Quantitative In Vivo Measurements of Tumor Perfusion Using Rubidium-8 1 and Positron Emission Tomography Simon R. Cherry, Paul Carnochan, John W. Babich, Franca Serafini, Nick P. Rowe!!, and Ian A. Watson Joint Department ofPhysics, Royal Marsden Hospital and Institute ofCancer Research, Sutton, United Kingdom; Perfusion Laboratory,
RoyalMarsden
Hospital,
London,
United Kingdom;
MRC Radiobiology
Department ofRadiotherapy, Royal Marsden Hospital, Sutton, UnitedKingdom; Hammersmith, London, United Kingdom
Rubidium-8i(t1,@ = 4.58 hr) was investigatedas a tumor perfusiontracerintheVX2carcinomaimplantedintorabbit thigh muscle using a large-area, multiwire proportional chamber positron emission tomography (PET) system. Perfusion was determined using the arterial reference sam
Unit, Chilton and
and MRC Cyclotron Unit,
largely extracted and retained on first-pass through a capillary bed. In the case of labeled microspheres, the extraction is due to capillary blockade, resulting in very
high first-pass extraction if the sphere size is carefully chosen and the capillary bed is uniform. The invasive nature of the technique which requires an intraarterial injection, together with the hazards associated with the
pIemethod,andthe resultsfrom PETimagingwere corn paredwith postmortemtissuesampling.Absolutequanti injection of large numbers of spheres into the arterial tation of tumor perfusionwas achievedusing external probesto estimatelocalextractionfraction.Redistribution circulation, resulted in the search for other tracers which of rubidium-8i (81Rb)was investigated using a dual-tracer show similar characteristics but are more suitable for clinical use. Potassium analogues have been widely technique. Average perfusion was found to be 13.5 and 3.7mI/mm/i00g intumorandnormalmuscle,respectively. investigated as perfusion tracers due to high first-pass The extractionfractionas estimatedfrom a two-cornpart extraction and prolonged retention following i.v. ad ment model ranged from 0.94 to 1.00. No significant ministration. The distribution of rubidium was first redistribution of 81Rb was observed in these tissues. Nine used as an indicator ofregional blood flow in the 1950s patientswith malignancieswere studied using 81Rband (4) and since then rubidium-86 (86Rb)has become a PET.Tumor perfusionin four patientswith carcinomaof standard tracer in animal tumor perfusion studies where the breast was elevated by a factor of 1.8 (range i .2—2.3) the accuracy of microspheres can be in doubt due to comparedto contralateralnormalbreast. shunting
of spheres through
tumor
vasculature
(5).
J NucIMed 1990:31:1307—1315
Another potassium analogue, thallium-20l (201Tl),has found routine clinical application in the assessment of myocardial perfusion (6,7) although its low photon energy makes it far from ideal for imaging. This, to
umor perfusion may have an important role to play in cancer treatments such as chemotherapy, radiother apy, and hyperthermia due to the importance of blood flow dependent parameters such as oxygenation, drug delivery, and heat transfer. Studies of tumor perfusion may thus lead to a better understanding of the limita tions of these treatments and will permit an objective
gether with the fact that potassium analogues do not cross the intact blood-brain barrier, prompted the de velopment of lipophilic blood flow tracers incorporat ing technetium-99m (99mTc)or iodine-123 (123!).Tech
assessment of methods aimed at the manipulation
of
tumor vasculature. Techniques for measuring tumor perfusion in ani mals have been widely reported (1,2,3) and are gener ally based on the use of radiolabeled
tracers, which are
netium-99m-hexamethyl-propylene amine oxime (HMPAO) has found widespread use as an indicator of cerebral blood flow and has been used in studies of both cerebral (8,9) and non-cerebral (10,11) tumor perfu sion.
The results from 99mTc and 123!based perfusion tracers when combined with single-photon emission computed tomography (SPECT) are at best semi-quan titative
due to the difficulty
of accounting
for attenua
tion and the limited statistics of SPECT images. The ReceivedJuly 17, 1989; revisionaccepted Feb. 21, 1990. advent of positron emission tomography (PET), in For reprints contact: Simon A. Cherry PhD, Department of Nuclear which it is inherently easier to obtain quantitative, high Medicine&Biophysics,Divisionof RadiOlOgiCal Sciences. UCLASchoolof Med@ane, Los Angeles,CA90024. resolution images of the tracer distribution (12), has
In Vivo Measurements of Tumor Perfusion • Cherry et al
1307
resulted in the development of a variety of perfusion tracers such as C'5O2 (13), H215O (14,15) and ‘3NH3 (16). There are also two positron-emitting isotopes of rubidium, both ofwhich have been employed to meas
brane, allowing the process to be described in terms of
ure myocardial perfusion (17,18,19). Rubidium-82 is
The rapid transport processes along with the large in
generally favored because of its short half-life (t112= 75 sec) which allows rapid sequential studies to be carried
tracellular rubidium
out under varying stress conditions. Drawbacks include the large positron energy (max. 3.35 MeV) that de
and long retention ofrubidium which makes it suitable as a perfusion tracer. While the rate constant describing the active transport across the cell membrane is rela tively independent of perfusion, diffusion through the capillary wall has been shown to be strongly flow de
creases image resolution
and the need to acquire data
at a high rate over a short time period which p!aces considerable
demands on the PET instrumentation.
In
a three-compartment model (Fig. 1), representing plasma, interstitial, and intracellular spaces with appro priate rate constants
between the compartments
(21).
potassium sink and slow backdiffusion of from the cell result in the high accumulation
tumor perfusion studies where the repeat studies are
pendent (22). An increase in perfusion generally leads
usually carried out at intervals of days or weeks, 81Rb
to a decreased residence time for rubidium ions in the
with a lower positron range (max. 1.11 MeV, 31% positron emission) and a half-life of4.58 hr may prove more suitable. The aims of this study were to measure tumor per fusion using 81Rb, comparing the results from nonin
capillary network, which results in a lowering of the
vasive PET imaging with post-mortem tissue sampling in animals, and to investigate possible redistribution of 81Rb during the time required for PET imaging. The
measurement of local extraction fraction was also in vestigated using an external probe system. The trans plantable VX2 carcinoma implanted into rabbit thigh muscle was chosen as the experimental system in order to produce a large volume tumor for PET imaging
extraction
fraction
across
the capillary
wall (23).
The
fall in extraction fraction as a function ofperfusion was observed to follow an exponential relationship (23,24) that was formulated mathematically by Renkin and Crone (25,26) as: E = 1 —exp@@,
(1)
ure the extraction fraction.
where E is the extraction fraction, f is the flow per unit volume (perfusion), and PS represents the product of capillary permeability and surface area. This flow de pendent extraction fraction leads to the well docu mented non-linear relationship between flow and up take of rubidium in tissue (18,27). The transport of rubidium across the cell membrane while independent
BACKGROUNDAND THEORY
of drugs or altered metabolic status. For example, the
combined with easy access for external probes to meas
of regional perfusion may be affected by the presence
Extraction and Retention of Rubidium Following i.v. injection, 81Rb passes virtually unex tracted through the pulmonary circulation and is dis tributed to tissues in proportion to cardiac output. On entering a capillary network, a large fraction of the rubidium diffuses across the microvascular wall into
the interstitial space from where it is actively trans ported into the cells by a combination of ATP-ase and Na@-K@-2Cl-cotransport (20). The major resistance to this transport occurs at the capillary wall and cell mem
presence of high plasma insulin levels has been dcm onstrated to increase rubidium extraction across the cell
membrane (28,29). Hypoxia, however, which may be encountered in rapidly growing tumors, does not appear to inhibit the uptake of thallium (30). Since similar uptake mechanisms apply to both thallium and rubid ium, rubidium is likely to be equally insensitive to the oxygenation status of tissue thus allowing accurate as sessment of perfusion in poorly oxygenated regions of tumor.
FiGURE 1 Three-compartment modelrepresent ingtransportof rubidiumfromplasma to cellularspace.The magnitudeof inter-compartmental rateconstantsare representedschematicallyby the thick ness of the arrows.
1308
FLOW
The Journal of Nuclear Medicine • Vol. 31 • No. 8 • August 1990
Calculation of PerfusionfromRubidium Uptake
LOW EXTRACTION
(E:O.1)
By applying the principle of mass conservation to a region of tissue i with perfusion f, the perfusion can be expressed as (31): fi—.
CT F_iJ
0
@(T)
(2)
Ca(t) dt, HIGH
EXTRACTION
(
[email protected])
where C@is the regional tissue concentration, Ca is the arterial concentration, and E@is the local tissue extrac tion fraction which expresses the fraction of the dcliv ered rubidium that is extracted and retained by the tissue. This will consist of a component from the cx traction across the capillary wall (Equation 1) and a second component
representing
the summed effects of
transport of rubidium across the cell membrane and
FIGURE2
redistribution.
Schematic illustration of first-pass rubidium kinetics illustrating
The regional tissue concentration
can be
measured directly from calibrated PET images, or al ternatively can be determined by post-mortem tissue counting in animal studies. The integrated arterial con
differencebetweenhigh(E = i .0)and low (E = 0.1)extraction
centration
ponent Cf can be described by (32):
is found by withdrawing
arterial blood at a
casesfollowinga typlcali.v.bolusinjection.
fixed rate during the injection of the tracer. This term has the effect of normalizing the equation for the in jected activity and cardiac output. Measuring the cx traction fraction has until recently only been possible
using extremely invasive techniques involving sampling both arterial inflow and venous outflow from a region (24). Thus, the majority ofperfusion
CI(T) = A T exp_BT,
where A and B arc constants. Assuming that there is no egress of rubidium from the trapped space, the trapped
component C becomes:
studies performed
with tracers which arc extracted and retained by tissue actually measure the perfusion index, f, x E1, which
C@(T) = K J
C@(T) dt,
(4)
0
increases as a function of f1but in a non-linear fashion,
@
(3)
leading to an underestimate at higher rates of perfusion. The recent introduction of a method to estimate cx traction fraction using small external counters (32) now enables measurements to be made relatively noninva sively in a variety of situations.
where K is the rate constant, representing exchange between the free and the trapped compartments. The total count rate C registered by the probe as a function oftime can therefore be described by (32):
Measurement of ExtractionFraction
C(T)= C@(T) +
The principle of measuring the extraction fraction involves placing small scintillation detectors over the region of interest (ROI) and monitoring the first-pass curve following a rapid i.v. injection of rubidium. Fig ure 2 qualitatively illustrates the difference between first-pass curves in situations ofhigh and low extraction for a good quality bolus injection. In the high extraction
@(T) (‘T
=
A
T
exp_BT
K
J
C@(T)
dt.
(5)
0
ing trapped rubidium and free rubidium. The free com
Fitting Equation S to the first-pass activity curve, the extraction fraction is given by the ratio CjC at the time of peak count rate (32), assuming that the bolus injec tion is short enough so that the peak count rate is a measure of the total activity delivered to that region. Another assumption of the technique is that the wash out of rubidium from the cells is negligible over the duration ofthe study. Since the measurement of extrac tion fraction is complete within a few seconds, this condition is not usually of concern. The probes will only give average extraction fraction measurements over a region which is dependent on the probe diameter and its depth response and therefore can only be used peripherally or on the surface of tissues exposed by surgery.
InVivoMeasurementsof TumorPerfusion• Cherryet al
1309
case, the curve rises rapidly initially as the bolus enters
the field of view of the detector. Most of the tracer is trapped and hence a plateau is established close to the
maximum count rate. In the low extraction case, the bolus enters the field of view as before, but now a substantial amount of the tracer flows straight through
the region without being extracted, resulting in a plateau level well below the peak level. The first-pass curves can be analyzed more quanti tatively by fitting a two-compartment
model represent
MATERIALS AND METhODS Animal Preparation
The experiments were performed on adult female New Zealand White rabbits, weighing in the range 3.5—4.5kg
after the conclusion of the experiment. The counting proce dure was repeated after an interval of five days following the decay of81Rb, allowing the counts per gram oftissue for both
81Rband @Rb to be calculated.Standards of known activity were taken from the injectate solutions to provide absolute
(median 4.0 kg). Tumors were implanted by deep intramus calibration in terms of the injected activity. Corrections for cular injection of VX2 carcinoma cell suspension (5 x 106 backgroundand radioisotopedecayweremade. cells) into the lower thigh. Tumors became palpable after 7PET imageswerereconstructedusingweightedbackprojec 10 days and rubidium studies were carried out at 21 days tion followedby deconvolutionand were correctedfor dead when the tumors were 2.5—3.5cm in diameter. The rabbits
were premedicated with Fentanyl-fluanisone (Hypnorm: CrownChemicalCo. Ltd., Lamberhurst,UK) 0.1 ml/kg i.m. prior to induction of generalanesthesiaby slow i.v. infusion of xylazine (Rompun: Bayer Bury St. Edmunds, UK) 3 mg/ kg and ketamine hydrochloride (Ketalar: Parke-Davis, Pon
typool, UK) 10 mg/kg.Followingcannulationof the right carotid artery (polyethylene tubing, i.d. 0.76 mm, o.d. 1.22
mm) for arterial blood sampling,anesthesiawas maintained by continuousinfusionof ketaminehydrochlorideat 0.8 mg/ kg/mm into a marginal ear vein. At the end of the study,
animals were killed using a lethal dose of pentobarbitone sodium.
was obtained commercially
(Amer
sham mt., Buckinghamshire, UK) and diluted to the desired concentration
81Rb,usingboth tissueactivityfrom the post-mortemsamples and three-dimensional ROl analysis (ROl volume 5.8 ml) from the calibrated PET image. Percent dose/gram was cal culated for 81Rband @Rb using the tissue samples in order to investigate redistribution. Equation 5 was used to fit the time activity curves obtained
from the CsI probes in order to
determine an average extraction fraction for tumor and nor mal thigh muscle.
RESULTS
Isotope Preparation Rubidium-86-chloride
time and the background caused by scattered and accidental coincidences (34), leading to a final image resolution of roughly 15 mm (33). Equation 2 was used to calculate perfusion as measured by
with 0.15 M NaCl. Rubidium-81-chloride
was
obtained by extraction of 81Rb from a 8IRb/8ImKrgenerator
(MRC Cyclotron Unit, London, UK) using 3 M HO. The acid solution was brought to dryness on a hot plate and redissolved in 0.15 M NaC1. Prior to injection, the 81RbCI solution was filtered and the final concentration adjusted by
dilution with sterile 0.15 M NaCI.
Extraction Fraction Typical first-pass time-activity curves from probes over the tumor and the normal thigh muscle are shown in Figure 3. The curves clearly correspond to the high extraction
case (Fig. 2). The
extraction
fraction
was
found to vary between 0.94 and 1.00 in both normal muscle and tumor tissue (Tables 1 and 2). The situation shown in Figure 3 where the curves rise to a plateau
with no obvious peak was interpreted as having an Experimental Protocol
extraction of 1.00 (29).
Approximately 0.4 mCi 81RbCl was injected as a rapid bolus into an ear vein. The volume of the injectate was 0.6
ml. Arterial blood sampling commenced 30 sec prior to injec tion, using a Harvard pump to withdraw I ml/min from the carotid artery for a total time of 5 mm. In four ofthe rabbits, two small collimated cesium iodide (CsI) probes (effective diameter 10 mm) were placed over the tumor and contralat cml normal thigh muscle and recorded time-activity
Tumorand Muscle Perfusion Average values for tumor and muscle perfusion as calculated from the PET images and post-mortem tissue sampling using Equation 2 are shown in Tables 1 and
curves
over the same period in order to estimate the extraction fraction. At 5 mm, arterial blood sampling was terminated and the rabbit was positioned in a large-area multiwire pro portional chamber PET camera (MUP-PET) (33). PET im aging commenced at 10 mm with data acquisition taking 2030 mm to collect between 0.6 and 1.0 x 106 coincidence events. On completion of the PET scan, a further injection of
0
2 MBq 86RbCl was given at 50 mm. Assuming
C.,
that the
z
C) LU Cl) U) I-
z
0
fractionation of cardiac output remains the same for both rubidium isotopes, the amount of 81Rb redistributing over a 50-mm timescale postinjection
can be assessed by comparing
the biodistribution of 81Rband 86Rb. 0 20 40 60 80 100 The rabbit was killed 2 mm after administration of @Rb, TIME (seconds) and the VX2 tumor was excised. The tumor was sectioned into 200-mg pieces which were classified as viable or necrotic FIGURE3 First-pass curves from CsI probes placed over tumor and on the basis of macroscopic appearance. Tissue samples were contralateralnormalthigh muscleduring i.v. bolus injectionof also taken from normal thigh muscle and all the major organs. 81Rb Extraction fraction is calculated by applying a two All samples were weighed and counted along with the arterial
compartment model represenling free and trapped rubidium
blood sample in an automatic gamma counter immediately
to these curves.
1310
TheJournalof NuclearMedicine• Vol. 31 • No. 8 • August 1990
I
2E@
ExtractionTABLE Fraction and Average Perfusion (±Estimated
CarcinomaPerfusionRabbit. Error)in the VX2 (ml/min/100
Extraction
sampling11.00 fraction 16.520.96
9)
PETimaging Tissue 15.5±2.1
I-
12.930.94
12.9±2.3
16.141.00
10.1±2.7
9.95—
11.7±1.9
13.76—
13.9±2.6
11.88—
11.6±1.6
‘-I
13.97— 11.5±2.4
-:3 D
notumor
D Gi
2. There is no significant difference (p < 0.05) between
the perfusion as calculated from the PET imaging and post-mortem tissue sampling, although there is a sus picion that perfusion may be overestimated
in low flow
regions and underestimated in high flow regions with PET. Such errors would be consistent with the incom plete removal of scattered coincidence events from the PET images. A transaxial PET image through the level of the center of the tumor in Rabbit 2 is shown in Figure 4. The image is scaled according to Equation 2 in order to represent absolute perfusion. Note the high perfusion of the tumor compared with the contralateral normal thigh muscle and also the rubidium uptake in the center
FiGURE 4 TransaxialPETimagethrough rabbit at the levelof the tumor, calibrated in perfusion units (mI/mm/i 00 g). Perfusion in the
tumor(right)is a factor of five higherthan normalmuscle perfusion(left).Uptakein the centeris due to activityaccu mulatingin the bladder.
shown in Table 3 demonstrate that regions containing predominantly
necrotic
tissue
show
lower
perfusion
than viable regions. The range of tumor perfusion re flccts the influence ofsampling and the degree of necro sis present. In Experiment 4, where considerable necro sis was noted, there was a factor of 18 variation in
of the image due to clearance of rubidium from the kidneys into the bladder. Due to modest sensitivity of the MUP-PET system (33) and the low positron emission decay fraction of
perfusion across the tumor.
81Rb (31 %), spatial resolution was limited to 15 mm in
Redistribution of 81Rb during the timescale of PET imaging would be detected by a difference in biodistri
Redistribution
these studies. Therefore, only average perfusion values could be obtained from the PET images. Fine tissue sampling (200 mg samples), however, revealed a very
bution between 81Rband 86Rbwhich was injected after
hetercogencous
and @Rbbiodistribution as a function of 81Rb uptake. The differences are randomly distributed around the
distribution,
with areas ofhigh and low
perfusion spread throughout the tumor. These results accord with the visual assessment ofviable and necrotic regions which was confirmed
by histology. The results
imaging. Figure 5 shows a plot ofthe difference in 81Rb
line ofequality and fitting the points by linear regression
reveals no significant pattern (slope = 0.04, intercept = 0.0006, p < 0.05) in the differential biodistribution of
2 TABLE 3
ExtractionTABLE Fraction and Average Perfusion (±Estimated
MusclePerfusionRabbit(ml/min/100 Error)in Normal
3.521.00 3.130.98 3.25—
9)ViableNecroticRange116.5(n=8)—14.6—17.6215.3 (ml/min/100
lissue
RabbitTumor
4.4±0.6 .00
3.16— 3.97— 4.38—
PET imaging
3.6±0.4 3.3±0.8
3.241
by Tissue Sampling perfusion
9)
Extraction
samplingI1.00 fraction
Heterogeneity of Tumor Perfusion Measured
2)9.9—16.1321.2(n=10)8.5(n=3)7.3—22.1424.3(n=10)5.7(n=3)3.0—54.15 (n = (n = 6)10.5
3.6 ±0.9
3.2±0.7 3.2±0.9 4.1±1.0
5.1±0.9
4.7
InVivoMeasurementsof TumorPerfusion• Cherryet al
1311
@
5@o
@o
CLINICAL STUDIES
0.03
The feasibility of using 81Rb and PET in clinical studies of tumor perfusion was investigated in a small
y= -O.0006+0.0415x 0.02 CD CD .@
semi-quantitative 0
0
.n
London, UK). One to 2 mCi (40—70MBq) of 81Rb in 1 ml were injected intravenously and PET imaging
000
E
B 0
commenced 5 mm postadministration lasting for 30 mm. Between 8 and 20 x l0@coincidence events were
B
-0.01
0
acquired
@0 -0.02
0.03
0.00
.,.y.i.j
0 02
0.04
study of nine patients with a variety
of malignancies (Table 4). Rubidium-8l was obtained as a sterile injectable solution (MRC Cyclotron Unit,
0.01
0.06
0.08
% dose/gram Rb-81
FIGURE5 Plot of difference in biodistribution of 81Rb and
in each scan. A boundary
attenuation
correc
tion method (33) was applied in studies ofthe neck. In the chest, the presence ofnonuniform attenuation from the lungs produces unacceptable errors with this 0.10 method and therefore these images remained uncor rected. In Patient 9, with metastases in the supraclavic ular nodes from carcinoma ofthe bronchus, the extrac
@Rb against
tion fraction
was measured
using the external
CsI
probes as in the animal studies. No arterial blood 81Rbuptake in tumor and normal thigh muscle.The random sampling was performed. Patient 8 was imaged twice, distribution of differences indicates there is no significant washout of rubidiumover a 50-mmperiod in these tissues. once prior to and once seven days after completing radiotherapy (17 Gy in 2 fractions, 7 days apart). Pa tients 1 and 2 also had SPECT perfusion scans using 81Rb and 86Rb over the range of uptake, which repre 99mTcHMpAO sents perfusion between 2.5 and 54 ml/min/100 g. This Analysis was carried out by placing ROIs over the indicates that