A METHOD FOR THE CONTINUOUS MEASUREMENT OFCEREBRAL BLOOD FLOW IN MAN BY MEANS OF

RADIOACTIVE KRYPTON (KRT9) *

By BENJAMIN M. LEWIS,t LOUIS SOKOLOFF, RICHARD L. WECHSLER,W. BUDD WENTZ AND SEYMOUR S. KETY

(From the Aviation Medical Acceleration Laboratory, U. S. Naval Air Development Center,Johnsville, Pa., the Laboratory of Clinical Science, National Institute of Mental

Health, Bethesda, Md., and the Department of Physiology and Pharma-cology, Graduate School of Medicine, Philadelphia, Pa.)

(Submitted for publication March 26, 1959; accepted May 15, 1959)

The measurement of cerebral blood flow bymeans of the nitrous oxide method of Kety andSchmidt (1) or the krypton-85 (Kr85) techniqueof Lassen and Munck (2) requires a steady stateof at least ten minutes, for it is not until the end ofsuch a period that the uptake of the gas by thebrain can be estimated indirectly from its concen-tration in cerebral venous blood. These methodsare, therefore, not applicable to the study of rapidor transient changes in cerebral blood flow. If,however, a Remitting inert gas such as krypton-79(Kr79) is employed, its uptake by the brain canbe followed by direct measurement of the y-radia-tion through the intact skull, thus permitting thecontinuous measurement of cerebral blood flowduring transient or rapid changes before equili-bration between brain tissue and cerebral venousblood is achieved.The description of the development, evaluation

and application of such a method is the subject ofthis report.

METHODTheory

Like the nitrous oxide method, the Kr' method isbased upon the Fick principle, which can be expressedas follows:

dQB CBF (CA - Cv), [1)]dt

where dQB/dt is the time rate of change of the quan-tity of Kr'm in the brain, CBF is the rate of blood flowthrough the brain assuming equal arterial inflow andvenous outflow, and CA and Cv are the arterial and rep-

* Opinions or conclusions contained in this report arethose of the authors and are not to be construed as neces-sarily reflecting the views or endorsements of the NavyDepartment.

t Present address: Department of Medicine, WayneState University. College of Medicine, Detroit 7, Mich.

resentative cerebral venous Kr79 concentrations, re-spectively.

Since accurate measurement of instantaneous rates ofchange like dQB/dt is difficult, it is preferable to inte-grate Equation 1 between definite intervals, short enoughto keep the resolution time of the method small and topermit the approximation that cerebral blood flow hadremained constant during the interval but long enoughto allow accurate determinations of the change in quan-tity of Kr'm in the brain. Thus, in the integrated andtransposed form,

ZAIQB t

CBF = _ Itit2(CA - Cv)dt

[2)]

Equation 2 states that the cerebral blood flow rateduring a given interval of time is equal to the changein brain Kr" content divided by the integrated arterio-venous Kr" concentration difference during the sameinterval of time. By measuring brain content and ar-terial and venous concentrations either continuouslyor repeatedly at fixed successive intervals of tinre, as,for example, the one minute intervals in these studies,cerebral blood flow can be calculated continuously oversuccessive one minute periods until the rate of change ofQB and/or the arteriovenous difference become too smallto be determined reliably. Equation 2 applies duringsaturation or desaturation of the brain with Kr'. Fur-thermore, since QB represents total brain Kr7" contentand not concentration in the cerebral tissues, CBF is ameasure of the blood flow to the whole brain rather thanto a representative fraction of it, as is the case with thenitrous oxide method.

QB, CA, and Cv are determined by the counting of they-radiation of the Kr' and are expressed as counts perminute (cpm) in the brain or cpm per milliliter of blood,respectively. Since the counting rates in brain and bloodare determined under different geometric conditions and,indeed, by different counters, counting efficiencies aredifferent for these two types of measurements, and anequivalence factor (F) obtained by counting a standardKr" source under the two conditions in a manner to bedescribed below must be included in the numerator ofEquation 2.

707

LEWIS, SOKOLOFF, WECHSLER, WENTZ AND KETY

0 2E46IN 1

E 16.0(MNU40

JUGLARVENUS R79CONENTATINSANDBANI7

12.0-_ 30

0C

0

TIME (MINUTES)

FIG. 1. TYPICAL CURVES FOR ARTERIAL AND INTERNAL

JUGULAR VENOUS KR"9 CONCENTRATIONS AND BRAIN KR'S

CONTENT DURING SATURATION WITH THE RADIOACTIVE GAS.

The formula and the method for the calculation of CBFduring a one minute interval (fifth minute) are illus-trated. AQB during the fifth minute is the numerator ofthe formula and is obtained as indicated; the shaded area

between the arterial and cerebral venous curves repre-

sents the integrated arteriovenous difference during thesame interval of time, and is the denominator of theformula. F is a proportionality constant which convertsblood and brain counting rates to equivalent units (seetext). To obtain continuous minute-to-minute valuesfor cerebral blood flow, the calculation is repeated foreach successive one minute interval of time.

The application of this equation to the calculation ofcerebral blood flow is illustrated in Figure 1.

Properties, preparation and administration of Kr79.Kr79 was chosen as the tracer substance for this methodbecause it is a chemically inert gas which diffuses freelyacross the blood-brain barrier, emits -y-radiation whichpermits its direct measurement in brain by means ofexternal counting techniques, and has a suitably shortphysiological half-life. Its physical half-life is 34 hours(3), and its decay scheme includes K-capture, positronemission, and at least 15 -y-transitions ranging in energy

from 44.5 to 833.4 Kev (4). It was produced for thesestudies by deuteron bombardment of potassium bro-mide in a cyclotron 1; the Kr79 formed by the ensuingBr-d-2n reaction is trapped in the crystal structure of thebromide. Other radioactive isotopes (Br' and Br'2) are

also formed in the process. The occluded Kr79 is re-

leased by solution of the potassium bromide in brominewater in an enclosed atmosphere containing air and a

barely measurable volume of inert carrier krypton. Thebromine water is then completely distilled into a dryice-acetone cold trap; the overlying gas is passed

1 Kr79 was produced for these studies in the CyclotronLaboratory of the Massachusetts Institute of Technol-ogy.

through a sodium hyposulfite scrubber to remove anyresidual radioactive bromine and finally collected in anevacuated 1 L tonometer. The yield of Kr79 was assayedby measurement of the -y-radiation from the tonometerat a distance of 100 cm with an ionization chamber typesurvey meter which had previously been calibrated witha known Kr79 source.2 The quantity of Kr"9 obtainedin the tonometer generally approximated 5 mc.The Kr79 was administered to the subjects in the in-

spired air which was prepared by flushing the contentsof the tonometer into a spirometer and diluting withroom air to a final concentration of 100 uc per L. Ad-ministration was usually through a non-rebreathing sys-tem with collecton of the expired air in a Douglas bag,but in a few cases to be discussed below, rebreathingwith CO2 absorption and 02 replacement was employed.Under the conditions of these experiments, the highestradiation dosage is in the lungs and has been calculatedto be less than 0.1 rad (5).Measurement of Kr79 content in brain. The quantity

of Kr79 in the brain is measured by means of a scintil-lation counter precisely located with respect to the brainand carefully shielded to exclude radiation from all bodilyareas except the brain and overlying extracerebral tis-sues. The scintillation probe consists of a cylindrical,thallium-activated sodium iodide crystal, 4 cm in diam-eter and 10 cm in length, a Dumont 6292 photomultipliertube, and a cathode-follower type pre-amplifier, all en-closed within a light-proof aluminum shield. With thesubject in the supine position, the probe is placed later-ally with respect to the head so that its proximal endis approximately 25 cm from the midsagittal plane ofthe head and its longitudinal axis is directed at the mid-point of the line connecting the opening of the externalauditory meatus and the top of the crown (Figure 2).The entire probe is surrounded by a 2 inch thick leadcylinder, 12 inches in diameter (Figure 2). Betweenthe side of the head and the proximal end of the leadcylinder, but flush against both, is a 2 inch thick flatlead plate which is cut in the shape of the contours ofthe brain as "seen" by the scintillation probe (Figure 2).The lead shield was designed to exclude from the probeall Kr79 radiation from the room, body and head exceptthat of the brain and directly overlying extracerebraltissues; its design was determined by experiments witha plastic model of the head containing multiple com-partments representing brain, extracerebral tissues, andrespiratory passages which could be filled with Kr79solutions or gas mixtures as indicated.The output pulses of the scintillation probe are first

amplified (Atomic Instrument Co. Linear AmplifierModel 204-B) and then counted' in a decade scaler(Atomic Instrument Co. Model 1070A) equipped witha pulse height discriminator adjusted to the minimumlevel necessary to exclude all tube noise and other elec-tronic artifacts. Continuous recording of the counting

2 This Kr79 source was standardized for us by Dr.Sherman Frankel, Department of Physics, Universityof Pennsylvania.

708

CONTINUOUS MEASUREMENT OF CEREBRAL BLOOD FLOW IN MAN

background and also for radioactive decay from thetime of the head counting procedure.Determination of the factor F. The factor, F,

described previously in the discussion of the theory,is actually a proportionality constant between the over-all brain and blood counting efficiencies and can be de-fined as follows:

F = B [3)]

FIG. 2. DIAGRAMMATIC REPRESENTATION OF EXPERI-MENTAL SETUP. A, overall view of physical arrange-

ment of subj ect, lead shields and apparatus; B, viewof head as seen by scintillation counter in lead shieldwith the X marking the point at which the longitudinalaxis of the scintillation crystal is directed.

rate is achieved by photography of the scaler face andtimer or by differentiation of the scaler output by means

of a count rate meter, the output of which is recorded on

an Esterline-Angus recording milliammeter.Measurement of Kr'9 concentration in blood. If three

complete counting channels are available, it is possibleto monitor the arterial and cerebral venous Kr79 con-

centrations continuously and simultaneously with thecounting from the brain. This can be accomplished bythe continuous sampling of blood at a constant ratethrough glass helices fixed in the wells of well-typescintillation counters. In the present studies, however,discrete, timed arterial and cerebral venous blood sampleswere drawn into oiled, heparinized syringes at intervalsas close to one minute as possible, and their Kr' con-

centrations were assayed subsequently by measurementof the counting rates in aliquots of each of them trans-ferred consecutively to a single glass helix-scintillationcounter system. Each blood sample was mixed by shak-ing with mercury and introduced into the helix via one

of the two-way stopcocks attached by plastic tubingto each opening of the helix. These stopcocks were

closed when the system was filled, and by transferringan excess of blood through the helix, anaerobic condi-tions were maintained for the sample being assayed.The use of the helix also assured a constancy of samplevolume and geometric conditions. Between samplesthe helix was flushed with water, then isopropyl alcohol,and finally dried by a current of air. Sufficient samplecounts were collected to yield a coefficient of variationof less than 3 per cent, and corrections were made for

where B equals the counting rate from a given quan-tity of Kr'9 in blood counted under the blood countingconditions and S equals the counting rate from the samequantity of Kr79 distributed throughout the brain andcounted under the brain counting conditions.

Since S cannot readily be determined in situ, a Plexi-glas phantom model of the head was constructed whichcontained separate "brain," "airway," and "extracerebral"compartments and conformed to average anthropologicalmeasurements for adult Caucasian males. The phan-tom model was utilized in the following manner. A 10ml syringe containing 5 ml of an aqueous Kr' solutionwas counted at various distances between 20 and 95 cmfrom the end of the scintillation crystal employed in themeasurement of Kr"9 in the brain. The contents of thesyringe were than quantitatively transferred into thewater-filled "brain" compartment of the Plexiglas model;the "extracerebral" compartment was filled with Kr79-free water. Counting rates from the brain model werethen measured at various distances from the scintillationcrystal which was directed at the same point in themodel as in the measurements from the head describedpreviously (Figure 2). The ratio of "brain" to sy-ringe counting rates was thus determined at various dis-tances between the center of each and the end of theprobe. It was found not to change systematically be-tween 25 and 95 cm and to average 0.67 over that range(Figure 3). Linear absorption coefficients for the over-all y-radiation of Kr"9 in water, Plexiglas, brain, andcalvarium were also experimentally determined; theirvalues indicated that for the distances involved dif-ferences in self-absorption in the model and human headwere negligible. It was concluded, therefore, that underthe counting conditions described, a given quantity ofKr"9 distributed in the brain yields a counting rate equalto 67 per cent of its rate at the same distance whendistributed in 5 ml of solution contained in a 10 mlsyringe. This information makes possible the deter-mination of S in Equation 3 and, therefore, also the fac-tor, F.

Since the efficiencies of the scintillation counters mayvary from day to day, F must be determined for eachexperiment. The procedure is as follows. A concen-trated solution of Kr"9 in blood is made by equilibratinga sample of the subj ect's blood with a small sample ofthe undiluted Kr"9-air mixture. Five ml of this solu-tion is drawn into a 10 ml syringe; gas bubbles arecarefully expelled, and the syringe is capped. Its con-tents are referred to in subsequent discussion as the"standard" Kr"9 solution. The syringe is placed at pre-

709

LEWIS, SOKOLOFF, WECHSLER, WENTZ AND KETY

cisely the same distance from the end of the scintillationcrystal as was the midsagittal plane of the head duringthe cerebral blood flow measurement, and its countingrate is determined under the same conditions and withthe identical equipment employed in the measurement ofbrain Kr79 content. A sample of the standard Kr79 solu-tion is then transferred to the glass helix and countedwith the same equipment and in the same manner as theexperimental blood samples.

If Nh and N. are the counting rates obtained from thestandard solution in the helix and in the syringe, re-spectively, and Vh and V. are the corresponding vol-umes of the solution in the helix and the syringe, thenfrom Equation 3 and the results of the experiments withthe phantom "brain" model,

B - Nh/Vh Nh V, [)vV F = _ = h/ E4)]S 0.67 N,/V, 0.67 N.Vh

Since the arterial and cerebral venous concentrationsare determined directly in counts per minute per helixvolume, the helix volume, Vh, cancels out when F isintroduced into Equation 2.The final form of the equation from which CBF is

calculated is then as follows:

IQB t - X Nh VsCBF =- E

(CA' - Cv') X 0.67 Nj

where CA' and Cv' are the arterial and cerebral venousconcentrations in counts per minute per helix volume.

ProcedutreAll studies reported here were performed in normal

young men in the supine position. Arterial and cerebralvenous blood was sampled as previously described (1).After establishment of the sampling systems, the headwas carefully placed with respect to the scintillationprobe and lead shield as described above.

In all studies Kr79 concentration in the inspired air wasadjusted to approximately 100 uc per L by dilution in a120 L spirometer with room air or a mixture of 15 percent N2O, 64 per cent N2, and 21 per cent 02. Thenitrous oxide mixture was employed in several studiesin which cerebral blood flow was measured simultane-ously by the nitrous oxide technique (1) for the pur-pose of comparison with the Kr79 method. Administra-tion was usually through a non-rebreathing respiratorysystem. A rebreathing system was used only when theKr79 was insufficient to dilute to the specified concentra-tion and volume necessary to sustain respiration for atleast 10 minutes. Cerebral blood flow was then meas-ured only during the desaturation period following 10minutes of rebreathing of the inspired gas mixture.During the rebreathing, CO2 was removed by a baro-lime absorber in the expiratory circuit, and oxygen wasfed into the spirometer to maintain constant the 02tension monitored by a Pauling oxygen analyzer in theinspiratory pathway. A three-way stopcock in the

inspired air line permitted rapid switching between roomair and spirometer gas as required; a similar stopcockin the expiratory pathway permitted switching to aDouglas bag during non-rebreathing or to the spirometerduring rebreathing (Figure 2). To facilitate shieldingfrom the scintillation probe, a rubber mouthpiece ratherthan a mask was employed for administration of the gas,and the nasal passages were blocked by a nose clip.Dead space was minimized by valves in the inspiratoryand expiratory pathways close to the mouthpiece. Thespirometer and Douglas bag were placed behind leadX-ray shields as far as possible from the scintillationprobe to reduce the contribution to the background fromthe relatively large quantities of Kr79 contained withinthem (Figure 2).During all periods of measurement, breathing of the

appropriate inspired air, counting from the head, andsampling of arterial and cerebral venous blood werebegun simultaneously at the zero time for that period.The first pair of blood samples was drawn at a con-stant rate of 10 ml per minute for the entire first min-ute to provide integrated values over that period. There-after, timed simultaneous arterial and venous blood sam-ples of approximately 5 ml each were drawn as rapidlyas possible (approximately 10 seconds) at intervals of1 to 1.5 minutes. Measurements during saturationlasted 10 minutes from the onset of breathing the Kr79gas mixture; measurements during desaturation lasted10 minutes from the onset of inhalation of room airfollowing a 10 minute period of breathing the Kr79 gasmixture.

In order to test the ability of the method to detectchanges in cerebral blood flow, several studies wereperformed during the infusion of l-norepinephrine, thebreathing of 7 per cent C02, or hyperventilation. Theseconditions were introduced only during the desaturationperiod after. several minutes of resting control valueshad been obtained.Blood oxygen and carbon dioxide contents were ana-

lyzed manometrically by the method of Van Slyke andNeill (6). Blood pH was determined anaerobically atambient temperature by means of a glass electrode andBeckman Model G potentiometer and corrected to 37° Cby the factors of Rosenthal (7). Blood pCO2 was cal-culated according to the nomogram of Peters and VanSlyke (8). Mean arterial blood pressure was measuredby means of an air-damped mercury manometer con-nected to the femoral arterial manifold and adjusted tocarotid artery level.

RESULTS

By means of the Kr79 method, it was possibleto follow total cerebral blood flow minute by min-ute. In Table I are presented the results of 14studies in 10 normal young men in which measure-ments were made continuously during steadystate periods of approximately 10 minutes. Sixof the studies were performed during saturation

710

CONTINUOUS MEASUREMENT OF CEREBRAL BLOOD FLOW IN MAN

and eight during desaturation. During the firstminute or two after the onset of saturation or

desaturation, the quantity of Kr79 in the lungs ischanging rapidly. Since it is enormously greaterthan the amount in the brain, even the small frac-tion of its scattered radiation which bypasses thelead shield significantly increases the rate ofchange of the head counting rate and leads to ar-

tifactually high values for cerebral blood flow.By the third minute, however, the lung Kr79 con-

tent is usually changing slowly and only negligiblyinfluences the slope of the head counting curve.

After the eighth minute, the head counting curve

frequently becomes so flat and the arteriovenousdifference so small that reading errors may greatlyinfluence the calculated values for blood flow.Therefore, only the values between the third andeighth minutes were considered reliable and theseare included in Table I. Over this period the min-ute-by-minute values for cerebral blood flow va-

ried relatively little; the overall standard devia-tion within studies was equal to 72 ml per min-ute or about + 6 per cent of the mean, and thevariance within studies was significantly less thanthe variance between them (p < 0.001). The to-tal mean of all 14 studies was 1,247 ml per minute(SD = 277). Mean cerebral blood flow in theten subjects was 1,236 ml per minute (SD = +246). The results of saturation and desaturationstudies were not significantly different. A slightbut systematic upward drift of the values withtime was observed in both the saturation anddesaturation studies, but it was considerably more

prominent in the former and may, perhaps, haveresulted from CO2 accumulation in the dead space

of the respiratory system.The Kr79 and nitrous oxide methods were

compared simultaneously in nine experiments ineight subjects; the inspired air mixture then con-

tained 15 per cent N20 in addition to the Kr79,

TABLE IMinute-to-minute cerebral blood flow in normal young adult men

Cerebral blood flow (mI/mix)Type During following time in minutes:of

Subj. study* 3 4 5 6 7 8 Mean ± SD

GM S 946 1,042 1,074 957 1,073 1,018 ± 59NP S 1,156 1,156 1,214 1,296 1,327 1,311 1,243 ± 74JB S 1,599 1,632 1,555 1,519 1,509 1,443 1,542 71DT S 1,188 1,182 1,337 1,250 1,352 1,262 i 70HC S 1,012 1,038 1,197 1,294 1,276 1,209 1,171 i 109HC D 864 771 841 817 828 863 827 4 69JL S 936 983 986 938 1,027 1,083 992 i 53JL D 1,033 932 897 858 866 879 911 ± 58GD D 1,000 1,133 1,116 1,084 1,090 1,112 1,089 i 45SK D 995 1,027 1,090 1,143 1,169 1,291 1,114 ± 99JS D 1,553 1,643 1,631 1,627 1,726 1,633 1,635 ± 58JS D 1,438 1,362 1,395 1,430 1,436 1,423 1,413 4 46EP D 1,605 1 695 1 675 1,592 1,536 1,527 1,605 4 72EP D 1,505 1,539 1,640 1,730 1,680 1,700 1,632 4- 81

Total 14 studiesMean 1,202 1,224 1,261 1,252 1,278 1,289 1,247SD +277

10 SubjectsMean 1,186 1,216 1,252 1,239 1,271 1,290 1,236SD 4246

6 Saturation studiesMean 1,140 1,172 1,227 1,209 1,261 1,262 1,205SD 4200

8 Desaturation studiesMean 1,249 1,263 1,286 1,285 1,291 1,304 1,278SD ±334

* S = saturation, D = desaturation.

711

LEWIS, SOKOLOFF, WECHSLER, WENTZ AND KETY

TABLE II TABLE III

Comparative cerebral blood flow values calculated by indirect Comparison between the results of the N20 and direct Kr79method from integrated cerebral arteriovenous N20 and methods for the determination of cerebral blood flow

.1ST9 W1.yt~f177H..tAr aiferencs-r

Cerebral blood flow(mil/OO g/min)

Subj. N20 Kr79

GD 55 66HC 58 59SK 56 60JL 54 49JL 51 50GM 66 58NP 48 48JB 68 60DT 62 65

Mean 57.5 57.3

Mean overall diff. 41 SD = 4.6 + 3.8

and measurements were made during a steadystate of at least ten minutes as required by thenitrous oxide technique (1). Cerebral blood flowwas calculated in two ways: 1) indirectly from theintegrated arteriovenous N20 difference over theentire ten minute period as described by Kety andSchmidt (1); and 2) indirectly from the inte-grated arteriovenous Kr79 difference exactly aswith nitrous oxide. The brain: blood partitioncoefficient for krypton employed in the latter cal-culations was 1.06, the average value found by

Cerebral blood flow

N20 Direct Kr79Subj. method method

ml/100 g/min ml/100 ml/minHC 58 86HC 55* 60SK 56 80JL 54 69JL 51 86JL 51 63GM 59* 71NP 48 82JB 68 99DT 62 81JS 64* 104JS 58* 89EP 58* 96EP 57* 98

Mean 57.1 83.1 tSD d5.3 4 13.7

Coefficient of correlation = 0.55 (p < 0.05)

* Calculated from integrated cerebral arteriovenous Kr79differences by indirect method.

t Significantly different statistically as determined bymethod of paired comparison (p < 0.001).

Lassen and Munck (2) in subj ects with normalhematocrits. Almost identical values for cere-bral blood flow were obtained with both gaseswhen calculated in this manner (Table II); there-

43

10 20 30 40 50 60

Distance (cm )

4

70 80 90 100

FIG. 3. RATIO OF COUNTING RATE FROM A GIVEN QUANTITY OF KR79 DISTRIBUTED

THROUGHOUT SOLUTION IN "BRAIN" COMPARTMENT OF PLASTIC HEAD MODEL TO COUNT-ING RATE AT SAME DISTANCE FROM SAME QUANTITY OF KR79 CONCENTRATED IN 5 ML OF

SOLUTION IN A 10 ML SYRINGE AS A FUNCTION OF DISTANCE BETWEEN THE MIDSAGITTAL

PLANE OF THE HEAD OR THE MIDDLE OF THE SYRINGE AND THE PROXIMAL END OF THE

SCINTILLATION CRYSTAL (SEE TEXT).

040

aoff 1.00

4J 0.90

,, 0.80U44bD 0.70

$k .600%

o 0.504.'

C 0.40-44

54m 0.30

' 0.200*A o. 10'4940

No. of 3 1Determinattons

712

Kl.,-l

CONTINUOUS MEASUREMENT OF CEREBRAL BLOOD FLOW IN MAN

fore, both methods were considered equally validfor comparison with the direct minute-to-minutemethod which was applied simultaneously in sevenof the above experiments as well as in seven addi-tional studies, including six in which Kr79 alonewas employed. In order to compare the indirectmethod, which measures blood flow per 100 g ofbrain, taken as a whole, per minute, with the di-rect method, which yields total cerebral bloodflow per minute, endocranial volume was calcu-lated from extracranial measurements by themethod of Lee and Pearson (9). No correctionswere made for endocranial volume not occupiedby the brain or for the specific gravity of thebrain. The average of the minute-to-minute bloodflow values between the third and eighth minutesper 100 ml endocranial volume was then com-pared with the blood flow per 100 g brain perminute obtained by the indirect method over theentire teRfminute period (Table III). The valuesobtained with the direct method uniformly ex-ceeded those of the indirect technique (mean val-ues equal 83.2 ml per 100 ml per minute and56.9 ml per 100 g per minute, respectively), butthe correlation between them was statistically sig-nificant (R = + 0.55; p < 0.05).

In five studies cerebral blood flow was experi-mentally altered in the midst of the measurementsto test the ability of the Kr79 method to detect andfollow rapid changes. All measurements weremade during desaturation, and the experimentalconditions were imposed after a two minute con-trol period. The studies included two on the ef-fects of breathing 7 per cent CO2 and one eachon the effects of hyperventilation of room air,hyperventilation of 100 per cent 02, and the intra-venous infusion of l-norepinephrine at a rate ofapproximately 16 lug per minute. The results

TABLE IV

Comparative changes in cerebral blood flow during variousprocedures as measured by direct Kr79 and N20 methods

Per cent change incerebral blood flow

Kr7'9 N20 methodProcedure method (ref. no.)

Inhalation, 7% CO2 +86 +75 (10)Hyperventilation, room air -31 -34 (10)Hyperventilation, 100% 02 -33I-Norepinephrine infusion -28 - 8 (11)

are graphically represented in Figure 4 and dem-onstrate the ability of the Kr"9 method to followcontinuous changes in cerebral blood flow. InTable IV are compared the maximum percentagechanges in cerebral blood flow caused by theseprocedures with those previously observed in ni-trous oxide method studies (10, 11). Althoughthe experimental conditions cannot be comparedprecisely, the direct Kr79 method appears to be atleast as sensitive as the nitrous oxide techniqueand to yield proportionate results.

DISCUSSION

Total cerebral blood flow values obtained withthe Kr79 method exceed those to be expectedfrom the results of the nitrous oxide techniquein a representative fraction of brain and the brainweight. This discrepancy is probably attributableto the collective effect of a number of factors.Although carefully shielded from the remainderof the body, the scintillation counter measuringbrain Kr79 content must also detect radiation fromthe extracerebral tissues overlying the brain.Changes in Kr7 9 content of these tissues con-tribute to the slope of the head count curve and,therefore, also to the calculated values for cerebralblood flow. In order to evaluate the contributionfrom the overlying soft tissues, a tourniquet wasapplied to the head just above the ears, and thesubject was allowed to breathe Kr79. Aftersix minutes the tourniquet was released; despitea noticeable reactive hyperemia of the scalp, therewas no significant change in the slope of the headcount curve. These results indicate negligiblecontributions from the scalp probably because ofits smaller volume and lower blood flow thanthe brain; the contribution from the skull remainsunevaluated but is probably similar. The extra-cerebral tissues may, perhaps, account for someof the discrepancy between the Kr79 and nitrousoxide methods but apparently do not obscure themeasurement of changes in cerebral blood flow.For example, the directions and degrees ofchange elicited by hyperventilation and carbondioxide inhalation are almost identical to thoseobserved with the nitrous oxide technique (TableIV), despite the fact that these procedures causeopposite changes in extracerebral blood flow(12).

713

LEWIS, SOKOLOFF, WECHSLER, WENTZ AND KETY

I-

l

U I 2 3 4 5 bTIME- MINUTES

A

4,4g/CC CarNOR-EPINEPHRINE

MBP

Boo-

, - PR04-4

d

L 2

o 2 3 4 5 6 7TIME- MINUTES

B

1400

a1200

1000

UI.800

a0

0° 600us 400 440_mEsow Jw 200 Z20_

0

_1200 ;

8

8 1000i~ iEEsElo 800 h35Jo0

K L0 iS4i600 7*O160

hi 7.451-uj 400jJS4

FIG. 4. MINUTE-TO-MINUTE VALUES FOR CBF AS DETERMINED BY THE KR"9 METHOD DURINGEXPERIMENTALLY INDUCED CHANGES IN CEREBRAL BLOOD FLOW. A, two studies during the in-halation of 7 per cent CO,; B, during the continuous intravenous infusion of 4 ,ug per mll-norepinephrine solution at a rate of approximately 4 ml per minute; C, during hyperventilationof 100 per cent 02 at a rate of approximately 30 L per minute; D, during hyperventilation ofroom air at a rate of approximately 30 L per minute.

Rapid changes in the Kr79 content of the lungsand respiratory passages also influence the slopeof the head count curve. This effect arises partly,perhaps, from incompletely shielded paranasalsinuses but mainly from scattered radiation fromthe chest. Although both sources represent a

very small fraction of the radioactivity in therespiratory system, they may still be significantcompared to the quantity in the brain because ofthe low tissue: gas partition coefficient (approxi-mately 0.05) of krypton (13). As previouslynoted, this effect is sufficient to invalidate meas-

urements made during the first and often secondminutes of saturation or desaturation and may

also contribute a little to the excessively highvalues for cerebral blood flow observed in subse-quent minutes. In one experiment a continuousrecord of chest radioactivity was made through-out the procedure, and its effect on the headcounts was calculated by means of a "scattering"factor determined in experiments with "phantomlungs." It was found to increase the calculatedvalue for cerebral blood flow during the three toeight minute period by 3.5 per cent.The greatest single cause of the high values

obtained with the Kr79 method is probably thedisproportionate weighting of different areas ofthe brain in the total head count. The overall

' 1800C,._US 1400390U. 1000a00

X 6004

0w 200hi oi 2 3 4 5 b6

TIME- MINUTES

C

-I000

s

.9 800

w 600

0

i 400

0 200

hiC.)

' 10o

EEt"Ia2

.Lso

0a

pH

1 2 3 4 5 6TIME- MINUTES

D

7

714

CBF

I

CONTINUOUS MEASUREMENT OF CEREBRAL BLOOD FLOW IN MAN

linear absorption coefficient for Kr79 y-radiationin brain tissue was measured and found to beapproximately 0.09 cm-' (one-half thickness = 8cm); there is, therefore, significant self-absorp-tion of the y-radiation within the brain. Becauseof this effect as well as the inverse square law,the Kr79 content in the cerebral tissues becomesprogressively less represented in the total headcounting rate in going from the surface of thebrain closest to the counter to the other side.It has been estimated on the basis of theoreticalconsiderations that, even with uniform Kr79 con-centration throughout the brain, the most proxi-mal 1 cm thickness of brain tissue would aloneaccount for at least 18 per cent of the total count-ing rate or three times the amount to be expectedif all areas were equally represented. Since themost superficial and, therefore, most efficientlyrepresented areas contain disproportionately highamounts of gray matter compared to the brainas a whole, the head counting rate is excessivelyweighted in favor of the cerebral cortical tissueswhich have considerably higher rates of bloodflow than most other areas of the brain ( 14).This effect undoubtedly accounts for much of thediscrepancy between the results of the Kr79 andNO methods; it may also render the methodsusceptible to errors arising from the redistribu-tion of blood flow between the deep and super-ficial areas of the brain.

Despite its defects, the Kr79 method performsquite well in studies of functions not otherwisemeasurable in man. It permits the estimation ofcerebral blood flow under nonsteady state condi-tions and, as illustrated in Figure 4, is capable offollowing rapid or transient changes in this func-tion. Although its descriptions of these changesmay not be quantitatively accurate, they are indi-cative of and probably proportional to those ac-tually occurring in the brain. It does not requirethe assumption of a constant cerebral metabolicrate as does the arteriovenous 02 differencemethod which can also be used for the detectionof rapid changes in cerebral blood flow (15). Infact, when combined with arteriovenous 02 dif-ferences, the Kr79 method can be employed to fol-low changes in cerebral metabolic rate as well.Many of the problems encountered in the de-

velopment of the Kr79 method are specifically re-

lated to the tracer substance and the form in whichit was administered. A nongaseous substancewhich fulfilled the necessary requirements andwhich could be injected intravenously would ob-viate the influence of the isotope in the lungs, airpassages and paranasal sinuses with considerablesimplification of the problem of shielding. Anti-pyrine iodinated with 1131 has been shown (16)to cross the blood-brain barrier rapidly and tohave a distribution coefficient between brain andblood of approximately unity (17); it appears tohave certain advantages with respect to thismethod which warrant further investigation.

SUMMARY

1. A method for the rapid continuous estima-tion of total cerebral blood flow in man by meansof radioactive krypton (Kr79) is presented; itstheoretical basis, developmental experiments, andprocedure are described.

2. Mean total cerebral blood flow in ten nor-mal young men was 1,236 (SD = ± 246); thestandard deviation among minute-to-minute val-ues within studies was ± 72 ml per minute orabout 6 per cent of the mean value.

3. Cerebral blood flow determined by the Kr79method and corrected for intracranial volume wasfound to be significantly higher than the valuesobtained simultaneously with the nitrous oxidemethod; in 14 experiments the mean values were83 ml per 100 ml intracranial volume per minuteand 57 ml per 100 g brain tissue per minute, re-spectively. The results of the two methods were,however, significantly correlated (R = + 0.55;p < 0.05).

4. Sources of error in the Kr7 9 method areevaluated, and the causes of the discrepancy be-tween the results of the nitrous oxide and Kr79methods are discussed.

5. The Kr79 method was found capable of fol-lowing rapid changes in cerebral blood flow suchas, for example, those induced by hyperventilation,CO2 inhalation, or l-norepinephrine infusion.

ACKNOWLEGM ENTS

The authors wish to express their gratitude to Dr.Sherman Frankel and Dr. Wilton M. Krogman forvaluable advice and suggestions, and to Mr. Gerald B.Pidcock, Mr. John Bowanko and Miss Phyllis Camp-bell for outstanding technical assistance.

715

LEWIS, SOKOLOFF, WECHSLER, WENTZ AND KETY

REFERENCES

1. Kety, S. S., and Schmidt, C. F. The nitrous oxidemethod for the quantitative determination of cere-bral blood flow in man: Theory, procedure andnormal values. J. clin. Invest. 1948, 27, 476.

2. Lassen, N. A., and Munck, 0. The cerebral bloodflow in man determined by the use of radioactivekrypton. Acta physiol. scand. 1955, 33, 30.

3. Hollander, J. M., Perlman, I., and Seaborg, G. T.Table of isotopes. Rev. mod. Phys. 1953, 25, 469.

4. Thulin, S. Studies in nuclear spectroscopy withelectromagnetically separated gaseous isotopes.II. Ark. Fysik 1955, 9, 137.

5. Wechsler, R. L. Effects of acceleration upon cere-bral metabolism and cerebral blood flow. PhaseI. Development of a new method for continuousmeasurement of cerebral blood flow in humansunder acceleration. U. S. Naval Air Develop-ment Center, Aviation Medical AccelerationLaboratory, Johnsville, Pa., Aug. 1952, Reportno. NADC-NA-5202.

6. Van Slyke, D. D., and Neill, J. M. The deter-mination of gases in blood and other solutionsby vacuum extraction and manometric measure-ment. J. biol. Chem. 1924, 61, 523.

7. Rosenthal, T. B. The effect of temperature on thepH of blood and plasma in vitro. J. biol. Chem.1948, 173, 25.

8. Peters, J. P., and Van Slyke, D. D. QuantitativeClinical Chemistry. Methods. Baltimore, Wil-liams and Wilkins, 1932, vol. II.

9. Lee, A., and Pearson, K.-V. Data for the problemof evolution in man.-VI. A first study of the cor-

relation of the human skull. Phil. Trans. A 1901,196, 225.

10. Kety, S. S., and Schmidt, C. F. The effects of al-tered arterial tensions of carbon dioxide and oxy-gen on cerebral blood flow and cerebral oxygenconsumption of normal young men. J. clin. In-vest. 1948, 27, 484.

11. King, B. D., Sokoloff, L., and Wechsler, R. L. Theeffects of l-epinephrine and l-nor-epinephrine uponcerebral circulation and metabolism in man. J.clin. Invest. 1952, 31, 273.

12. Schmidt, C. F., and Hendrix, J. P. The action ofchemical substances on cerebral blood vessels.Res. Publ. Ass. nerv. ment. Dis. 1937, 18, 229.

13. Lawrence, J. H., Loomis, W. F., Tobias, C. A., andTurpin, F. H. Preliminary observations on thenarcotic effect of xenon with a review of valuesfor solubilities of gases in water and oils. J.Physiol. 1946, 105, 197.

14. Landau, W. M., Freygang, W. H., Jr., Rowland,L. P., Sokoloff, L., and Kety, S. S. The localcirculation of the living brain; values in theunanesthetized and anesthetized cat. Trans.Amer. neurol. Ass. 1955, 80, 125.

15. Patterson, J. L., Jr., and Cannon, J. L. Posturalchanges in the cerebral circulation, studied bycontinuous oximetric and pressure-recordingtechniques (abstract). J. clin. Invest. 1951, 30,664.

16. Sapirstein, L. A., and Hanusek, G. E. Cerebralblood flow in the rat. Amer. J. Physiol. 1958, 193,272.

17. Hansen, D. B. Unpublished studies.

SPECIAL NOTICE TO SUBSCRIBERS

Post Offices will no longer forward the Journal when you move.Please notify The Journal of Clinical Investigation, Business

Office, 333 Cedar Street, New Haven 11, Conn., at once when youhave a change of address, and do not omit the zone number ifthere is one.

716