-weighted fast field echo (T1-FFE) source image (three-dimensional T1-FFE sequence; flip angle, 45� ; TR = 46ms;TE = 3ms; voxel size, 0.25mm3) (d). Maps of the magnitude of the relative R2* response (�ΔR2*/R2*) for both hyperoxia (e) and hyperoxic hypercapnia(f) are also shown. The arrow highlights a gray matter area. Hyperoxic respiratory challenge generally leads to a decrease in the gray matter relaxivitythat is larger during carbogen breathing (c, f) than during the inhalation of pure oxygen (b, e).

C. ROSSI ET AL.


Effects of respiratory challenges

In addition to acting on the local dHb concentration, the breath-ing of hyperoxic gas mixtures, which contain a vasoactive agent,induces physiological changes, such as altered magnitudes ofregional cerebral blood flow (CBF) and volume (22). Carbogenbreathing has been reported to induce an increase of up to50% in CBF (23), whereas Bulte et al. (24) measured a decreaseof less than 10% in CBF during the breathing of pure oxygen.The physiological changes discussed could affect the signal in

our three-dimensional, radiofrequency-spoiled, FFE acquisitionvia several mechanisms. Variation in blood flow could alter theapparent R1 relaxation rates, as a result of changed inflow effects,as well as the apparent R2* relaxation, as a result of changed

degrees of intravoxel dephasing and signal attenuation causedby flow or the convection of water molecules. Blood volumechanges could alter voxel-wide averages of R1 and R2* relaxationrates and observable proton densities because of differences inthese values for intravascular or extravascular water, and ofchanges in voxel-wide average diffusivity. In our study, the esti-mation of the intercept of the signal at TE = 0ms showed thatthe influence of inflow effects on the gray matter signal changesshould be negligible. Nevertheless, a residual contribution ofinflow to the estimated R2* relaxivity cannot be excluded. How-ever, the computation of R2*, rather than the estimation of thesignal changes via single-echo measurements, allows the re-moval of the contribution of the apparent R1 value to theresponse of the MR signal to the challenge (25). The possibleinfluence of the inflowing spins and of the paramagnetic natureof the dissolved oxygen on the longitudinal relaxation rate of theMR signal should be especially accounted for when performing asingle-echo experiment with gas mixtures containing a vasoac-tive agent (10).

In this study, a set of multi-echo MR images was acquired forthe pixel-wise computation of the relaxation rate R2* of the MRsignal (26). This strategy was selected for three main reasons:(i) local R2* values are expected to react sensitively to theabove-mentioned physiological changes (20); (ii) they are rela-tively straightforward to quantify; and (iii) they represent a possi-ble way to overcome the unsolved problem of the meaningfulquantification of the BOLD effect (27).

Challenge-induced R2*changes

At all field strengths, both respiratory challenges, i.e. hyperoxia(100% O2 breathing) and hyperoxic hypercapnia (carbogenbreathing), were found to induce a decrease in the R2* relaxationrate of cortical gray matter in the equilibrium phase of the chal-lenge (Table 1). Although the dependence of the R2* values onthe imaging strategy (e.g. voxel size) may hinder direct compari-son with the results reported in the literature (28), the magnitudeof the observed respiratory-induced R2* changes compares wellwith the values reported elsewhere. Several studies have per-formed a single-echo-based evaluation of the changes in theMR signal during respiratory challenges (8,29,30). In thesestudies, an estimation of the maximum achievable BOLD contrast

Figure 4. Magnitude of the respiratory challenge-induced ΔR2* and ΔR2*/R2* gray matter response as a function of the static magnetic field strengthB0. Mean data from all volunteers at each of the three field strengths are shown with error bars that represent the standard deviation of the mean. Theincrease in the ΔR2* magnitude with the field strength in (a) was well fitted with the quadratic curve aB20 þ bB0 þ g for both oxygen (a=0.093 s–1 T–2;b=�0.157 s–1 T–1; g=0.417 s–1; norm of residuals, 2.68� 10–15) and carbogen (a=0.092 s–1 T–2; b=0.121 s–1 T–1; g=0.303 s–1; norm of residuals, 2.99� 10–15).The magnitude of the relative change (ΔR2*/R2*) increased with the field strength with an order slightly larger than unity for both challenges (b). Alinear regression applied to the data provided a slope of 2.66 T–1 and an intercept of 0.99 for carbogen breathing and a slope of 1.98 T–1 and anintercept of 0.51 for oxygen breathing.

Figure 5. Respiratory-induced gray matter MR signal changes expectedfor a sequence with pure T2* weighting of the image contrast as a functionof field strength and TE, expressed as a percentage of the maximum base-line signal at TE= 0ms. For increasing field strength, the maximum effectsare expected to increase and shift to smaller TEs. The maximum effect islarger for carbogen breathing than for oxygen, but is expected to be ob-served at only minimally shorter TEs than under hyperoxia (i.e. duringthe breathing of 100% O2). For single-echo observation of a challenge re-sponse, the selection of a TE half as long as the TE with maximum effectshould offer improved image quality at the expense of a reduction in theeffect size by c. 20%.



(i.e. ΔR2*/R2*) was provided by the measurement of the relativechange in the MR signal (ΔS, %) at the selected TE (Fig. 5 andTable 2). Losert et al. (8) and Rauscher et al. (29) reported a BOLDcontrast on the order of 3% over the cortical gray matter during100% O2 breathing at 1.5 T. These values are in optimal agreementwith the magnitude of the relative R2* change measured in thisstudy in similar experimental conditions (ΔR2*/R2* = 3.2� 1.9%).Carbogen breathing has been reported to lead to a BOLD contrastof c. 4 % at 1.5 T (29) and c. 8% at 3 T (30). In this study, inhalation ofcarbogen led to absolute ΔR2*/R2* values of 5.3� 1.7% and8.7� 2.8% at 1.5 and 3 T, respectively. A slightly higher change inR2* (absolute ΔR2*: 2.4� 0.5 s–1 versus 1.49� 0.49 s–1) has been es-timated by Remmele et al. (20) at 3 T during carbogen breathing.To the best of our knowledge, no studies have been reported sofar using comparable settings and similar gas mixtures at 7 T.

Field dependence of the challenge-induced R2*changes

One of the aims of this study was the evaluation of the fieldstrength dependence of the MR response to respiratory chal-lenges. In this work, we found a distinct increase in the observedΔR2* magnitudes when going from 1.5 to 3 T (the ratio of theΔR2* values was 2.2 for carbogen and 2.0 for oxygen) as well asfrom 3 to 7 T (the ratio of the ΔR2* values was 3.8 for carbogen

and 4.9 for oxygen). The dependence of the ΔR2* magnitudeson the field strength fits very well with a quadratic function, forboth oxygen and carbogen challenge (Fig. 4a).Ogawa et al. (31) and Boxerman et al. (32) proposed two bio-

physical models to predict the dependence of the BOLD signaloriginating from the extravascular compartment of the tissueon the strength of B0, CBF and oxygen consumption. Accordingto these models, a quadratic dependence of ΔR2* on the fieldstrength should be expected for voxels mainly containing micro-vessels (i.e. vessels with a radius smaller than 8mm), whereas arather linear dependence should be observed in proximity tolarger vessels. The quadratic dependence of ΔR2* on B0, foundin our study, suggests that the ROIs drawn over the cortical graymatter mainly contained a microvascular contribution. This resultis in accordance with the data reported by Yacoub et al. (33) us-ing a voxel volume of 3.2mm3 at 4 and 7 T. A supralinear depen-dence of ΔR2* on B0 was observed by Turner et al. (12) and Gatiet al. (16). However, other studies investigating the field strengthdependence of the BOLD signal have reported a linear increasein the effect with B0 (21,34). The apparent discrepancy betweenthe data reported in the literature may be partially explainedby considering the dependence of the contribution of themacrovasculature to the BOLD signal on the size of the voxel.Lai et al. (35) showed that, in voxels containing both large and

Table 1. Results of the region of interest (ROI) analysis performed over the cortical gray matter. Local R2* values measured duringmedical air breathing (baseline) and relative deviations of the transverse relaxivity measured during the inhalation of 100% O2 (O2)and carbogen (CB) are reported for 1.5, 3 and 7 T. For each subject, the mean value and standard deviation resulting from a multi-ROI analysis are reported. The statistical significance of the data is highlighted using roman type for nonsignificant data (p> 0.05),italic type for significant data (p< 0.05) and bold type for highly significant data (p< 0.01)

1.5 T 3 T 7 T

Subject

R2* (s–1) ΔR2*/R2* (%) R2* (s

–1) ΔR2*/R2* (%) R2* (s–1) ΔR2*/R2* (%)

Baseline O2 CB Baseline O2 CB Baseline O2 CB

1 11.3� 1.0 �5.6� 3.2 �6.8� 3.7 17.2� 0.8 �4.2� 7.2 �7.8� 2.1 29.2� 2.5 �16.1� 8.4 �20.8� 5.02 12.8� 1.8 �1.3� 0.9 �4.9� 2.7 18.3� 2.0 �4.2� 3.7 �12.1� 2.6 28.5� 1.8 �10.1� 5.7 �20.5� 1.83 11.0� 1.8 �5.6� 6.1 �6.7� 2.6 16.9� 1.0 �5.3� 2.8 �8.1� 4.3 28.3� 4.5 �17.9� 7.5 �23.0� 9.04 13.4� 2.0 �2.1� 2.2 �2.7� 2.0 17.2� 0.8 �0.4� 2.1 �4.7� 9.1 26.8� 4.3 �19.8� 7.1 �19.7� 11.45 13.3� 0.7 �4.0� 2.2 �6.7� 3.2 15.8� 1.0 �6.6� 2.4 �9.6� 4.6 27.9� 5.1 �9.6� 4.5 �17.5� 6.06 11.9� 0.5 �2.0� 2.7 �7.1� 1.8 16.7� 0.8 �0.8� 2.2 �6.1� 2.4 33.8� 2.5 �12.3�2.4 �19.8� 2.37 19.5� 2.5 �0.8� 2.7 �4.4� 1.4 15.6� 1.6 �7.1� 9.9 �8.6� 5.6 31.6� 4.7 �9.1� 5.4 �15.7� 3.38 13.1� 1.3 �3.8� 3.6 �3.3� 2.1 17.2� 0.6 �8.4� 5.0 �12.9� 3.6 26.0� 3.3 �13.8� 2.2 �20.3� 1.1Mean 13.3� 2.7 �3.2� 1.9 �5.3� 1.7 16.9� 0.9 �4.6� 2.9 �8.7� 2.8 29.0� 2.6 �13.6� 4.0 �19.7� 2.2

Table 2. Maximum blood oxygenation level-dependent (BOLD) signal changes (ΔSmax) achievable at optimal echo time (TEopt)and maximum BOLD signal changes expressed as a percentage of the maximum baseline signal at TE = 0ms are reported for1.5, 3 and 7 T. BOLD signal changes versus TE were computed for the rates of relaxation assessed in the cohort of volunteers duringrespiratory challenges by assuming a mono-exponential signal decay and no inflow effects

Oxygen Carbogen

B0(T)

TEopt(ms)

ΔSmax as % of Sbaseline(TEopt)

ΔSmax as % of Sbaseline(0ms)

TEopt(ms)

ΔSmax as % of Sbaseline(TEopt)

ΔSmax as % of Sbaseline(0ms)

1.5 76 3.0 1.1 77 5.3 2.03 61 4.7 1.8 62 8.9 3.47 37 13.1 5.2 38 19.3 7.9

C. ROSSI ET AL.


small vessels, the macrovasculature dominates the BOLD con-trast. In an experiment performed with a relatively low spatialresolution, each voxel is expected to contain a contribution ofboth large and small vessels. Depending on the relative propor-tion of the macrovasculature within the voxel, the BOLD contrastmay show a linear dependence on the field strength or a mixtureof linear and quadratic terms. Therefore, the image resolution isan important parameter in the experimental assessment of thefield strength dependence of the BOLD response to the chal-lenge of cortical gray matter.The relatively small voxel size of 0.25mm3 used in this work

for R2* mapping suggests a minor contribution of venules andveins to the relaxivity of the MR signal measured in cortical graymatter. However, it is worth noting that the accuracy of theassessment of the field strength dependence of the BOLD re-sponse to the respiratory challenge may be limited by the smallsize of the effect measured at 1.5 Tesla.The relative R2* changes (ΔR2*/R2*) induced by carbogen and

oxygen breathing increase with the field strength. This resultconfirms that the MR signal response at the optimal TE is higherat higher magnetic field strength (16,18). The data fitted wellwith a linear curve. Nevertheless, a hyperlinear dependence ofthe ΔR2*/R2* ratio on the strength of the static magnetic fieldcannot be excluded (Fig. 4b).

Three-dimensional FFE signal change and optimum TE

Signal changes with TE can be calculated for a T2*-weighted im-aging sequence from R2* and ΔR2*. The optimum TE that yieldsthe largest signal change in relation to the maximum baselinesignal observed at TE = 0 can be determined (Fig. 5). This, in turn,allows an estimation of the maximum relative signal change thatmay be observed at the optimum TE on respiratory challenge(Table 2). The range of relative signal changes is large. The detec-tion of effects caused by oxygen inhalation at 1.5 T on the orderof 3% may require dedicated post-processing methods (8). Incontrast, the large effects induced by carbogen breathing at7 T, on the order of 19%, may allow direct visual assessment ofthe regional inhomogeneity of the gray matter response to thechallenge.At high magnetic field, a relatively short TE may prevent signif-

icant signal loss in T2*-weighted images (28). The dependence ofthe MR signal response on TE (Fig. 5) suggests that, in a single-echo experiment, some of the MR response may be traded forimproved overall image quality and minimum scan time.

Paramagnetic effects of molecular oxygen

Increasing oxygen supply results in an excess of molecular oxy-gen dissolved in the blood plasma of the lungs. Dissolved oxygenmoves from the lungs into the tissue capillaries, where it binds tohemoglobin (36). Changes in the amount of dHb cause BOLDcontrast. In addition to this effect, the slightly paramagneticmolecular oxygen that remains dissolved in the capillary bed ofthe tissue, or in the CSF, as well as the excess of paramagneticgaseous oxygen in the upper (36) and lower (37) airways, maylead to non-BOLD signal changes in T2*-weighted images.As a result of the weak transverse relaxivity of molecular oxy-

gen, signal changes originating from dissolved oxygen in cere-bral tissue are usually considered to be small (38). However, bulksusceptibility changes originating from variation in the amountof dissolved oxygen in the airways may affect significantly the

MR signal in several brain areas (36,37). Pilkinton et al. (36)showed that, in brain regions close to the nasal and oral cavities,the paramagnetic effects of dissolved oxygen may be of thesame order of magnitude as the BOLD-induced signal changes(i.e. 0.1 ppm at 3 and 7 T). If not taken into account, these para-magnetic effects may lead to region-dependent non-BOLDsignal changes that may affect the quantification of the BOLDresponse to the challenge. These changes are expected to causepredominantly R2* shortening on hyperoxic challenges, and tovary strongly with the spatial vicinity to structures holding in-creased molecular oxygen concentrations.

Three aspects of the experimental set-up of our study shouldhave mitigated the paramagnetic effects of molecular oxygenon the BOLD signal. First, automatic shim adjustments were per-formed before the acquisition of each multi-echo T1-FFE dataset.Second, the slab orientation was carefully selected to avoid theupper airways (Fig. 1a). Finally, the high resolution of the datasetsshould limit the influence of macroscopic field inhomogeneitieson the voxel signal.

In conclusion, this study found that the R2* response of corti-cal gray matter to hyperoxic respiratory challenge was higherunder hyperoxic hypercapnia (carbogen breathing) than underhyperoxia (oxygen breathing). The magnitude of the R2* changeincreased quadratically with the strength of the static magneticfield. The quantification of the MR response to respiratory chal-lenge at high magnetic field may find clinical application inthe development of new strategies for tumor therapies, in thefollow-up of acute and chronic brain injuries and in the evalua-tion of degenerative brain processes.

Acknowledgements

CR was supported by the Foundation for Research at the Facultyof Medicine, University of Zurich (grant no. 34270124).

REFERENCES

1. Busl KM, Greer DM. Hypoxic–ischemic brain injury: pathophysiol-ogy, neuropathology and mechanisms. NeuroRehabilitation 2010;26: 5–13.

2. Oliver L, Olivier C, Marhuenda FB, Campone M, Vallette FM. Hypoxiaand the malignant glioma microenvironment: regulation and impli-cations for therapy. Curr. Mol. Pharmacol. 2009; 2: 263–284.

3. Buxton RB. Interpreting oxygenation-based neuroimaging signals:the importance and the challenge of understanding brain oxygenmetabolism. Front. Neuroenerg. 2010; 2: 8.

4. Rostrup E, Larsson HB, Toft PB, Garde K, Henriksen O. Signal changesin gradient echo images of human brain induced by hypo- andhyperoxia. NMR Biomed. 1995; 8: 41–47.

5. Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imag-ing with contrast dependent on blood oxygenation. Proc. Natl. Acad.Sci. USA 1990; 87: 9868–9872.

6. Overgaard J. Hypoxic radiosensitization: adored and ignored. J. Clin.Oncol. 2007; 25: 4066–4074.

7. Johnston AJ, Steiner LA, Gupta AK, Menon DK. Cerebral oxygenvasoreactivity and cerebral tissue oxygen reactivity. Br. J. Anaesth.2003; 90: 774–786.

8. Losert C, Peller M, Schneider P, Reiser M. Oxygen-enhanced MRI ofthe brain. Magn. Reson. Med. 2002; 48: 271–277.

9. Kastrup A, Kruger G, Neumann-Haefelin T, Moseley ME. Assessmentof cerebrovascular reactivity with functional magnetic resonanceimaging: comparison of CO(2) and breath holding. Magn. Reson.Imaging 2001; 19: 13–20.

10. Rostrup E, Law I, Blinkenberg M, Larsson HB, Born AP, Holm S,Paulson OB. Regional differences in the CBF and BOLD responsesto hypercapnia: a combined PET and fMRI study. Neuroimage2000; 11: 87–97.



11. Van der Zande FH, Hofman PA, Backes WH. Mapping hypercapnia-induced cerebrovascular reactivity using BOLD MRI. Neuroradiology2005; 47: 114–120.

12. Turner R, Jezzard P, Wen H, Kwong KK, Le Bihan D, Zeffiro T,Balaban RS. Functional mapping of the human visual cortex at4 and 1.5 tesla using deoxygenation contrast EPI. Magn. Reson.Med. 1993; 29: 277–279.

13. Frahm J, Merboldt KD, Hänicke W, Kleinschmidt A, Boecker H. Brainor vein – oxygenation or flow? On signal physiology in functionalMRI of human brain activation. NMR Biomed. 1994; 7: 45–53.

14. Mürtz P, Flacke S, Müller A, Soehle M, Wenningmann I, Kovacs A,Träber F, Willinek WA, Gieseke J, Schild HH, Remmele S. Changes inthe MR relaxation rate R(2)* induced by respiratory challenges at3.0 T: a comparison of two quantification methods. NMR Biomed.2010; 23: 1053–1060.

15. Coleman TF, Li Y. On the convergence of interior-reflective Newtonmethods for nonlinear minimization subject to bounds. MathematicalProgramming 1994; 67: 189–224.

16. Gati JS, Menon RS, Ugurbil K, Rutt BK. Experimental determination ofthe BOLD field strength dependence in vessels and tissue. Magn.Reson. Med. 1997; 38: 296–302.

17. Li TQ, Yao B, van Gelderen P, Merkle H, Dodd S, Talagala L, Koretsky AP,Duyn J. Characterization of T2* heterogeneity in human brain whitematter. Magn. Reson. Med. 2009; 62: 1652–1657.

18. Peters AM, Brookes MJ, Hoogenraad FG, Gowland PA, Francis ST,Morris PG, Bowtell R. T2* measurements in human brain at 1.5, 3and 7 T. Magn. Reson. Imaging 2007; 25: 748–753.

19. Uludag K, Müller-Bierl B, Ugurbil K. An integrative model for neuronalactivity-induced signal changes for gradient and spin echo func-tional imaging. Neuroimage 2009; 48: 150–165.

20. Remmele S, Dahnke H, Flacke S, Soehle M, Wenningmann I, Kovacs A,Träber F, Müller A, Willinek WA, König R, Clusmann H, Gieseke J,Schild HH, Mürtz P. Quantification of the magnetic resonance signalresponse to dynamic (C) O2-enhanced imaging in the brain at 3 T:R2* BOLD vs. balanced SSFP. J. Magn. Reson. Imaging 2010; 31:1300–1310.

21. Van der Zwaag W, Francis S, Head K, Peters A, Gowland P, Morris P,Bowtell R. fMRI at 1.5, 3 and 7 T: characterising BOLD signal changes.Neuroimage 2009; 47: 1425–1434.

22. Vavilala MS, Lee LA, Lam AM. Cerebral blood flow and vascular phys-iology. Anesthesiol. Clin. N. Am. 2002; 20: 247–264.

23. Sedlacik J, Kutschbach C, Rauscher A, Deistung A, Reichenbach JR.Investigation of the influence of carbon dioxide concentrations oncerebral physiology by susceptibility-weighted magnetic resonanceimaging (SWI). Neuroimage 2008; 43: 36–43.

24. Bulte DP, Chiarelli PA, Wise RG, Jezzard P. Cerebral perfusion re-sponse to hyperoxia. J. Cereb. Blood Flow Metab. 2007a; 27: 69–75.

25. Howe FA, Robinson SP, Rodrigues LM, Griffiths JR. Flow and oxygen-ation dependent (FLOOD) contrast MR imaging to monitor the

response of rat tumors to carbogen breathing. Magn. Reson. Med.1999; 17: 1307–1318.

26. Baudelet C, Gallez B. How does blood oxygen level-dependent(BOLD) contrast correlate with oxygen partial pressure (pO2) insidetumors? Magn. Reson. Med. 2002; 48: 980–986.

27. Hennig J, Speck O, Koch MA, Weiller C. Functional magnetic reso-nance imaging: a review of methodological aspects and clinicalapplications. J. Magn. Reson. Imaging 2003; 18: 1–15.

28. Reichenbach JR, Venkatesan R, Yablonskiy DA, Thompson MR, Lai S,Haacke EM. Theory and application of static field inhomogeneityeffects in gradient-echo imaging. J. Magn. Reson. Imaging 1997; 7:266–279.

29. Rauscher A, Sedlacik J, Barth M, Haacke EM, Reichenbach JR. Nonin-vasive assessment of vascular architecture and function during mod-ulated blood oxygenation using susceptibility weighted magneticresonance imaging. Magn. Reson. Med. 2005; 54: 87–95.

30. Gauthier CJ, Madjar C, Tancredi FB, Stefanovic B, Hoge RD. Elimina-tion of visually evoked BOLD response during carbogen inhalation:implications for calibrated MRI. Neuroimage 2011; 54: 1001–1011.

31. Ogawa S, Menon RS, Tank DW, Kim SG, Merkle H, Ellermann JM,Ugurbil K. Functional brain mapping by blood oxygenation level-dependent contrast magnetic resonance imaging. A comparisonof signal characteristics with a biophysical model. Biophys. J.1993; 64: 803–812.

32. Boxerman JL, Bandettini PA, Kwong KK, Baker JR, Davis TL, Rosen BR,Weisskoff RM. The intravascular contribution to fMRI signal change:Monte Carlo modeling and diffusion-weighted studies in vivo. Magn.Reson. Med. 1995; 34: 4–10.

33. Yacoub E, Shmuel A, Pfeuffer J, Van der Moortele PF, Adriany G,Andersen P, Vaughan TJ, Merkle H, Ugurbil K, Hu X. Imaging brainfunction in humans at 7 Tesla. Magn. Reson. Med. 2001; 45: 588–594.

34. Driver I, Blockley N, Fisher J, Francis S, Gowland P. The change in ce-rebrovascular reactivity between 3 T and 7 T measured using gradedhypercapnia. Neuroimage 2010; 51: 274–279.

35. Lai S, Hopkins A, Haacke E, Li D, Wasserman B, Buckley P, Freitlman L,Meltzer H, Hedera P, Freidland R. Identification of vascular structuresas a major source of signal contrast in high resolution 2D and 3Dfunctional activation imaging of the motor cortex at 1.5T: preliminaryresults. Magn. Reson. Med. 1993; 30: 387–392.

36. Pilkinton DT, Gaddam SR, Reddy R. Characterization of paramagneticeffects of molecular oxygen on blood oxygenation level-dependent-modulated hyperoxic contrast studies of the human brain. Magn.Reson. Med. 2011; 66: 794–801.

37. Raj D, Paley DP, Anderson AW, Kennan RP, Gore JC. A modelfor susceptibility artefacts from respiration in functional echo-planar magnetic resonance imaging. Phys. Med. Biol. 2000; 45:3809–3820.

38. Bulte DP, Chiarelli PA, Wise RG, Jezzard P. Cerebral perfusionresponse to hyperoxia. J. Cereb. Blood Flow Metab. 2007b; 27: 69–75.

C. ROSSI ET AL.


manipulation of cortical gray matter oxygenation by hyperoxic respiratory challenge: field...

Documents