Brain magnetic resonance imaging with contrast dependent on blood oxygenation

Proc. NatI. Acad. Sci. USA

Vol. 87,pp.9868-9872, December 1990 Biophysics

Brain

magnetic resonance imaging with contrast dependent on

blood oxygenation

(cerebral blood flow/brainmetabolism/oxygenation)

S. OGAWA, T. M. LEE, A. R. KAY, AND D. W. TANK

Biophysics Research Department, AT&T Bell Laboratories, Murray Hill, NJ 07974

Communicated by Frank H. Stillinger, September 24, 1990 (received for review August1, 1990)

ABSTRACT Paramagnetic deoxyhemoglobin in venous blood is a naturally occurring contrast agent for magnetic resonance imaging (MRI). By accentuating the effects of this agent through the use of gradient-echo techniques in high fields, we demonstrate in vivo images of brain microvasculature with image contrast reflecting the blood oxygen level. This blood oxygenation level-dependent (BOLD) contrast follows blood oxygen changes induced by anesthetics, by insulin-induced hypoglycemia, and by inhaled gas mixtures that alter

metabolic demand or blood flow. The results suggest that BOLD contrast can be used to provide in vivo real-time maps

ofbloodoxygenation in the brain undernormal physiological conditions.BOLDcontrastadds an additional feature to mag-netic resonance imaging and complements other techniques that are attempting to provide positron emission tomography-likemeasurements related to regional neural activity.

Magnetic resonance imaging (MRI) is a widely accepted modality for providing anatomical information. Current

re-search (1) involves extending MRI methods to provide in-formation aboutbiological function, in addition tothe

con-comitant anatomical information. In addition to localized

spectroscopy (2) and chemical shift imaging (3) that are

applicable tomanychemical species, MRI ofwaterprotons

has been functionally extended to NMR angiography (4), perfusionimaging(5, 6), andperfusionimagingenhancedby

exogenous contrast agents (7). Since water is by far the predominant molecule in tissue, and since its signal

domi-nates the information content in protonimages, one would ideally liketoexploit changes in thewater signalthat arise fromphysiological events. Except forcases ofwater move-ment, such as bloodflow, these changes arenormally very small.

It has previously been demonstrated (8, 9) that the pres-enceofdeoxyhemoglobin inbloodchanges theprotonsignal fromwatermoleculessurroundingablood vessel in gradient-echo MRI, producing blood oxygenation level-dependent (BOLD) contrast. BOLD contrasthas its origin in the fact thatwhennormallydiamagnetic oxyhemoglobingivesupits

oxygen,theresultingdeoxyhemoglobinisparamagnetic.The presence of paramagnetic molecules in blood produces a

difference inmagneticsusceptibilitybetween the blood

ves-sel andthesurroundingtissue. Thissusceptibilitydifference is "felt" both by the watermolecules in theblood andby

thosein thesurrounding tissue,theeffectextending signifi-cantlybeyondthevessel wall. This increasein thenumber of spins affected bydeoxyhemoglobin isaformof amplification. When the susceptibility-inducedlocal field differences exist within an imaging voxel, there isa resultant distribution of

shifts in water resonance frequencies. Inthegradient-echo method, a phase dispersion ofwaterproton signals is

pro-duced at the echo time. This dispersion reduces the signal intensity and the voxel appears dark in the image. These

intensity losses, whichathigh magnetic fields (-4 T) extend

significantly beyondtheboundary ofthe bloodvessel, arethe

source of BOLD contrast. This form of contrast is not observed in spin-echo images. Through simulations (9), we haveshownthatvessels as small as 50

,um

indiameter can be

detected in images withapixel size of100pum.We have also demonstrated that the size ofthesusceptibility-induced local fielddependson(i)the concentration ofparamagnetic deox-yhemoglobin and (ii) the orientation of the vessel relativeto

the main magnetic field (8, 9).

Since BOLD contrast depends on the state of blood oxygenation, physiological events that change the

oxy/

deoxyhemoglobin ratio should lend themselves to

noninva-sivedetection through the accentuation of BOLDcontrastin gradient-echo proton images at high magnetic fields. We report here that this is indeed the case and demonstrate changes in BOLD-contrast microimages of brain produced by changes in inhaledgasmixture under urethaneanesthesia, by insulin-induced hypoglycemia under diazepamsedation, and by changes in the levelof halothane anesthesia. The observed changes in BOLD contrast correlate with the anticipated changes in bloodoxygenlevel producedby altered metabolic

load orblood flow(10).

MATERIALS AND METHODS

Physiology and Anesthetics. Sprague-Dawleyrats (female,

250± 25g)wereanesthetized with thedescribed anesthetics andventilated without assistance. Animal care and the ad-ministration of anestheticswerein accordance with National Institutes of Health guidelines. For MRI experiments, the animal was placed in a thermally regulated compartment

withinthe magnet and rectaltemperatureand

electrocardio-gram weremonitored. Blood wasperiodicallysampledfrom thefemoral artery and bloodgasconcentrations were

mea-sured withaRadiometer model ABL330gasanalyzer. Blood glucosewasestimatedusingaBoehringer-Mannheim Accu-Check Ilm. To measure the electroencephalogram (EEG) duringMRIexperiments, small tubularsaltbridges (0.8mm

i.d.)wereimplanted insmallholesthroughthe skull thatwere

locatedinside theareaencircledby the surface coil used for

NMRsignal detection. Carewastakennottobreakthedura.

Magnetic ResonanceMicroimaging. Gradient-echo images

were obtained with a 7-T horizontal magnet at 300 MHz (Spectroscopy Imaging Systems, Fremont, CA). Surface coils of20 to 25-mmdiameterwerepositionedonthesurface ofthecraniumand usedfor both excitation and detection of signals. Themaximumfieldgradientusedtoproduceimages

was4 G/cm(1 G = 0.1 mT). The slice width of theimages shownwas500

,4m

and thein-planepixelsizewas117,um X

Abbreviations: MRI, magnetic resonance imaging; BOLD, blood

oxygenation level-dependent; EEG, electroencephalogram.

9868 Thepublicationcostsof this articleweredefrayedin partbypagecharge payment.This articlemusttherefore beherebymarked"advertisement" in accordance with18U.S.C. §1734solelytoindicate thisfact.

Proc. Natl. Acad. Sci. USA 87 (1990) 9869

FIG.1. Effect of bloodCO2levelonBOLDcontrast.(a)Coronal slice brain image showing BOLD contrast froma rat anesthetized with urethane. The gas inspired was100%02-(b) The same brain but with90%02/10%CO2 as the gas inspired. BOLD contrast is greatly reduced. The arrow points to the sagittal sinus, showing increased signal intensity, indicative of increased blood oxygenation. The magnetic field was perpendicular to the slice plane. Theimaging time was 9minwitharepetition time of 0.26 sec andanechotime of 12 msec.Thepixel size of theimageswas 117Amx117,um withaslice thickness of 550Aum. Thescale represents lineardimension incm. Corebody temperature wasmaintainedat35-36.50C.

117

Am.

Theecho time of theimagingwas12-18 msecand the

imageacquisition time was6-15 min.

Measurement of Blood FlowVelocity and Effect of Blood

Oxygenation on Signal Intensity at the Sagittal Sinus. The

venousblood flowvelocity(cm/sec)atthesagittal sinuswas

measured by themethod ofpresaturation recovery by flow (11). Proton spins within a 1-mm- or2-mm-thick slicewere

selectively presaturated and after a specified delay time a

0.5-mm-thick slice at the same location was selected for

image observation by a gradient-echo pulse sequence with

12-msececho time. Themajor direction of the blood flow in thesagittal sinus wasnormal tothe sliceplane. The venous

bloodsignalrecoveredwithincreasing delay time, providing

a measure of blood flow velocity. In this presaturation-recovery-by-flow method,thecontributiontothe flowsignal by thevenousblooddrainageintothesagittal sinus within the presaturation slice was minimized because ofthe slow

lon-gitudinal relaxation time (T1)recovery ofthese spins.

Inthe standardgradient-echoimages shown inthefigures (nopresaturation), thesignal intensity ofthe venousblood

water atthesagittal sinus dependsontheblood oxygenation. This is because thetransverserelaxation time (T2) of blood

waterisstronglydependentonbloodoxygenation (12).Atthe

7-Tfield strengthused inthisstudy, T2 varies from50 msec

FIG. 2. Effect ofinsulin-induced hypoglycemiaon BOLD con-trast.(a)Afterinjectionofbovineinsulin,BOLDcontrastis apparent atbloodglucoselevels¢20mg/dl.(b)Lossofcontrastisobserved atblood glucose levels << 20mg/dl. (c) BOLD contrast can be restoredbyintraperitonealinjectionofmannose.Inhalant gas,50o

02/50%o

N2;core

body

temperaturewasmaintainedat

34-35°C;

at 100% oxygenation to 4 msec at 0% oxygenation. The venous bloodsignal becomes very weakwhen the T2value

becomescomparableto orshorter than theechotime of the

signal acquisition. At60%ooxygenation level,the estimated

T2 value is 18 msec, similar to the echo time used in this report.

Simulation ofBOLD-Contrast Images. The image simula-tion (see ref. 9) assumed a cylinder of radius a oriented

Proc. Nati.Acad. Sci. USA 87 (1990)

perpendicular to the applied magnetic field and having a volume magnetic susceptibility difference of Ax relative to thesurrounding tissue. The frequency shift(o) as a function ofposition used in the simulation was then

outside thecylinder: wz/wo = 2,irA(r/a)2[(2cos26 - 1)] inside thecylinder: wS/w0 =

-(2T/3)AX,

where r (r, 6)is the position vector relative to the center of the cylinderwith angle 6 to the magnetic field. The resonance frequency without the local field is

o".

The field inside the cylinderwasassumed tobe homogeneous, but the T2 decay rate of blood water had a quadratic dependence on the oxygenation level (12). The voxel size in the simulation was 100 ,um x 100

Atm

x 500 ,m and the pixel size infrequency along the read gradient was 100 Hz. The maximum frequency shift at the surface of the vessel

(2rcoAX)

was 200Hz, similar tothat measured in in vitro experiments with deoxygenated blood (9). The echo time was 10 msec.

RESULTS

Effects ofAnesthetics and Blood Flow on BOLD Contrast. Coronal slice images of rat brains with BOLD contrast

demonstrating thatimagecontrastisreducedasbloodflow is increased are shown in Fig. 1. Rats under urethane (850 mg/kg i.p. with atropine at 0.04 mg/kg s.c.) anesthesia with 100% 02asthe gasinspired had strong BOLD contrast(Fig. la), manifested as numerous dark lines in the vicinity of venous bloodvessels. In neocortex these vessels lie perpen-dicular to the cortical surface. This BOLD contrast was

completely lost (Fig. lb) when the gas inhaled was switched to10% C02/90% 02. The addition of CO2 to the inhalant gas increased bloodCO2 levels from 50 to 80 mmHg (1 mmHg = 133 kPa). In general, increased blood CO2 produces increased blood flow. This increased blood flow provides a greater supply of oxygen to the brain and, in the absence of achange inthe metabolicload, should increase venous blood oxygen-ation, consistent with our observed loss of BOLD contrast with change in blood CO2 level. By using a presaturation-recovery-by-flow MRI method, we could directly measure

theincrease in blood flow produced in the sagittal sinus. The measured flow velocity (cm/sec) increased from 0.4 to 1.7 cm/sec. Note that there was a concomitant signal intensity increase at thesagittal sinus (Fig.

lb,

arrow) in the gradient-echo image. This increase also indicates a higher level of

'Av~~4#A&FM:~~Y#Mp

200 uV 2 sec

FIG. 3. Dependenceof BOLDimagecontrasthalothane dose.Brainimagesofaratinspiring0.75% halothane in oxygen(a),3% halothane inoxygen(b),or100%N2(c)areshown. AlsoshownareEEGmeasurements ateachhalothane levelshortlybefore theimageswereacquired. 9870

Biophysics:

Ogawa

etal.

Proc. NatL. Acad. Sci. USA 87 (1990) 9871

oxygenation in the venous blood, because of the known dependenceofwaterT2 onbloodoxygenation. Thus the loss of BOLD contrast and the high venous blood oxygenation

level measured at thesagittal sinuswereconsistent withthe largeincrease inbloodflow velocity inducedby theincreased systemic CO2.

Insulin-Induced Hypoglycemia Reduces BOLD Contrast. As

shown in Fig. 2, under diazepam sedation (10 mg/kg) and with 50%02/50% N2 as theinhalantgas, a rat brain image demonstrated a BOLD contrast slightly more pronounced thanthat inFig. la. (Thesignalintensityatthesagittal sinus

waslessthan thatin Fig.la,suggestingalower levelof blood oxygenation under these conditions.) After injection of in-sulin (bovine, 16units/kg; Sigma), BOLD contrast did not

change while blood glucose remained >20mg/dl (Fig. 2a). When bloodglucose droppedtowell below20mg/dl,BOLD

contrast wasgreatly reduced(Fig. 2b). Addition of glucoseor mannose(400mg/kg i.p.) reversed this change, restoring the

contrast(Fig. 2c).

Thereduction in BOLDcontrastduring hypoglycemiawas

correlated with a modest increase in blood velocity at the sagittal sinus. Paco2 levels in blood before insulin-induced hypoglycemiawere 35-40 mmHg withblood velocityat the sagittal sinus of0.4-0.6 cm/sec. When the loss of BOLD

contrast occurred, Paco2 was 30-38 mmHg and blood

ve-locitywas0.7-1.1cm/sec. However,itseemslikelythat the

loss ofBOLDcontrastproduced bystronghypoglycemiawas notentirely produced by theincrease in blood flow but also fromareducedmetabolism, since,atnormalglucose levels, modestincreases in blood flow produced by other methods didnot cause acomplete elimination of BOLDcontrast(data

notshown).

BOLD-Contrast Reduction Without Increased Blood Flow. When the inhalation anesthetic halothane isused, itis

pos-sibletovarythestateof anesthesia of the animalbyvarying theconcentration of halothane andtodemonstrateachange in BOLDcontrast not produced by changes in blood flow velocity. We have comparedaseriesof brainimagesfrom the

sameanimal produced under identicalimagingbut different degrees of anesthesia. To obtain an estimate of overall neuronalactivity, the EEG signalwasmeasuredin situbefore

orafterimage acquisitionfrom electrodes positionedonthe duraneartheimagedcoronalsection. The dosedependence of the BOLD contrast is shown in Fig. 3, along with the corresponding EEG signals. With 0.75% halothane inoxygen as the inhalant gas, radiating venous blood vessels were

observed in the neocortex, demonstrating BOLD contrast

(Fig. 3a). Increasing the halothane content to 3% led to

depression of the EEG(brief periodsof isoelectricsignalare

evident)andconcomitant loss ofcontrastin theMRI image (Fig. 3b). Finally,toexhibit maximumpossiblecontrast,the anesthetized animalwasventilated with

100%o

nitrogen, lead-ingto anisoelectric EEG.Thehighly deoxygenated blood in thebrain under this conditiongave very high BOLDimage

contrast(Fig. 3c). Unlike the urethaneexperimentsdescribed above, it is unlikely that the decreased image contrast ob-servedwithincreasinghalothanecontent canbeexplainedas an effect of halothaneoncerebral blood floworvolume. In separate experiments the venous blood flow velocity was

measuredatvarious halothane levels. Flowvelocityactually decreased from 0.25 cm/sec at 0.75% halothane to 0.12

cm/sec at3% halothane. The decreased BOLDcontrast in Fig. 3b is therefore probably due to the lowered oxygen

consumption at 3%halothane, where theanimal reached a state of low brainactivity asjudgedby theEEG.

Image Simulation. Asa steptowardquantifying expected changes in BOLD contrast,wehave usedimage simulationto

estimate its dependenceonbloodoxygenation. Image simu-lationswereperformedforacylindricalblood vessel

perpen-dicular to the main magnetic field, using experimentally

determined values for themagnetic parametersofthe blood and the surrounding tissue. The relation between BOLD

contrastand blood oxygenation thatresulted from the sim-ulations is shown inFig.4aforavarietyof blood vessel sizes.

Thenormalized BOLDcontrast, (IO-

])/Io,

changed in the

range of blood oxygenation between 90% and 50%, the

anticipatedrangeofvenousbloodoxygenation.The apparent

width ofthe vessel in the image is also blood oxygenation dependentfor those with diameters that exceedthe

pixel

size (9). An anoxic brain image is shown with a corresponding signal intensity profile in Fig. 4b. An evaluation of the

sensitivity of BOLD contrast to the change in the oxygen

tension in the blood can be made for this image from the simulation results. If the minimum detectablechangein the

contrast/noise

ratio (Al/N) is assumed tobe2 in an

image

with a

signal/noise

ratio

(IO/N)

of40, the

corresponding

change in the blood oxygenation level (Y) at Y 0.5 in a 50-tim size vessel would be 0.15. Taking account of the

cooperative

oxygen

binding

to

hemoglobin

in

blood,

this changein Yrelatesto achange of20%oin the bloodoxygen

tension. The sensitivity of the method will be reduced for vessels thatare notorthogonaltothemagneticfield direction.

Inthisregard,the vasculatureofthecortexisadvantageous

asit runs

predominantly

in the coronal plane.

a

1.0

0.8-0.6 0 0.4 0.2 Diameter, ,m 200 100

>t 4

M~~~~~~~oo

0.50 0.25 Blood

oxygenation

FIG. 4. (a) Relationship between BOLD contrast and blood

oxygenation.Theimagesimulationwasperformedforablood vessel with itslongitudinalaxisperpendiculartothemagneticfield and also

perpendiculartotheread-gradientaxis in the sliceplane. (b)Signal intensityprofile alongthe line shown in theaccompanyingimageof ananoxic brain(ventilationwith

100%o

N2).

Biophysics: Ogawa

etal.

Im

Proc. Natl. Acad. Sci. USA87 (1990)

In order to use BOLD contrast toquantitatively determine regional venous blood oxygenation, in vivo calibration will be necessary. For example, images obtained at two different averageblood oxygenation levels, produced by different 02 compositionsin theinhalantgas,provide sufficient informa-tion so that local contrast changes in subsequent images of the same area can be related to changes in local blood oxygenation quantitatively.

DISCUSSION

Wehave demonstrated that BOLD-contrast MRI of brain can track the changes inblood oxygenation expected from anes-thetic inhaled gases and from insulin-induced hypoglycemia that change thephysiological state of the animal. Since the contrast is entirely dependent on blood oxygenation, it is determined by the balanceof supply (blood flow) and demand (extraction by tissue) of oxygen. Under normoxic conditions arterial blood isfully oxygenated and does not contribute to BOLD contrast, while venous blood vessels containing de-oxygenated blood contribute dark lines to the image. The results shown here indicate that BOLD contrast can be used tononinvasively monitorin real time theblood oxygenation levels of brain areasin response to central nervous system drugs thataffect basal metabolismorblood flow. Although BOLD-image contrast is enhanced athigh magnetic fields, the effect isobservedat4.7T*, afield strength that is close to the highest field strength (4 T) presently available for human subjects.

Many MRI methods are being explored as alternative approaches to positron emission tomography (PET) in the study of regional brain activity, including methods that measure regional blood flow and volume using exogenous contrastagents(7). PETimagingrelieson afamilyoftracer

methodsformeasuring different physiological quantities

in-cluding blood volume, blood flow, and regional oxygen extraction (13). BOLD contrast adds to a similar,emerging set of functional MRI methodologies that are likely to be complementary to PET imaging in the study ofregional brain activity. Unlike most tracer methods, BOLD contrast relies onan intrinsic contrast agent and image acquisitioncan be precisely synchronized to external stimuli with good time resolution.Eventransient changes in blood oxygenation (14) can,in principle, be measured.

The underlying mechanism of BOLD contrast is the ac-centuationbygradient-echo imaging at highmagneticfields of effectsproduced by magnetic susceptibilityvariationthat is caused by an endogenous paramagnetic agent. The same mechanism may alsobe useful in the study of other iron-containing systems, including the spleen, liver, and bone marrow.

1. Axel,L.(1990) Magn.Reson.Med. 14,171.

2. Hanstock, C.C., Rothman, D. L., Prichard, J. W. & Jue, T. (1988) Proc.Nail. Acad. Sci. USA 85, 1821-1825.

3. Brown, T. R.,Kincaid, B. M. & Ugurbil, K. (1982) Proc. Natl. Acad. Sci. USA 79, 3523-3526.

4. Haacke, E. M., Masaryk, T. J., Wielopolski, P. A., Zypman, J. A.,Tkach, J. A., Amartur, S., Mitchell, J., Clampitt, M. & Paschal,C. (1990) Magn. Reson. Med. 14, 202-221.

5. Kim,S.-G. &Ackerman,J. J. H.(1990) Magn. Reson. Med. 14, 266-282.

6. LeBihan, D.(1990) Magn. Reson. Med. 14, 283-292. 7. Rosen, B. R., Belliveau, J. W., Vevea, J. M.& Brady, T.J.

(1990) Magn. Reson. Med. 14, 249-265.

8. Ogawa, S.,Lee, T. M., Nayak,A.&Glynn,P.(1990)Magn. Reson.Med. 14, 68-78.

9. Ogawa,S. & Lee, T. M. (1990) Magn.Reson. Med.16, 9-18. 10. Siesjo, B. K.(1978) in Brain Energy Metabolism(Wiley,New

York), pp. 237-240.

11. Wehrli, F. W. (1990)Magn. Reson. Med. 14, 187-193. 12. Thulborn, K.R., Waterton, J. C.,Mathews, P. M. & Radda,

G. K.(1982) Biochim.Biophys.Acta714,265-270.

13. Sokoloff, L., ed. (1985) Brain Imaging and Brain Function (Raven, NewYork).

14. Frostig,R.D.,Lieke,E.E., Ts'o, D. Y.&Grinvald,A.(1990) Proc.Nati. Acad. Sci. USA 87, 6082-6086.

*Ogawa, S., Lee, T. M., Chu, S. & Foxall, D., Eighth Annual Meeting SocietyofMagneticResonanceinMedicine, August 12-18, 1989,Amsterdam,abstr. 1158.