Brain Blood Flow Measured with Intravenous H215O. II. Implementation and Validation

Abstract

We have adapted the well-known tissue autoradiographic technique for the measurement of regional cerebral blood flow (CBF), originally proposed by Kety and his colleagues, for the measurement of CBF in human subjects using positron emission tomography (PET) and intravenously administered oxygen-15-labeled water. This report describes the steps necessary for the implementation of this PET/autoradiographic technique. In order to establish the accuracy of the method, we measured CBF with intravenously administered oxygen-15-labeled water and PET in anesthetized adult baboons and compared the results with blood flow measured by a standard tracer technique that uses residue detection of a bolus of oxygen-15-labeled water injected into the internal carotid artery. The correlation between CBF measured with PET and the true CBF for the same cerebral hemisphere was excellent. Over a blood-flow range of 10-63 ml/(min X 100 g), CBF (PET) = 0.90 CBF(true) + 0.40 (n = 23, r = 0.96, p less than 0.001). When blood flow exceeds 65 ml/(min X 100 g) CBF was progressively underestimated due to the known limitation of brain permeability to water.

Full text

In 1951, Kety presented a theory of inert-gas cx
change at the lungs and tissues and its possible applica
tion to the measurement of tissue blood flow (1). A short
time later he and his colleagues described in detail (2,3)
a tissue autoradiographic technique for the measurement
of local brain blood flow in laboratory animals based on
this theory of inert-gas exchange. This autoradiographic
technique involved the intravenous infusion of radiola
beled trifluoroiodomethane. At the end of 1 mm the
animal was killed and the brain was removed and sliced
in the desired plane for quantitative autoradiography.
These quantitative autoradiograms and the time-activity
curve of the blood radioactivity formed the basis for the
computation of local brain blood flow using Kety's
equations (1 ). This technique has been widely applied
and revised on several occasions (4,5) as newer ra
diotracers were substituted for the volatile and diffi
cult-to-handle trifluoroiodomethane.
Received Dec. 21, 1983; revision accepted Apr. 28, 1983.
For reprints contact: M. E. Raichle, MD, Box 8I31, Washington
University School of Medicine, St. Louis, MO 63110.
Because positron emission tomography (PET) pro
vides a quantitative measure oflocal tissue radioactivity
and, hence, an in vivo autoradiogram (6), we reasoned
that local cerebral blood flow could be measured in
human subjects using the principles of the Kety auto
radiographic technique in conjunction with PET and a
freely diffusible, inert pharmaceutical labeled with a
positron emitter. In Part I (7) we described the theo
retical basis for our adaptation of the Kety approach for
PET and considered the impact of several potential er
rors. In this paper we describe the implementation of this
technique using H2150 as the diffusible tracer, and
present direct experimental evidence concerning its ac
curacy.
A preliminary report of this work has been presented
(8).
METHODS
Animal preparation. CBF was measured in six adult
baboons (Papio papio) weighing 18 to 25 kg. To facili
tate the injection of small aliquots (approximately 0.2
ml) of the baboon's blood labeled with oxygen-IS-water
790 THE JOURNAL OF NUCLEAR MEDICINE
BrainBlOOdFlowMeasuredwithIntravenousH2150.
II.ImplementationandValidation
M. E. Raichle, W. R. W. Martin, P. Herscovltch, M. A. Mintun, and J. Markham
Washington University School of Medicine, St. Louis, Missouri
We have adapted the well-known tissue autoradiographlctechniquefor the
measurement of regional cerebral blood flow (CBF), orlginaliy proposed by Kety
and hiscolleagues,for the measurementof CBF In humansubjectsusingpositron
emission tomography (PET) and Intravenously administered oxygen-15-labeled
water. This report describesthe steps necessaryfor the Implementationof this
PET/autoradiographictechnique.Inordertoestablishtheaccuracyofthemethod,
we measuredCBF with IntravenouslyadmInIsteredoxygen@15-labeledwater and
PETin anesthetizedaduftbaboonsandcomparedthe resultswithbloodflowmea
suredbya standardtracertechniquethatusesresiduedetectionofa bolusofoxy
gen-15-Iabeled water injected Into the internal carotid artery. The correlation be
tween CBF measured wIth PETand the true CBF for the same cerebral hemisphere
was excellent. Over a blood-flowrange of 10—63mll(mIrrlOO g), CBF (PET)
0.90 CBF(true) + 0.40 (n = 23, r 0.96, p < 0.001). When blood flow exceeds 65
ml/(mln•100g) CBFwasprogressivelyunderestImatedduetothe knownlImitatIon
of brainpermeabilityto water.
J Nuci Med 24: 790—798,1983

CLINICAL SCIENCES
INVESTIGATIVENUCLEAR MEDICINE
into the internal carotid artery, the baboons were anes
thetized with phencyclidine (2 mg/kg) at least 2 wk
before experiment proper, and the right external carotid
artery was ligated at its origin from the common carotid.
At the later experiment the radiotracer was injected into
the common carotid artery through a small catheter
(0.021 cm diam) positioned there under fluoroscopic
control.
For the actual comparison of the PET autoradiogra
phic method (7) with the standard residue detection
technique using the intracarotid administration of oxy
gen-l 5-labeled water (9) the baboons were anesthetized
with ketamine (2 mg/kg), paralyzed with gallamine,
intubated with a cuffed endotracheal tube, and passively
ventilated on a gas mixture containing 70% nitrous
oxide and 30% oxygen. The baboons were then posi
tioned on a special couch that permitted placement of
the head in the imaging device or over a 3 in. X 2 in.
NaI(Tl) scintillation detector appropriately collimated
and positioned under the animal's head so as to ensure
essentially uniform detection of a single cerebral hemi
sphere.
To permit the intracarotid injection of oxygen-l5-
labeled water, a small catheter was inserted percuta
neously into the femoral artery and its tip positioned in
the right common carotid artery under fluoroscopic
control. To prevent clotting in this arterial catheter
system, which was used for the injection of the labeled
water, monitoring of blood pressure, and sampling of
arterial blood, all animals were heparinized at the be
ginning of the experiment. Arterial pH, PCO2,and P02
were measured before and after each injection of tracer.
To permit the intravenous injection of oxygen-i 5-labeled
water, a small venous catheter was placed percuta
neously in the femoral vein.
Positron emission tomography was performed with
the PETT VI system (10), the design and performance
characteristics of which have been discussed elsewhere
(1 1,12). Data are recorded simultaneously from seven
slices with a center-to-center separation of 14.4 mm. All
studies were done in the low-resolution mode, giving an
in-plane (i.e., transverse) resolution of 11.7 mm full
width at half maximum (FWHM) in the center of the
field of view and a slice thickness of 13.9 mm FWHM
at the center of the image.
The head of the baboon was positioned with the aid
of a vertical laser line such that the center of the lowest
slice corresponded to a line running through the center
of the cerebral hemispheres. A lateral skull radiograph
with this line marked by a vertical radio-opaque wire
provided a permanent record of the positionof the lowest
PET slice. Because of the size of the adult baboon brain
(approximately 150 cc) only data from this bottom slice
were used in these experiments. Attenuation correction
was uniquely determined for each animal by obtaining
a transmission scan using a ring source of activity
(germanium-68) fitted to the tomograph as pre
viously described (10).
Single-probe data collection. The signal from the single
NaI(Tl) scintillation detector produced by the intra
carotid injection of H215O was processed by a pulse
height discriminator with a 60-keV energy window
symmetrically bracketing the annihilation peak to
minimize scattered radiation. The accepted events
(counts) from this single detector were stored in a small
laboratory computer. Processing of these data was per
formed in the computer, including corrections of the
count rate for deadtime losses, physical decay of 0-1S
(T112 123 sec), and background, and conversion to a
time-activity plotout. Optimal temporal resolution was
achieved in the initial portion of the recording by using
sampling integration times of 0.1 sec. Statistically
smooth recordings were ensured by injection of sufficient
activity into the carotid artery to achieve peak counting
rates from 10,000 to 20,000 cps.
PET data collection. For the measurement of CBF
with PET a 40-sec emission scan (whose modification
to permit longer scan times is discussed below) was
performed following an intravenous, bolus injection of
5 ml of saline containing 20-30 mCi of oxygen-I 5-la
beled water. Data collections by PET were started at the
time of arrival of radioactivity in the brain as judged by
a sudden increase in the bank pair coincidence counting
rate of the PETT VI system. This was usually 10—i5 sec
after tracer injection. Zero time for the study was always
the actual time tracer administration commenced. The
preparation of the radiowater has been described else
where (13). Arterial blood samples were drawn about
every 5 sec from the indwelling carotid catheter, starting
at the onset of injection and continuing throughout the
scan. These samples were weighed and counted in a well
counter to obtain 0-i 5 activity as cps/g blood, corrected
for physical decay from the time of injection to the time
of measurement. The time-activity curve was then con
structed.
Calibration of the tomograph to obtain the regional
0-1 5 concentration in the brain from the reconstructed
image (cps/cc tissue) was performed by imaging a
phantom divided into six wedge-shaped chambers of
equal size. The chambers were filled with varying con
centrations of carbon- 11-labeled bicarbonate. Aliquots
from each chamber were counted in the same well
counter used for the measurement of blood radioactivity,
and the observed counting rate decay-corrected to the
start of the phantom scan. From these data (Cinitjai, in
cps/cc), the total counts presented to the scanner by 1
cc of target during the length of the scan (Ctotai, in
counts/cc) were obtained by integrating the decay curve
over the length of the scan T5:
T@
Ctotal @1i: Cinitial exp(—kt)dt, (1)
791
Volume 24, Number 9

.—@ 5 5 25 35
Time from beginning of introcorotid injection (eec)
FIG.1. BraIntime-activitycurve(semilog)resultingfromintracarotid
Injectionof H21@Oinakift baboon. @aphloextrapolationof relatively
slow clearance of labeled tissue water back to maximum of perfu
sionpeakallowscalculationoffractionoflabeleçlwaterextracted
by tissue (E) durIng single capillary transIt (E B/A). Application
ofcentralvolumeprincipleoftracerkinetics(seeteXt)tOthesedata
allows computationof cerebral bloodflow (CBF),which we desig
nate here as CBF(true).DataInthisfigurewere obtainedat cerebral
blood flow of 102 ml/(mln.100 g).
PET system. The blood curve and scan data were
analyzed according to the general principles of inert-gas
exchange developed by Kety (1) and later embodied in
the tissue autoradiographic technique for the measure
ment of local brain blood flow in laboratory animals
(2—5).With this method the local blood flow is obtained
by numerically solving the following equation for the
constant K
C,(T) = AK S0T Ca(t) exp[—K(T—t)]dt
= AKCa(T)* exp(—KT) (6)
where C(T) is the instantaneous local radiotracer con
centration (cps/g) at time, T, derived from a quantitative
autoradiogram of a brain slice,and Ca(t) isthe measured
concentration of radiotracer in arterial blood as a func
tion of time, [cps/ml]; and A is the brain-to-blood
equilibrium partition coefficient for the tracer [ml/g].
Aequals0.95inourexperiment(9).Theoperationof
convolution is denoted by the asterisk. The constant K
is defined by (3):
K=mf/X (7)
where m is a constant between 0 and 1 that represents
the extent to which diffusion equilibrium between blood
and tissue is achieved during passage from the arterial
to the venous end of the capillary. As further defined by
Kety (3)
m 1 —exp(—PS/f) (8)
Thus, m is equivalent to E [Eqs. (7), (8)], a fact that will
become important in the analysis of the PET/autora
(5) diographic data (see below).
RAICHLE, MARTIN, HERSCOVITCH,MINTUN. AND MARKHAM
where k is the decay constant (per sec) for the tracer.
After the phantom image was reconstructed, a regression
equation was obtained comparing the relative scan data
and the directly measured activity in the phantom. From
this relationship, the actual local 0-1 5 concentrations
can be obtained from each scan.
Because our scanner does not correct for radioac
tive decay during the data collection, it is necessary to
correct scan data for tracer decay that occurs during the
study. Our method isderived by assuming a function of
activity that would be constant were it not for decay. This
is equivalent to computing the average decay over the
scan interval T5, i.e.,
average decay = 51's exp(—kTs)dt/Ts (2)
= 1 —exp(—kTs) 3
kTs
Inversion of the average decay yields an “average―
decay correction. Simulation studies (unpublished) of
various functions equivalent to time-varying head-ac
tivity curves likely to be encountered during our studies
demonstrate that this method of decay correction is quite
adequate (i.e., maximum error <4%).
Data analysis. Single detector system. The time
activity curve obtained subsequent to the intracarotid
injection of an aliquot of oxygen-iS-labeled water was
used to compute the mean cerebral transit time for la
beled water (tH2O) (9). The calculation of CBF from
tH2o—whichwe will call CBF(true) to distinguish it
from CBF(PET), the CBF deduced from a PET
study—isbased on the well-established central volume
principle of tracer kinetics (14,15).
The fraction of oxygen-l 5-labeled water extracted by
the brain during a single capillary transit (E) was de
termined from the same residue curve used for the
measurement of CBF(true) following intracarotid in
jection of H2150. This single-injection, external-regis
tration technique uses the first 30 secof the residue curve
after the intracarotid injection oflabeled water, a method
developed in our laboratory for use in vivo with cycio
tron-produced positron emitters (9,16,1 7)@ With this
method E is obtained by graphically extrapolating the
relatively slow clearance of the labeled water from brain
tissue back to the maximum of the perfusion peak, and
computing the ratio
E=B/A (4)
as shown in Fig. 1. As developed in detail elsewhere by
ourselves (I 7) and others (18—20)the quantity E can be
related to the tissue blood flow f[ml/(sec. g)J, the cap
illary surface area (S[cm2/g] ) and its permeability
(P(cm/secj) in the following manner:
E = 1 —exp(—PS/f)
,A “0-Water
E - B/A
10K
0
U
a 1K
0
C-)
792 THE JOURNAL OF NUCLEAR MEDICINE

CLINICAL SCIENCES
INVESTIGATIVENUCLEAR MEDICINE
PET scanners, including the one used in this study
(10) do not have adequate temporal resolution to mea
sure tissue radioactivity, C1(T), instantaneously. Thus,
to apply the autoradiographic technique to in vivo studies
with PET, a scan must run for many seconds, essentially
summing the instantaneous radioactivity over time. We
have therefore modified the operational equation for this
model by an additional integration over the time of the
scan (i.e., T2 —T1) as follows:
I' T2
C= I C,(T)dT
JT1S@T2
= XK % Ca(T)* exp(—KT)dT
JT1
Here, C is the local number of counts per unit weight of
tissue recorded by the tomograph from a region of brain
tissue during the scan. The time, T1, is chosen as the time
at which there is an appreciable increase above back
ground in a selected bank-pair coincidence counting rate
of the PETT VI scanner, thus signifying the arrival of
radioactivity of the head. In practice, the constant m is
assumed to be 1 in the solution of both the tissue [Eq.
(6)] and PET [Eq. (9)] operational equations for K.
Thus, from Eq. (7), the flow, f, is equal to AK.
This model relates regional CBF to regional tissue
counts and the arterial blood radioactivity curve. Un
fortunately, the operational Eq. (9) cannot be solved
explicitly for blood flow. It can be solved numerically,
however, by means of an interactive parameter estima
tion technique, although this approach requires signifi
cant computing time and would be unwieldy given the
large number of spatial data to be analyzed. An alter
native approach is to use the operational Eq. (9) to
generate a bookuptable that relates tissue counts to flow
for closely spaced values of flow. This involvesnumerous
evaluations of the equation and would require storage
and repetitive searching of the table. Instead, we have
expressed the operational equation relating blood flow
to tissue counts in terms of a second-order polynomial
equation:
flow = A (counts)2 + B (counts).
With this function, a parabola going through the origin,
the relationship between flow and counts can be fitted
with a less than 0.8% inaccuracy in flow for any given
number of counts. In comparison, a first-order linear
approximation would introduce errors of up to 5% be
cause the operational equation does deviate somewhat
from true linearity (see Fig. 2, Ref. 7). The addition of
a second-order polynomial term into the fit thus provides
greater accuracy.
CBF was calculated for a single, computer-generated
region of interest in the center of the baboon brain. This
region of interest was a square with 18 pixels on a side.
Because each pixel in the tomographic system is 0.27 cm
by 0.27 cm in the horizontal plane and approximately
1.4 cm thick, the volume of brain sampled was ‘--‘33cm3
or, assuming an average brain density of 1.05 g/cc, 34.7
g. Because the adult baboon brain weighs about 150 g,
the volume of tissue we sampled for our data is well
below a value that might be in error because of partial
volume effects due to sampling of noncerebral tissue.
Placing our region of interest centrally, with equal weight
given to both cerebral hemispheres, seemed appropriate
despite the fact that our CBF(true) was measured only
on the right cerebral hemisphere because no significant
asymmetries between hemispheres were noted in our
data.
(9\ Experimental procedure. CBF was varied in the ba
‘I boons in two ways. First, the arterial carbon dioxide
tension was varied by altering the respiratory rate. At
least 20 mm were allowed between changes in the res
piratory rate to permit a new steady state to be achieved.
Measurements of arterial blood gases were obtained
before and after all measurements of CBF. In addition
to varying the arterial CO2 tension we used continuous
infusions of sodium pentothal (2 g in 500 ml saline) to
achieve blood flows less than 20 ml/(min.lOO g).
In order to evaluate the effect of PET scan length on
the accuracy of the PET/autoradiographic technique
we used the list-mode data-gathering capabilities of the
system (10), which permitted us to collect data over a
period of 120 sec and then reconstruct them into scans
of 0—405cc, 0—80sec, or 0—120 sec.
Human studies. A single normal adult study* is pre
sented to illustrate the performance of the PET/auto
radiographic technique for the measurement of CBF.
The subject was prepared for the study by the percuta
neous insertion of a small radial-artery catheter under
local anesthesia to permit frequent sampling of arterial
blood, and the insertion of an intravenous catheter for
tracer injection in the opposite arm. The subject's head
was positioned in the same manner as that described,
above, for the baboons, including alignment with a ver
tical laser so that the lowest PET slice corresponded to
the subject's orbito-meatal line. A lateral skull radio
graph (Fig. 2) was obtained with this line marked by a
vertical radio-opaque wire to record permanently the
orientation of the PET slices in relation to the bony
landmarks of the skull. A molded plastic face mask
prevented significant head movement during the PET
scan. This system, described in detail elsewhere (10),
enables accurate repositioning of a subject undergoing
sequential studies on different days. After the subject's
head was in place in the tomograph, a transmission scan
for attenuation correction was performed with a ring
phantom containing germanium-68. During the actual
measurement of CBF, the room lights were dimmed and
the subject instructed to close his eyes. His ears were not
plugged. Ambient noise consisted almost entirely of
cooling fans from the electronic equipment in the
room.
Volume 24, Number 9 793

, H,@D@ ‘ ‘ Iin@of/
@ @000@.l04
j,o@o
I4C
- 12C
7c,0(
E
E@80
@ 60
C-,41
21
CO @0 40 60 80 100 120 140 160 @0 20 40 60 80 00 20 40 60
CBF(True),mlrrin'h(' CBF(True),mlmun@h('
FIG. 3. Comparison between (a) cerebral blood flow measured in adult baboons using PET/autoradiographic technique [CBF(PET)]with
intravenous,bolus injectionof H21@O,and (b)cerebral bloodflowmeastred in same animalusingintracarotklinjectionof H@O and residue
detection [CBF(true)]. Data on left represent experimental data. In addition to line of identity, figure contains theoretical line based on
computed product of brain permeabilty times surface area (P-S)for water for these experiments, 104 ml/(min-100g). This theoretical
line is product of CBF(true) and E estimated from Eq. (8) using CBF(trUe)as F in Eq. (8) along with measured P@S.For tracer with P-S
104 ml/(min-100 g), CBF(PET) progressively underestimates CBF(true) when latter exceeds 50 mI/(min-100 g). CBF(PET) data on right
have been corrected for measured extraction of H2150(see text for details).
:@:
@& 001$ ‘(:0.93+1.26
p n=34
@40 :90 r:0.96
RAICHLE, MARTIN, HERSCOVITCH,MINTUN, AND MARKHAM
curve by the amount of the difference measured in sec
onds.
RESULTS
Figure 1 presents the first 30 sec of a typical time
activity curve recorded by our single NaI(Tl) scintilla
tion detector collimated to view the injected right cere
bral hemisphere of an adult baboon. From the initial
portion of this time-activity curve the cerebral hemi
sphere extraction (E) of oxygen-i S-labeled water is
calculated according to Eq. (10) and the procedure de
picted in the figure. These same data were used to
compute CBF(true), the true cerebral blood flow, ac
cording to Eq. (6).
The relationship, based on 36 paired measurements,
between the CBF(true) and the CBF(PET) is shown in
Fig. 3 (left). Between a CBF of 10and 60 ml/m(min.iOO
g) the relationship between the two measurements is
excellent (Y = 0.90x + 0.40; r = 0.96), but above a CBF
of 60 ml/(min.iOO g) the CBF(PET) technique pro
gressively underestimates the CBF(true). Because the
intracarotid, residue-detection technique allows us to
compute not only the CBF(true) but also the extraction
(E) of labeled water [Eq. (4); Fig. 1] for each measure
ment of CBF(PET) we can evaluate the effect of this
diffusion limitation of water on our PET measurement
of CBF by dividing each CBF(PET) value by the cor
responding value for E. In effect, this corrects for the fact
that the value of m [Eq. (9)] is less than the assumed
value of 1. The result of this manipulation of the
CBF(PET) is shown in Fig. 3 (right). With this correc
tion there occurs excellent agreement between
CBF(true) and CBF(PET) over a blood-flow range of
10—155 ml/(min-100 g). Corresponding to this obser
vation is the fact that there is excellent agreement be
FIG.2. Lateralskullradiographof normal,adulthumansubject
showing position (horizontal, radio-opaque line) of lowest of seven
PETslicesdepictedinFig.5. Slicesareseparatedby 14mm (center
tocenter).Radio-opaquewireis1mmindiameter.
Because blood samples are drawn from the radial
artery in most human subjects (our study follows arte
riography, in which case the femorai artery is used) to
construct the arterial blood time-activity data for Eq.
(10), we anticipated that the intravenously injected ra
dioactivity would arrive at our peripheral sampling site
at a time that was different from its arrival time in the
brain (usually it arrives later at the peripheral site be
cause of the greater distance from the heart). Because
the accuracy of our PET/autoradiographic technique
is very sensitive to such differences (7) we correct for
them in our human studies by determining the arrival
time in the head of the subject by observing and re
cording the time of an abrupt increase in the coincidence
counting rate (sampling from a single bank pair
(10) once every second) and the arrival time at the
peripheral sampling site from the arterial-blood time
activity curve (see Fig. 5). Differences in the recorded
delays are reconciled by shifting the arterial time-activity
794 THE JOURNAL OF NUCLEAR MEDICINE

CLINICAL SCIENCES
INVESTIGATIVENUCLEAR MEDICINE
experiments, because at low CBF where E is unity, P.S
isindeterminate.
Varying the length of data collection by PET on the
CBF computed by the PET/autoradiographic tech
nique—fromthe usual 40 sec (i.e., all data in Fig. 2) to
80 or 120 sec—leadsto estimates of CBF that decrease
as a function of the acquisition time. Thus, in a typical
case of CBF decreased from 38 ml/(min.lOO g) as
measured with the usual 40-sec scan to 32 ml(min.lOO
g) with an 80-sec scan and 27 ml/(min.lOO g) with a
120-sec scan. A similar decline was observed when the
scan length was maintained at 40 sec but the starting
time following injection was progressively delayed.
Data from a single, normal, young adult male are
shown in Figs. 4 and 5. Typical intravenous injections
of 60 to 80 mCi of H215Oresult in bank pair coincidence
count rates of 30,000 to 45,000 cps with the PETT VI
tomograph in adult subjects. Figure 4 shows the blood
time-activity (corrected for decay of 0-15) resulting from
the intravenous (antecubital) bolus injection of 82 mCi
of oxygen-i 5-labeled water in 6 ml of normal saline.
Figure 5 shows data obtained from this study (scan du
ration = 40 sec).
DISCUSSION
To the best of our knowledge, these data represent the
first direct comparison of the classic Kety tissue auto
radiographic technique (1—3)with another established
technique (i.e., the central volume principle; 14,15) for
I
Sec After Injection
FIG.4. TypIcalhumanarterial-blood.time-activitycurve Obtained
from radialartery following bolusinjectionof H@O@ 6 ml of salk@e
intoantecubitalvein.NotedelayInappearanceofradiOaCtMtyin
radial-artery blood (activity injected at time 0). ThIs delay is
compared with delay Inarrival of activity In subject's head as sig
nifled by abrupt Increase In coIncidence counts from PETr VI
scannar. When differences we observed (radial-artery delay usually
exceeds head delay by several seconds) blood curve Is shifted to
correspond to head delay.
tween the data in Fig. 3 (left) and a line predicted on the
basis of the measured P.S for these experiments in the
adult baboon [PS = 104 ±11.2 (s.d.); n = 25]. This
theoretical line is simply the product of CBF(true) and
E estimated from Eq. (8), using CBF(true) as F in Eq.
(8) along with the measured P.S for these experiments.
Note that P.S could be computed on only 25 of the 36
455
42@
28*
24@ @i4
:4 I
ii@ ‘N
!5I
eN
@:‘4'
S:.
:4@@
4:@@@
:44 •@ @:@;
4:@ :.@
:@h.
. : @.@
£: 4@ .
I : 4t@ :
:@l'
@:*@4
:@ ••
C:@
_:‘t :. ::@@ :@ ::@@
::4@ 1@@
FIG. 5. TypIcalquantitative meastrement of local cerebral blood flow (@BF)In normal. aluft, male subject usk@gintravenous bolus injection
of82mCiofH21@OandadaptatiOn(7)ofclassicKetytissueautor@ilo@aphlctechnk@ue(3).Datawerecollectedover40sec.Thequantitative
grayscale [ml/(mln.100g)Jwas set to same maximum[I.e.,70 ml/(min.1009)] foreach slice to permitmore accwate visualcomparison.
Specific regions have been selected in each slice (I.e., numbered boxes) to Illustrate local variations In blood flow. Values for these regions
are listed just below each slice along with standard deviation of 25 pIxels In each region.
Volume 24, Number 9 795

RAICHLE, MARTIN. HERSCOVITCH, MINTUN, AND MARKHAM
the measurement of CBF with diffusible, inert indicators.
The choice of the latter technique as our standard for
comparison seemed especially appropriate to us because
it is not affected by the known permeability limitation
of the brain for water (9,15). Thus, failure of the
PET/autoradiographic technique to estimate accurately
the true CBF as the result ofa tracer that does not diffuse
freely between blood and brain tissue should be easily
detectable and quantifiable.
Our data (Fig. 3) clearly indicate that a brain per
meability limitation for H2150 causes an underestima
tion of CBF, using the PET/autoradiographic technique,
when the true brain blood flow exceeds about 60 ml/
(min.l00 g) in the adult baboon. Values for CBF(PET)
above this flow rate are predictably scaled by the ex
traction of H215O [i.e., E, Eq. (4), or m, Eq. (10)] for
that blood flow. Three observations support this inter
pretation of our data. First, division of the CBF(PET)
value for blood flow by the extraction of H2150 for each
experiment, measured by the intracarotid injection of
H2150 [Eq. (4); Fig. iJ, produces a linear relationship
between CBF(true) and CBF(PET) over the entire range
of blood flow examined (Fig. 3, right). This manipulation
of the experimental data is equivalent to specifying a
unique value for m [Eq. (7)], which is normally assumed
to have a value of I (2,3). Second, the average value for
the product (P.5) of brain permeability and surface area,
estimated from our data [Eq. (8)] allows us to predict
the relationship between the blood flow measured by the
PET/autoradiographic technique and the CBF(true).
We have illustrated this in Fig. 3 (left) where we have
plotted the product of f [i.e., true blood flow per unit
weight of tissue, Eq. (6)] and m [Eq. (8)] as a function
of CBF(true). It can readily be seen that with a tracer
whose apparent P.S product is 104 ml/(min.100 g), CBF
is progressively underestimated when blood flow exceeds
a value of about 40 ml/(min.i00 g). It is also apparent
that our experimental data correspond quite well to this
hypothetical relationship. Finally, studies of the PET/
autoradiographic model using C-i 1butanol as the dif
fusible, inert tracer (to be published separately) dem
onstrate a linear relationship (i.e., line of identity) be
tween CBF(PET) and CBF(true) [maximum blood flow
120 ml/(min. 100 g)] . Because C- 11 butanol freely dif
fuses across the blood-brain barrier (16), this observation
further supports our hypothesis that our experimental
results reflect the known brain permeability limitation
of H2150.
Other investigators have previously noted that the
tissue autoradiographic technique is sensitiveto the brain
permeability of the diffusible tracer chosen (5,21,22).
These observations led to several changes in the choice
of tracer for tissue autoradiographic work in small lab
oratory animals in which CBF can occasionally reach
200-400 ml/(min.lOO g).
It is important to understand the clinical implications
of this brain-permeability limitation of H215Owhen it
is utilized as the diffusible tracer for the PET/autora
diographic technique. Two facts make us optimistic that
H2' 50 will perform quite satisfactorily for studies in
humans despite its P.S limitation. First, the average brain
permeability for water in humans is somewhat higher
than we have measured in the baboon in these experi
ments [i.e., ‘@-‘i04ml/(min.100 g)]. Paulson and his
colleagues report an average P for water in the human
brain of 2.4 X iO—@cm/sec (23). Assuming an average
human brain, with capillary surface area of 100 cm2/g
(see below), this yields an average P-S (i.e., for gray plus
white matter) of 144 ml/(min.i00 g) for the human
brain. From these data we can estimate that CBF(PET)
for the brain as a whole will appreciably underestimate
CBF(true) only when the blood flow exceeds “-55ml/
(min.100 g). Second, the above analysis as well as the
presentation of our data (Fig. 3) assume a uniform P.S
for the brain. Although it is probably reasonable to as
sume a uniform vascular permeability for the brain, it
is not reasonable to assume a uniform capillary surface
area when contemplating regional studies in man with
PET. In fact, capillary surface area is estimated by
others to vary from 190 cm2/cm3 in human cerebral
cortex to 57 cm2/cm3 in human cerebral white matter
(20,24). Using these data for S one can estimate the true
water P.S of gray and white matter from the estimate of
human whole-brain P provided by Paulson et al. (i.e., 2.4
x i0—@cm/see;Ref.23).SuchanestimateofwaterP.S
for gray and white matter, and the implication it has for
the accuracy of the PET/autoradiographic technique
are shown in Fig. 6. Based on this information we are
quite confident that the PET/autoradiographic tech
nique will perform accurately when applied to regional
studies in human subjects (e.g., see Fig. 5). An important
test of this assumption will be a direct region-by-region
comparison of local CBF measured sequentially with the
FIG. 6. Hypothetical relationship between cerebral blood flow (@8F)
measured with PET/autoradlo@-aphlc technique in humans (y-axis)
andtrueCBF(x-axls)assumlngproductofbrainwhite-matterper
meability times surface area (P.S) for H2150 of 80 ml/(min•100g)
and brain gray matter P.S for H2150 of 336 ml/(mln.100 g). These
valuesforP.SofH2150are based onwhole-brainaverageforhuman
of 140ml/(mln.100g)(24 Errorsprediotedonbasisofthisanalysis
are shown as stippled deviations from line of identity.
796 THE JOURNAL OF NUCLEAR MEDICINE

CLINICAL SCIENCES
INVESTIGATIVENUCLEAR MEDICINE
PET/autoradiographic technique in the same human
subject using H215Oand C-i 1butanol. We are currently
planning this study.
We emphasize the advantages of H215O over alter
native inert tracers, labeled with longer-lived radionu
clides, that might be selected because of their greater
blood-brain barrier permeability (e.g., C-i 1 butanol,
C-i 1 iodoantipyrine). Because the data collection time
for the PET/autoradiographic method must be short
(i.e., 40 5cc; see below), approximately equivalent
millicurie doses of radiopharmaceutical must be used
regardless of the radionuclide. This results in an appre
ciably higher dose to the subject when longer-lived,
positron-emitting radionuclides are used, and more
prolonged residual background activity must be toler
ated. The consequence of these effects is to reduce
greatly the number and frequency of measurements that
can be made safely in a single subject. Using H2150, for
example, we have made as many as eight sequential
measurements of CBF in a single sitting with the mea
surements spaced about 12 to i 5 mm apart. Such a
protocol, used in our case for functional mapping of the
brain, would be difficult, if not impossible, with tracers
labeled with N-i3, C-i 1, or F-18.
There is one very important constraint that must be
placed on the use of the PET/autoradiographic tech
nique as well as on the classic tissue autoradiographic
technique itself (5). The duration of the study must not
exceed i mm, as originally prescribed by Kety and others
(2—5).Data collections in excess of this time period show
an underestimation of CBF that increases with time.
This phenomenon is also shown in the data of Ekiof et
al. (21). In the latter study the classic autoradiographic
technique was used but direct tissue sampling was sub
stituted for the actual tissue autoradiography. Exam
ining several tracers with differing permeabilities (i.e.,
antipyrine, ethanol, water, and xenon), Eklof et al. (21)
observed a decline in the measured blood flow as the
tissue data collection was delayed from 30 sec to 60 or
120sec.Wedonothaveanexplanationforthisphe
nomenon. It is quite clearly not related to any aspect of
the PET adaptation of the original Kety autoradiogra
phic technique because it occurs in applications of the
original technique (21 ) as well as in our work. It is def
initely not due to the permeability of a particular tracer.
We have observed the phenomenon with C-i 1 butanol
as well as H2150 (unpublished data). Eklof et al. (21)
have observed it with an additional group of tracers with
widely differing permeabilities. It is not due to tissue
inhomogeneity, as Eklofet al. (21) discuss in detail. Until
this problem has been solved, investigators must be
cautioned to use this technique—whether in the original
form or in the PET adaptation—with strict attention to
the length of data collection. This must not exceed 1 mm
for accurate, quantitative results.
One practical feature of the PET/autoradiographic
technique should be emphasized. Because the tissue
concentration of radionuclide is related to local blood
flow in a nearly linear manner (7), a PET image of the
distribution of H2150 or other suitable radiopharma
ceutical (e.g., C-i 1butanol) is an accurate representa
tion of local tissue blood flow. Where regional compar
isons in the same brain provide the desired clinical in
formation, sampling of arterial blood may be unneces
sary. In addition, accumulating data in humans (to be
published separately) indicate that comparisons between
sequential scans in the same person can be made when
the scans are scaled to the exact amount of H21 @Oin
jected and careful attention is paid to the positioning of
the subject in the PET scanner.
Finally, when sampling ofarterial blood from a radial
artery is necessary for a truly quantitative measurement
of CBF using the PET/autoradiographic technique, we
stress that this can be accomplished with relative safety
and minimal discomfort to the human subject. The risk
to the subject is minimized by performing an Allen test
(25) to document a dual arterial supply to the hand; and
using a 20-gauge or smaller Teflon catheter for relatively
short periods of time (26). At the time of this writing, we
have performed approximately 360 such catheterizations
without a complication. Our experience agrees with the
much greater experience of others that suggests a com
plication rate (i.e. major ischemic complication) of less
than one in i000 (27—29).
FOOTNOTE
C Permission for this study was obtained from the subject in ac
cordance with guidelinesapproved for this study by the Human Studies
Committee of Washington University and the Radioactive Drug Re
search Committee(WashingtonUniversity)of the Foodand Drug
Administration.
ACKNOWLEDGMENT
This research was supported by NIH grants NS06833, NS14834,
HL13851, RR00396, and RROI38O. Drs. Herscovitch and Martin
were Fellows of the Medical Research Council of Canada.
REFERENCES
1. KEn' SS: The theory and application ofthe exchange of inert
gas at the lungs and tissues. Pharmacol Rev 3:1—41,1951
2. LANDAU WM, FREYGANGWH JR, ROWLAND LP, et al:
The localcirculationof the livingbrain;valuesin the un
anesthetizedandanesthetizedcat. TransAmNeurolAssoc
80:125—129,1955
3. KETY55: Measurementoflocalbloodflowbytheexchange
of an inert, diffusible substance. Methods Med Res 8:228-
236,1960
4. REIVICH M, JEHLE J, SOKOLOFF L, et al: Measurement
ofregional cerebral blood flow with antipyrine-'4C in awake
cats. I App! Physiol 27:296-300, 1969
5. SAKURADA 0, KENNEDY C, JEHLE J, et al: Measurement
of local cerebral blood flow with iodo['4C]antipyrine. Am J
Physiol234:H59-H66,1978
6. RAICHLE ME: Quantitative in vivo autoradiography with
positron emission tomography. Brain Res Rev 1:47-68,
1979
Volume 24, Number 9 797

RAICHLE, MARTIN, HERSCOVITCH,MINTUN, AND MARKHAM
7. HERSCOVITCHP, MARKHAMJ, RAICHLEME: Brain
blood flow measured with intravenous H2150. I. Theory and
error analysis. J Nucl Med 24:782—789,1983
8. HERSCOVITCH P, MARKHAM J, RAICHLE ME: Accuracy
of the Kety autoradiographic measurement of cerebral blood
flow adapted for positron emission tomography. JNucl Med
23:Pl3,1982(abstr)
9. EICHLING JO, RAICHLE ME, GRUBB RL JR, et al: Evi
dence of the limitation of water as a freely diffusible tracer
in brain of the Rhesus monkey. Circ Res 35:358—364,1971
10. TER-POGOSSIAN MM, FICKE DC, HOOD JT SR, et al:
PETT VI: A positron emission tomograph utilizing cesium
fluoride scintillation detectors. J Comput Assist Tomogr 6:
125—133,1982
11. FICKE DC, BEECHERDE, HOFFMAN GR, et al: Engi
neering aspects of PETT VI. IEEE Trans Nucl Sci 29:
474—478,1982
12. YAMAMOTO M, FICKEDC, TER-P000SSIAN MM: Per
formance studyof PETT VI, a positroncomputed tomograph
with 288 cesium fluoride detectors. IEEE Trans Nuci Sci
29:529—533,1982
13. WELCH MJ, LIrroN JT, TER-POGOSSIAN MM: Prepa
ration of millicurie quantities of oxygen-IS-labeled water. J
Labelled Compd 5:168-172, 1969
14. ZIERLER KL: Equations for measuring blood flow by cx
ternal monitoring of radioisotopes. Circ Res 16:309—321,
1965
15. ROBERTSGW, LARSON KB, SPAETH PE: The interpre
tation of mean transit time measurements for multiphase
systems. J Theor Biol 39:447—475,1973
16. RAICHLE ME, EICHLING JO, STRAATMANN MG, Ct al:
Blood-brain barrier permeability 1‘C-labeledalcohols and
‘50-labeledwater. Am J Physiol 230:543—552,1976
17. RAICHLE ME, LARSON KB: The significance of the
NH3-NH4@ equilibrium on the passage of ‘3N-ammoniafrom
blood to brain. A new regional residue detection model. Circ
Res 48:913—937,1981
18. SANGRENWC, SHEPPARDCW: A mathematicalderivation
of the exchange of a labeled substance between a liquid
flowing in a vessel and an external compartment. Bull Math
Biophys 15:387—394,1953
19. RENKIN EM: Transport of potassium-42 from blood to tissue
in isolated mammalian skeletal muscles. Am J Physiol 197:
1205—1210,1959
20. CRONEC: The permeability of capillariesin variousorgans
as determined by use ofthe “indicatordiffusion―method. Ada
PhysiolScand58:292—305,1963
21. Eiu.OF B, LASSEN NA, NILss0N L, et al: Regional cerebral
blood flowin the rat measured by the tissue samplingtech
nique; a critical evaluation using four indicators C'4-antipy
rine, C'&ethanol,H3-waterand xeno&33.Acta Physiol Scand
91:1—10,1974
22. ECKMAN WW, PHAIR RD, FENSTERMACHERJD, et al:
Permeability limitation in estimation oflocal brain blood flow
with [‘4C]antipyrine.Am J Physiol 229:21 5—221,1975
23. PAULSON OB, HERTZ MM, BOLWIG TG, et al: Filtration
and diffusion of water across the blood-brain barrier in man.
MicrovascRes 13:113—124,1977
24. COBB5: Cerebrospinalbloodvessels.In Penfleld's Cytology
and Cellular Pathology ofthe Nervous System. New York,
Hoeber, 1932, Vol II, pp 577—610
25. OH TE, DAVIS NJ: Radial artery cannulation. Anaesth
Intens Care 3:12—18,1975
26. DAVIS FM: Radial artery cannulation: influence of catheter
size and material on arterial occlusion. Anaesth Intens Care
6:49—53,1978
27. DALTON B, LAyER MB: Vasospasmwith an indwelling
radial artery cannula. Anesthesiology 34:194—197,1971
28. BEDFORDRF, WOLLMAN H: Complications of percuta
noonsradial-arterycannulation:an objectiveprospectivestudy
in man. Anesthesiology 38:228—236,1973
29. DAVIS FM, STEWART JM: Radial artery cannulation: a
prospectivestudy in patients undergoing cardiothoracic sur
gery. BrfAnaesth 52:41—47,1980
The Society of Nuclear Medicine Computer &Instrumentation Councils willpresent “TheTechnology of NMR,―a symposium
for scientists and physicians, February 5—6,1984 in Orlando, Florida.
Youare invitedtosubmitpaperswhichaddressthe technologicalproblemsassociatedwiththisnewmodality.Ofspecific
interest are design of and requirements for NMR instrumentation, imaging techniques for spatial encoding, problems of
pulse sequencing, image processing and display, as well as spectroscopy. Development of methodology and phantom
design for applications to system evaluation are also of interest.
Submit a 200 word abstract and four copies to:
Dr. R.E. Johnson
Department of Radiology
ImagingDivision
The School of Medicine
University of North Carolina
Chapel Hill, North Carolina 27514
Authors willbe required to submit a camera-ready manuscript at the meeting for publication in the proceedings.
Abstracts must be received by October 1, 1983.
798 THE JOURNAL OF NUCLEAR MEDICINE
The Society of Nuclear MedicineComputer & InstrumentationCouncils
The Technologyof NMR
A Symposiumfor Scientistsand Physicians
February 5—6,1984 Call for Papers Orlando, Florida