Evaluation of the 11CO2 Positron Emission Tomographic Method for ...

Journal of Cerebral Blood Flow and Metabolism

9:859-873

©

1989 Raven Press, Ltd., New York

Evaluation of the IlC02 Positron Emission Tomographic Method for Measuring Brain pH. II. Quantitative pH Mapping in Patients with Ischemic Cerebrovascular Diseases

Michio Senda, Nathaniel M. Alpert, Bruce C. Mackay, Richard B. Buxton,

John A. Correia, Steven B. Weise, Robert H. Ackerman, *David Dorer, and tFerdinando S. Buonanno

Departments of Radiology, *Psychiatry, and tNeurology, Massachusetts General Hospital, Boston, Massachusetts, U.S.A.

Summary: A practical method has been developed that, using IICOZ and positron emission tomography (PET), computes and maps (a) "effective pH" (pH,), a weighted average of intra- and extracellular pH, and (b) "clearance" (K\), product of blood flow and llCOZ ex­ traction. This method, together with measurements of ce­ rebral blood flow (CBF) and oxygen extraction fraction (OEF), was applied to 1 2 patients with cerebral ischemia or stroke. The regional K\ was positively correlated with CBF (n = +0.78). The k\/CBF ratio, representing the extraction fraction ratio of llCOZ to H/50, was nega­ tively correlated with CBF (r 0.54), suggesting that Ileoz extraction decreases as flow increases. In five acute stroke patients within 2 days of onset, the injured cortex had lower CBF (20.6 mi/min/IOO g), higher OEF

(78. 1 %), and lower pH, (6.96) than the contralateral cor­ tex (CBF 4 1 .4 mi/min/IOO g, OEF 53.3%, pH, 7.00), suggesting intracellular acidosis with intact cell membranes. In three stroke patients 5-8 days after onset, the injured cortex had higher CBF (60.9 mllmin/IOO g), lower OEF (32.0%), and higher pH, (7. 1 2) than the con­ tralateral cortex (CBF = 45.3 ml/min/IOO g, OEF 58.0%, pH, 7.06), which suggested an increase in ex­ tracellular volume compartment reflecting loss of cell membrane integrity. This method provides information on the regional tissue acid-base status and cell membrane integrity, which may be prognostic of tissue viability. Key Words: Brain pH-Positron emission tomography­ \\ COz-Stroke-Cerebral acid-base status-Hydrogen ion concentration .

Brain pH is tightly regulated by physicochemical buffering, by metabolic breakdown and production of acids, and perhaps by active transport of H + or HC03 . Such tight control is necessary at least in part because of the pH dependence of the activity of several enzymes involved in energy metabolism and

storage. In spite of this control, brain pH is altered by systemic processes and by local tissue pathology such as ischemic injury; in the latter case, such al­ terations may accompany or exacerbate tissue dam­ age. There is growing evidence linking brain acidosis and tissue injury. In animal studies, tissue acidosis has been found following ischemic injury (Anderson and Sundt, 1 983; Hakim, 1 986; Meyer et aI., 1 986; Smith et aI. , 1 986; Paschen et aI. , 1 987) and in acute infarction (Kogure et aI. , 1 980), while alkalosis has been found in the recovery period following tran­ sient ischemia (Mabe et aI. , 1 983; Yoshida et aI., 1 985) . A study on frozen brain sections revealed that the disruptions in pH extended beyond the boundaries of corresponding disruptions in NADH and ATP (Kim et al., 1 985) .

=

=

=

=

=

=

-

-

Received March 1 3 , 1 989; revised June 5, 1 989; accepted June 8, 1 989 Address correspondence and reprint requests to Dr. N. M . Alpert a t Division o f Nuclear Medicine, Massachusetts General Hospital, Fruit Street, Boston , MA 02 1 1 4, U . S . A . Abbreviations used: ANOVA, analysis o f variance; CBF, ce­ rebral blood flow; CBV, cerebral blood volume; CMR02 , cere­ bral metabolic rate of 02; CSF, cerebrospinal fluid; CT, com­ puted tomography; DMO , 5 , 5-dimethyl-2 , 4-oxazolidinedione; FWHM , full width at half-maximum; GLM , General Linear Model; OEF , oxygen extraction fraction; OM , orbitomeatal; PET, positron emission tomography; ROI , region of interest; TIA , transient ischemic attack.

859

860

M. SENDA ET AL.

In addition to being markers of metabolic change, alterations in pH itself can affect cellular processes directly . Many authors have observed that, when exposed to ischemic insult, hyperglycemic animals showed more profound neuronal damages and more accumulation of tissue lactate than normoglycemic animals, leading to the hypothesis that lactic acido­ sis may worsen ischemic brain injury (Myers and Yamaguchi, 1 977; Welsh et aI ., 1 980; Ginsberg et aI . , 1 980; Pulsinelli et aI ., 1 982) . Kraig et ai . ( 1 985) used H + -selective microelectrodes to measure brain interstitial [H+ ] ([H+ ]0) in cardiac arrest and normo- and hyperglycemic rats, and found that peak [H+ ]0 was bimodally distributed according to tissue lactate content, indicating a step function re­ lationship between [H+]0 and tissue lactate content rather than a linear titration . Harris et ai . ( 1 987) induced partial ischemia in the brain of baboons and observed that extracellular pH remained stable until regional blood flow dropped to 20 mllmin/lOO g, be­ low which [H+ ]0 increased exponentially . Those [H+ ]o responses to ischemic injury were not ex­ plained by physicochemical buffering alone, and suggested both the presence of an active ion­ transport mechanism to maintain [H+ ]0 and a [H+ ] threshold beyond which all brain cells are de­ stroyed . Petito et ai . ( 1 987) showed that direct in­ jection of sodium lactate solution (pH = 4 . 5) caused neuronal necrosis as early as 1 h postinjec­ tion, and necrosis of astrocytes, occurring later, at 3-6 h . Yet another relevant line of investigation has suggested that treating brain acidosis reduces mor­ tality in experimental brain injury (Rosner and Becker, 1 984) . There is further evidence from human studies documenting the importance of alterations in acid­ base balance . Studies such as the one performed by DeSalles et ai . ( 1 986) on subjects with severe head injury showed that subjects with a poor outcome had a higher ventricular cerebrospinal fluid (CSF) lactate level than did those with moderate disabili­ ties or a good outcome . In three head-injured pa­ tients, microelectrode measurements of extracellu­ lar brain pH showed brain tissue acidosis in areas of contusion or compression by mass lesion (De Salles et aI . , 1 987) . Alterations in acid-base balance have also been found with positron emission tomography (PET) in human subjects . Regions of relative tissue alkalosis were observed in patients with brain tumors (Brooks et aI ., 1 986), and in patients with stroke 1 0- 1 9 days after onset (Syrota et aI ., 1 983) . More recently, regions of tissue acidosis were observed in stroke patients within 48 h of onset (Hakim et al ., 1 987) . J

Cereb Blood Flow Metab, Vol. 9, No.6, 1989

Regional brain pH measurements with PET are potentially useful in the evaluation of stroke, par­ ticularly when combined with other PET measure­ ments, such as regional cerebral blood flow (CBF), oxygen extraction fraction (OEF), metabolic rate of oxygen (CMR02), and cerebral blood volume (CBV) . Furthermore, the ability to take minimally invasive measurements, repeated over short time intervals on the same subject, permits monitoring of pH changes during the evolution and treatment of stroke injury . Two tracers have been used for local pH mea­ surement by PET : [ 1 l C] 5 , 5-dimethyl-2,4-oxazo­ lidinedione (DMO) (Rottenberg et aI ., 1 983; Kear­ fott et aI ., 1 983; Syrota et aI ., 1 983) and l l C02 (Bux­ ton et aI ., 1 984, 1 987; Brooks et aI ., 1 984) . Because DMO equilibrates slowly over times longer than the 20 min half-life of II C, poor statistical precision may accompany pH measurements with this technique . The IIC02 method, first proposed by Raichle et ai . ( 1 979), was considered to suffer from problems re­ lated to the fixation of II C to organic compounds. We have developed a compartmental model to ac­ count for the IIC fixation, introduced a continuous inhalation of IIC02 for 1 0- 1 5 min instead of a single bolus inhalation to decrease the importance of fixed II C, and established a method to calculate local pH using dynamic PET measurement and least-squares fitting (Buxton et aI ., 1 984) . This technique is based upon a two-compartment model with four parame­ ters (k l , k2 ' k3' and k4) and requires nonlinear least­ squares fitting of regional time-activity curves . Ca­ nine studies have shown that measures of brain pH by PET agree with other methods both in magnitude and in response to plasma Peo2 variations (Buxton et aI ., 1 987) . The precision of the measurement was about ± 0 . 03 pH units in test-retest measurements in the same animal, and ± 0 . 04 in dog-to-dog stud­ Ies . Although the I l C02 technique was shown to be valid, the nonlinear least-squares fitting is a time­ consuming process, and is impractical for routine studies if the parameters are to be calculated pixel by pixel to generate parametric images . In this pa­ per, we describe an analytic parameter estimation technique that is accurate and fast enough for pixel by pixel mapping . The pH mapping is applied to normal human subjects and to patients with isch­ emic cerebrovascular disorders, in combination with measurements of CBF, OEF, CMR02• THEORY AND ALGORITHM

The measurement strategy advocated by Buxton et al . ( 1 984) involves continuous inhalation of l l C02

BRAIN pH MAPPING WITH IIC02 AND PET

gas, a two-compartment kinetic model, and itera­ tive nonlinear least-squares fitting. A diagram of the kinetic model is shown in Fig. 1 . Tissue is repre­ sented with two compartments : one [C;Ct)] contains dissolved l l eo2 gas and bicarbonate ions, and the other [Cm(t)] contains all other labeled compounds that evolve due to metabolic processes. Based on simulation studies, it was shown that the confound­ ing effects of l ie in chemical forms other than e02 and He03 - were much reduced with continuous inhalation of IIeo2 as compared to bolus adminis­ tration (Buxton et aI., 1 984). Measurements of pH with l l eo2 in human subjects (Brooks et aI. , 1984, 1 986) and in dogs (Buxton et aI., 1 987) have confirmed the simulation studies, showing that the contribution of these "other" labeled compounds to the measurement during the inhalation period was negligible. Analyses have also shown that dur­ ing, and for some time after, the inhalation period, the data can be adequately fit with a one­ compartment model (Buxton et aI., 1 987); as a re­ sult, it is possible to simplify the kinetic modeling equations, neglecting k4 and using a nominal con­ stant value for k3• Equations for the simplified model are shown below :

Cit) = C;(t)

+

Cm(t)

=

Kle-(k2+ k3lt ® Ca(t)

(1)

where Ct(t) represents the tissue concentration of lle at time t, which is the sum of C;Ct), i.e., l l e in the form of e02 and He03 -, and Cm(t), i.e., lie in "other" forms; Ca(t) represents the plasma concen­ tration of l ie at time t; KI represents the forward rate constant for capillary-tissue exchange, and is

861

theoretically the product of flow and IIeo2 extrac­ tion; k2 represents the reverse constant for capil­ lary-tissue exchange; k3 represents the nominal rate constant accounting for incorporation of l ie in "other" chemical forms and is fixed ( = 0. 003/min) (Buxton et aI. , 1987); and ® represents the convo­ lution operator. Because Cm(t) is small and its con­ tribution to Cit) is negligible, we can approximate Cit) with C;Ct) as shown above. Formally, Eq. ( 1 ) is identical to the well-known equation for measuring eBF with e 1502 (or H2150) using Kety's model (Kety, 1 960; Herscovitch et aI., 1 983). Several computational techniques have been described for producing regional estimates of KI and (k2 + k3) from time series data of Ct (t) and Ca(t) in Eq. ( 1 ). The one we implemented is based on the technique described by Alpert et ai. ( 1984). It uses the ratio of moments in a table look-up scheme to determine k2 for each pixel from the following equation :

JOT tCt(t)dt JOT t[e-(k2+k3)t ® Ca(t)]dt JOT Ct(t)dt JOT [e-(k2+k3)t ® Ca(t)]dt

(2)

JOT Ct(t)dtlJ: e-(k2+k3)t ® Ca(t)dt

(3)

where T is the end of data acquisition ( = 20 min). The parameter KI is also determined pixel by pixel using the equation KI =

Studies by Koeppe et ai. ( 1 985) have shown that this technique performs well when compared to the least-squares method. From kl and k2 thus obtained, L, the equilibrium partition coefficient for Il eo2, is calculated as (4)

kl --

0-

""--

-----------

l1C0 2 Hl1CO:!

k2

C.(t)

ARTERIAL BLOOD

Ci(t)

k3

- - .. � ""--

-

k4

Ct(t)

OTHER FORMS

and L is related to the effective tissue pH (pHt) by the relation L = (1

Cm(t)

TISSUE

FIG. 1. Kinetic tracer model of 11C02 in the brain. The tissue is composed of two compartments. The first is 11C02 and H11C03 - , with activity concentration expressed as C;(t). The second is all other fixed forms, with activity concentration Cm(t). K , K2• K3• and K4 are rate constants between the 1 arterial blood and the tissue compartments as shown.

+

1 0(pH,-pK'l)/(l

+

1 0(pH a - pK' l)

(5)

where pHa is the arterial plasma pH, and pK' is the apparent ionization constant ( = 6. 1 2) (Buxton et aI. , 1 984). We have studied the performance of this tech­ nique on ensembles of computer-simulated data, in which the tissue curve Cit) included statistical fluc­ tuations of SD/mean 0. 1 8 at peak count, a noise level encountered in a typical patient study. For the true value of KI = 0. 25 and L = 0. 5, the estimates (mean ± SD) were KI = 0. 3 6 ± 0. 2 1, L = 0. 48 ± 0. 04, and pHt = 6. 996 ± 0. 045. These studies =

J

Cereb Blood Flow Metab.

Vol. 9. No.6, 1989

862

M. SENDA ET AL.

showed that this analytic parameter estimation method determined L with good precision. The Kj was determined with less accuracy and precision than L as was also found with least-squares estima­ tion. It should be emphasized that, compared to the least-squares estimation, the new method does not require iterative procedures, and hence is much faster, about 1 min per 1 28 x 1 28 frame on a DEC­ V AX-ll/780 computer. MATERIALS AND METHODS

Human studies were performed further to evaluate the pH mapping method and to demonstrate the clinical fea­ sibility for studying the pathophysiology of ischemic ce­ rebrovascular disorders. Subjects The subjects in our study included three normal men (ages 42-43 years) and 12 patients with cerebral ischemic disorders: 8 with acute or subacute stroke, 1 with post­ stroke low-flow ischemia, and 3 with a history consistent with transient ischemic attack (TIA). Table 1 summarizes the clinical information and arterial blood gas data of the patients studied. They were classi­ fied into three groups according to clinical diagnosis and interval from stroke onset. Group I includes acute stroke patients within 2 days of onset (nos. 1-5). Group II in­ cludes subacute stroke patients 5-8 after onset (nos. 6--8). Group III includes patients with TIA or poststroke isch­ emia (nos. 9-12). The stroke patients were diagnosed by clinical history, neurological examination, and the presence of low den­ sity area in the x-ray computed tomography (CT) taken later that could explain the neurological findings. The TIA patients were diagnosed by recent history of neuro­ logical deficit that resolved within 24 h and the presence of angiographically demonstrated obstructed arteries that could explain their symptoms. No low-density area was found on x-ray CT in TIA patients or in the ischemic area of patient no. 9. None of the patients showed any symp-

toms, signs, or findings that suggested bilateral lesions. None had undergone bypass surgery or endarterectomy. The three normal subjects had a pH study, but no mea­ surements of CBF or oxygen metabolism were made. All of the patients underwent both pH and CBF/OEF studies successively. Informed consent was obtained from every subject. pH study Supine subjects were positioned in the gantry of a Scanditronix PC-384 PET camera (Litton et al., 1984) with their head fixed by an individually molded plastic­ foam headholder. The head was positioned so that the PET slices were parallel to the orbitomeatal line (OM line), with the lowest slice set to either 16 or 32 mm above the OM line. The subject inhaled llC02 gas in a concen­ tration of 10-20 mCi/L pumped from the cyclotron at a rate of 0.5-1.0 Llmin for 15 min, and then, the gas supply being switched off, breathed room air for 5 min. Sequen­ tial scans were performed at 30 s/frame all through the 20 min period. The arterial blood was sampled every minute (l ml per sample) through a small catheter placed in the radial artery. The blood samples were centrifuged, the plasma drawn off into weighed tubes, and the activity was measured with a well counter to generate the time course of arterial plasma-activity concentration. Additional blood samples were drawn to measure arterial pH, P02, and Pc02. Three slices with center-to-center spacing of 28 mm were used for quantitative calculations. The PET images were reconstructed by a filtered back-projection method yielding an in-plane spatial resolution of 8.5 mm (FWHM) and an axial resolution of 12 mm (FWHM). Corrections were applied to the data, as described by Bergstrom et al. (1982) to account for photon attenuation, scattered radi­ ation, and detector nonuniformity. The tissue activity concentration in the PET images was cross-calibrated against a well counter using a cylindrical phantom filled with 18F solution. The parametric images of pHt and Kl were generated using the techniques described in the previous section with an assumed blood volume correction (4%) for the entire brain. The radiation dose, estimated from dog studies, is ap-

TABLE 1. Clinical informations and arterial blood gases of studied patients No.

Age , sex M M F M M

Interval from stroke onset" 7h II h 12 h 15 h 1 . 5 days

Symptoms R R L R L

hemiplegia, global aphasia hemiplegia, global aphasia hemiparesis , facial palsy, dysarthria hemiplegia , hemianopsia, global aphasia hemiplegia, sensory los s , hemianopsia

Po2 , Pco2 , pH. (mm Hg, mm Hg, units) 67, 81, 81, 89, 80,

34, 40, 38, 37, 38,

7 . 43 7 . 42 7 . 44 7 . 45 7 . 47

1 2 3 4 5

53 38 56 89 56

6 7 8

63 M 62 M 62 F

5 days 6 days 8 days

R hemiparesis , sensory loss , global aphasia Conduction aphasia, dyslexia, dysca1culia R hemiparesis , sensory los s , agraphia

65 , 43 , 7 . 43 94, 39, 7 . 45 72 , 47, 7 . 40

9 10 I I 12

58 M 74 M 40M 43 M

3 weeks TIA TIA TIA

L R L R

71 , 84, 82, 85 ,

hemianopsia, amnestic aphasia hemiparesis facial palsy, hand numbne s s , dysarthria hemiparesis

45 , 37, 37, 39,

7 . 40 7 . 43 7 . 46 7 . 40

a Patients no . 1 -8 are stroke case s , with the above-shown interval from onset. Patient no. 9 is a 3-week-old stroke case having a small infarction in the right occipital lobe and a large persistent low-flow ischemic area in the entire right hemisphere , the latter being analyzed . Patients no. 1 0- 1 2 are TIA cases. Horitzontal lines separate subjects into three groups (see text) .

J

Cereb Blood Flow Metab, Vol. 9, No.6, 1989

BRAIN pH MAPPING WITH 11C02 AND PET proximately 0.75 rad in the lungs and 0.275 rad in the whole body in a typical case in which a total of 25 mCi of l 1C02 is administered, CBF/OEF study The regional CBF, OEF, and CMR02 were measured based on the 150 steady-state method developed by Frackowiak et al. ( 1980). The subject inhaled CI502 gas of 40 mCilL pumped at 0.5 Llmin for CBF measurement, and then 1502 gas of 80 mCilL for OEF and CMR02 mea­ surement at the same position. With each gas, we allowed 10 min to reach an equilibrium before a 6-min PET scan. The arterial blood was sampled every 2 min, starting 1 min before the start of the scan. Each blood sample was collected for 25 s to reduce the effects of respiratory fluc­ tuations, and the activity concentrations of whole blood and plasma were measured with a well counter. The im­ ages were reconstructed and calibrated in the same way as the pH study, and quantitative CBF, OEF, and CMR02 maps were generated using a technique devel­ oped in our laboratory (Senda et al., 1988), which cor­ rects for variation of arterial concentration. Activity in the blood pool was corrected assuming a uniform blood volume fraction (4%) in the entire brain. The radiation dose was estimated to be 1.5 rad in the lungs and 0. 15 rad in the whole body. Data analysis In each patient, regions of interest (ROIs) were drawn over the injured gray matter, injured white matter, con­ tralateral gray matter, and contralateral white matter as discussed below. The ROIs for the injured gray were drawn by inspec­ tion over the cerebral cortex of the diseased side that had changes in either CBF, OEF, or CMR02 compared to the contralateral side. The ROIs for the injured white matter were made over the part of the centrum semiovale that had similar abnormalities on the images. In patients no. 3, 5, 8, and 9, white matter ROIs were not drawn because the lesion did not include any part of the centrum semi-

863

ovale. No white matter ROIs were made on other white matter structures because of difficulties in discriminating white from adjacent gray or CSF space. Central core le­ sions characterized by extremely low CBF, if present, were excluded from analyses, because the decreased de­ livery of radioactivity leads to partial volume effects and poor statistical precision of the parameter estimates. Table 2 summarizes the size and approximate anatom­ ical location of the ROIs for the injured area in each case. The ROIs for the gray and white on the contralateral side were made approximately at the homologous posi­ tion to the opposite side. In normal subjects, ROIs were drawn in a similar man­ ner over the normal gray and white in the slice containing upper frontoparietal lobes and the centrum semiovale. A grid with a unit width of 5. 1 mm (2 pixels) was placed over each ROI to divide it into square boxes made of 2 x 2 4 pixels. The regional parameter values (CBF, OEF, CMR02, pHo and K ) were determined for each box as I the average of the 4 pixels, and used as the fundamental database for the following statistical analyses. To reduce the correlation between observations, the boxes were resampled by one out of four so that any two of them were one box length or more apart (center­ to-center distance of lO.2 mm or more). The mean and standard deviation of the resampled boxes were calcu­ lated for each ROl. A repeated-measures analysis of vari­ ance (ANOV A) was carried out using the General Linear Model (GLM) program in SAS (Freund et al., 1 986) (a) to estimate pH, in undamaged areas (normal subjects and contralateral areas in patients), and (b) to compare the injured areas and the corresponding contralateral areas in the patients. Additional detail on the form of the ANOVA model is presented in the Results section. To evaluate the validity of this resampling strategy, the residual was calculated as the difference of observed pH, and predicted pH, for the ANOVA model, and the corre­ lation coefficients between the residuals of neighboring boxes were calculated. A significant correlation was =

TABLE 2. Location of ROIs for injured area

Patient no .

2 3 4 5 6 7 8 9 10 11 12

Gray or white

Size (no . of boxes)

Gray White Gray White Gray Gray White Gray Gray White Gray White Gray Gray Gray White Gray White Gray White

63 15 28 22 6 17 12 59 56 41 34 19 52 71 1 23 25 1 02 14 54 53

L L L L R L L R L L L L L R L L R R L L

Approximate anatomical location of ROIs for injured area

Slice level (OM + mm)

prefrontal , premotor, motor, sensory , superior parietal (Broca's area) frontal motor, premotor (parietal) frontal (parietal) motor (excl . core) premotor, motor (parietal) frontal (parietal) middle temporal, occipital (temporooccipital) motor, inferior frontal , superior temporal frontal inferior parietal, angular parietal supramarginal, angular motor, sensory , inferior parietal, superior occipital motor, parietal, superior temporal parietal (frontal) motor, premotor, inferior frontal,inferior parietal,superior temporal frontal frontal , parietal frontal , parietal

58,8 1 81 86 86 72 81 81 30 44,72 44,72 86 86 51 72 58,86 86 58,86 86 72 72

Those areas within parentheses were excluded because of extremely low blood flow (core lesions) . 2 Size of ROI: one box equals 4 pixels , or 26 mm •

J Cereb Blood Flow Metab, Vol.

9,

No.6,

1989

M. SENDA ET AL.

864

found between contacting boxes: r = 0. 40 (p < 0.001) for horizontal contact, r = 0. 44 (p < 0.001) for vertical con­ tact, and r = 0.33 (p < 0.001) for diagonal contact. On the other hand, the correlation was small between boxes that are one box apart: r = - 0. 06 (p > 0.3) for horizontal, r = 0.03 (p > 0.5) for vertical, and r = 0.10 (p > 0.3) for diagonal. Since the correlations between the resampled boxes are small compared to the underlying spatial deter­ ministic variation and the random noise, which are not explained by the ANOVA model, the resampled data can be treated as statistically independent observations. The relationship among pH" CBF, and OEF on injured areas (mean values within the ROI) were examined using principal components analysis. The K] value and K /CBF ratio were also calculated for 1 each ROI. Because K represents the product of regional l flow and extraction fraction of llC02, and because the CBF measured with 150, in the strict sense, is the product of regional flow and extraction fraction of H2150 (Lam­ mertsma et aI. , 1981), the K]/CBF ratio represents the ratio of llC02 extraction to H2 150 extraction. RESULTS

All patients exhibited focal abnormalities either in CBF, OEF, or CMR02 in such areas as were considered by clinical criteria to be responsible for their symptoms. The injured areas were character­ ized by decreased CBF and increased OEF in group I, a state called misery perfusion by Baron et al. ( 1 9 8 1 ); increased CBF and decreased OEF (luxury perfusion) in group II; and decreased CBF and in­ creased OEF in group III. No patients showed ev­ ident focal abnormalities on the contralateral side in any of the parametric images.

50 t:l/I 0 0

......

"'-.. c::

'6 E '--"

"'-..

..-..�

� U z « a:: « (1l ...:l U

X

40

30

x

+

*

x

20

:k � +

4

10

0

+

xX

0

X

+

+

+ +

+

x

+

+

x

+

�+ x

20 CBF

80

60 40 (rnl/min/ 100g)

FIG. 2. Relationship between Kl and CBF with a regression line. Each plot represents average value within each ROI. Gray matter and white matter are plotted together. x: injured area. +: contralateral area.

for the entire population, 0. 66 ± 0. 24 for gray mat­ ter, and 0. 70 ± 0. 1 8 for white matter (mean ± SD). No significant correlation was observed between K/CBF ratio and OEF (r = 0. 24, p > 0. 1 ). Results on pHt

The pHt in each of the three normal subjects was 7. 05 ± 0. 08, 7. 07 ± 0. 05, and 7. 00 ± 0. 05 (average of 7. 04) in the gray matter, and 6. 87 ± 0. 1 3, 6.95 ±

Results on K)

The average K) value within each ROI was plot­ ted against CBF (Fig. 2). A significant positive cor­ relation was observed between K) and CBF in the entire population (r = + 0. 78, p < 0. 0002) with the regression line K) = 8. 25 + 0. 3 8 1 x CBF. No sig­ nificant difference was observed between the re­ gression lines for the injured areas (K) 8. 28 + 0. 374 x CBF) and the contralateral areas (K) 8. 1 5 + 0. 388 x CBF) (p > 0. 25), or between the gray matter (K) 1 0. 85 + 0. 342 x CBF) and the white matter (K) = 9. 85 + 0. 247 x CBF) (p > 0. 1 ). The K/CBF ratio was plotted against CBF in Fig. 3. A negative correlation was observed (r = - 0. 54, p < 0.00 1 ) with the regression line of K)/CBF 0.9 1 - 0.0069 x CBF. No significant difference was observed between the regression lines for the injured areas (K/CBF 0. 93 - 0. 0070 x CBF) and the contralateral areas (K)/CBF 0. 88 0. 0064 x CBF) (p > 0. 25), or between the gray matter (K)/CBF 0. 99 - 0. 0080 x CBF) and the white matter (K/CBF = 0. 97 - 0. 0 1 1 4 x CBF) (p > 0. 25). The average K)/CBF ratio was 0. 68 ± 0. 2 =

=

=

=

=

=

=

J

Cereb Blood Flow Metab, Vol. 9, No.6, 1989

1.2 ---� Q:I U

';;;. � '-" 0

e:::

« a::

Z 0 E-< U « a:: E-< X �

x 1.0

�

+ X

0.8 x

0.6

x

q.

x

0.4 0.2

o

+ �

+ + x

20 CBF

+ +

+

+ + x

+ x

+

40

+

+

.(!-

+

60 (rnl/min/ 100g)

x

x

80

FIG. 3. Relationship between K /CBF ratio and CBF with a 1 regression line. Each plot represents average value within each ROI. Gray matter and white matter are plotted together. x: injured area. +: contralateral area.

BRAIN pH MAPPING WITH llC02 AND PET

0.05, and 6. 9 1 ± 0. 04 (average of 6. 9 1 ) in the white matter (mean ± SD of the resampled boxes within the ROI). To examine the pHt difference between gray and white in undamaged tissues, ANOY A was per­ formed using a factorial model with "tissue type" (gray or white), "subject," and their interaction ef­ fects, on the three normal subjects and the contra­ lateral areas of the eight patients, in which ROls were obtained both in gray and white matter. The results indicated a significant interaction effect [F 5. 2, df ( 1 0, 255), p < 0. 00 1 ], suggesting that the gray-white difference depends on the subject. The least-squares average of pHt in the undamaged ar­ eas of the above-mentioned 1 1 subjects was 7.022 ± 0.005 for gray and 6. 929 ± 0. 008 for white, the dif­ ference being 0. 093 ± 0. 009 (mean ± SEM). Tables 3 and 4 present CBF, OEF, CMR02, and pHt values expressed as the mean and standard de­ viation of the resampled boxes within each ROI in each patient. The difference in pHt between injured and con­ tralateral areas was analyzed using a factorial ANOYA with "tissue state" (injured or contralat­ eral), "tissue type" (gray or white), "group" (three levels), and " patient" effect ( 1 2 levels, nested within "group"). Eliminating nonsignificant third­ or second-order interaction effects, the ANOY A =

865

model was reduced to four main effects (tissue state, tissue type, group, patient) and four interac­ tion effects (state * type, state * group, state * patient, type * group). The injured- contra­ lateral (I-C) difference in CBF, OEF, CMR02, and pHt was evaluated for each patient and tissue type, and the significance of the I-C contrast for this ANOYA model was indicated with asterisks in Ta­ bles 3 and 4. The injured cortex was acidotic in three of five patients in group I (misery perfusion), alkalotic in all three patients in group II (luxury perfusion), and comparable to contralateral in all five patients in group III (ischemia). When pHt in injured cortex was compared to con­ tralateral cortex using the I-C contrast for each group in the ANOYA model, group I was mildly acidotic (6. 96 vs. 7. 00, t 2. 4, p < 0.05), group II was alkalotic (7. 1 2 vs. 7. 06, t 5. 0, p < 0. 00 1 ), and group III showed comparable pHt between in­ jured and contralateral (7. 02 vs. 7. 00, t = 0. 91, p > 0. 1). The relationship between CBF, OEF, and pHt is expressed in Fig. 4. To facilitate plotting, the pHt value was expressed as a pHt score, which uses an integer number ranging from 0 (acidotic) to 9 (alka­ lotic) to represent the range of pHt values as shown in Table 5. Figure 4A plots the mean CBF, OEF, and pHt in

=

=

-

=

TABLE 3. Parameter values measured with PET in gray matter of each patient Patient no .

2 3 4 5 6 7 8 9 10 11 12

ROI

CBF (ml/min/lOO g)

I C I C 1 C 1 C I C

2 1 .9 5\,4 16.5 39.3 16.6 37.7 15.0 32.4 33.0 46 . 3

1 C I C I C

67. 1 43 . 3 38.9 35.2 76 . 6 57.4

I C 1 C I C I C

19.6 35 . 3 37.7 44. 8 39.5 64.9 57.3 69 . 6

±

±

±

±

±

±

±

±

± ±

±

±

±

±

±

± ± ±

±

±

±

±

±

±

8 . 2*** 10.4 2 . 9*** 5.3 6.2 3.8 2 . 2* 4.9 10.7 7.6

86. 1 55.2 77. 0 45 .9 90. 1 60. 1 80.0 56.7 57.4 48 . 4

30. 5*** 6.4 13.0 4.3 24. 8** 9.2

40. 8 63 . 2 38.7 66. 1 16.5 44. 7

4 . 0* 6.8 7 . 2* 7.4 8 . 1 *** 1 2.8 1 1 .0** 15.4

69 . 8 49 . 0 46 . 6 41.3 55.3 43 . 6 37.8 35.5

OEF (%)

CMR02 (mllmin/lOO g)

±

1 4. 7*** 5.3 1 0 . 7*** 4.9 1 \ . 3** 3.5 1 2 . 3 *** 5.5 1 1 . 1* 6.7

3 . 67 5 . 79 2 . 62 3 . 72 2 . 64 4 . 03 1 . 86 2 . 86 3 . 64 4.39

±

1 \,7*** 6.0 1 7.0*** 6.3 7. 1 *** 5.6

5 . 08 5 . 69 2.32 3.61 2.01 4 . 23

±

9 . 6*** 7.0 6 . 3 *** 6.2 7.9*** 5.4 4.2 4.2

2 . 82 3 . 54 3 . 07 3 . 26 3 . 74 4 . 85 4.32 4.87

±

±

± ±

±

± ±

±

± ± ±

±

± ±

± ±

± ±

± ± ±

± ±

±

±

±

±

±

±

±

±

±

±

± ± ±

± ±

± ±

± ±

±

±

±

pH ,

0. 72*** \.01 0 . 5 4*** 0.52 0 . 99 0.30 0.22 0.28 0 . 99 0 . 62

7.01 7 . 07 6 . 92 7.00 6.98 6 . 95 6.84 6 . 92 7 . 08 7 . 07

0 . 88** 0 . 73 0 . 8 1 ** 0.57 0 . 9 1 *** 0.63

7. 1 0 7 . 07 7.09 7 . 04 7. 1 8 7 . 07

0.58 0 . 62 0 . 49 0.68 0.84*** 1 . 08 0 . 92* 0.92

7.01 7 . 00 7 . 00 7.01 7.00 6.97 7.06 7 . 04

±

±

±

±

±

±

±

±

±

±

±

±

± ±

± ± ±

±

±

±

± ±

± ±

0. 06** 0 . 03 0.08** 0 . 03 0 . 03 0 . 03 0. 06*** 0 . 02 0 . 05 0 . 06 0 . 05* 0.04 0 . 06* 0.03 0 . 04*** 0 . 05 0 . 07 0 . 07 0.04 0 . 04 0 . 08 0 . 06 0.04 0.03

Mean ± SD o f boxes within R O I (I: injured area,C : contralateral area) . * p < 0.05 , **p 0.3). When principal components analysis was applied to the data, the first principal component (0.59 x CBF - 0.58 x OEF + 0.56 x pHt) explained 89% of the entire variability. Figure 4B illustrates the relationship among CBF,

A

OEF, and pHt in the injured white matter (n = 8). There was a weak negative correlation between CBF and OEF (r - 0.65, 0.05 < p < 0. 1 ). The correlation coefficients between pHt and other pa­ rameters were r = 0.77 (p < 0.05) to CBF, r = - 0.79 (p < 0.02) to OEF, and r = 0.45 (p > 0.25) to CMR02. The first principal component (0.56 x CBF - 0.57 x OEF + 0.60 x pHt) explained 83% of the variability. =

Results on particular cases

In patient no. 1, studied 7 h after the onset, the injured area had lower CBF, higher OEF, lower 8 100

100 4

80 __

�

'"--'

� r.:l 0

5

60

80 6

40

4

4

6

6

7

20

40

60

80 CBF

4

40

8

0

34-

3

60

20 0

o

5 4

20 o

o

20

(ml/min/100g)

FIG. 4. Plots of the average CBF, OEF, and pH, within the ROl on the injured gray matter stands for the pH, score keyed to Table

J

Cereb Blood Flow Me/ab, Vol. 9, No.6, 1989

5.

6

Each number represents one patient.

40 (A) and

60

white matter

(8).

80

The number

BRAIN pH MAPPING WITH lIeo2 AND PET TABLE S. pH! score for three-dimensional plotting pH,

Score

6 . 80 or less 6 . 80-6 .85 6 . 85-6 . 90 6 . 90-6.95 6 . 95-7. 00 7.00-7.05 7.05-7.10 7. 10-7. 15 7. 15-7 . 20 7.20 or more

0 1 2 3 4 5 6 7 8 9

CMR02, and more acidotic pHt than the corre­ sponding contralateral area both in gray and white matter (Fig. 5). In patient no. 5, 1.5 days after the onset, the in­ jured gray matter did not show any significant dif­ ference when pHt was averaged over the entire ROI (Table 3). However, when the regional pHt score within the injured gray ROI was plotted box by box against CBF and CMR02 (Fig. 6), a positive corre­ lation was observed between CBF and CMR02, and pHt increased as CBF (or CMR02) decreased. The correlation coefficients were r = + 0.83 (p < 0.00 1) between CBF and CMR02, r = - 0.70 (p < 0.0 1) between pHt and CBF, and r = - 0.62 (p < 0.05) between pHt and CMR02. Figure 7 shows parametric images of patient no. 6. High CBF, low OEF, and slightly low CMR02 were observed in the left motor cortex and inferior frontal (Broca's) areas. Low OEF extended to sur­ rounding areas, including part of the superior fron­ tal and temporal lobes. The pHt was increased in injured areas.

867

Figure 8 illustrates the relationship of CBF, OEF, and pHt score in the injured area of the same pa­ tient. In both gray and white matter, the plots in the injured area were displaced toward the lower right (higher CBF and lower OEF), with pHt being more alkalotic as the CBF and OEF deviate from the val­ ues of the contralateral side. DISCUSSION

K1 and extraction A significant positive correlation was observed between KI and CBF ( r = 0.78, p < 0.0002) (Fig. 2). Considering that KI is theoretically the product of blood flow and extraction of IIC02 (Buxton et aI. , 1 984) and that the CBF measured here is the prod­ uct of blood flow and extraction fraction of H2150, this observation supports the validity of our II C02 tracer model. The KI/CBF ratio, representing the extraction ra­ tio of IIC02 to H2150, was negatively correlated with CBF ( r = - 0.54, p < 0.00 1 ) (Fig. 3). Because the extraction of H2150 is known to decrease as the flow increases (Eichling et aI., 1 974; Raichle et aI., 1 976), this finding suggests that IIC02 extraction decreases more than H2150 extraction does as flow increases. For a passively transported tracer with limited diffusion, the one-way extraction (E) is, in general, related to flow (F) and permeability x sur­ face product (PS) by the equation E = 1 e PS/F (Renkin, 1959). Our finding suggests that the per­ meability of CO2 through the blood-brain barrier is smaller than that of H20. When the regression lines of KI (or kI/CBF) on -

-

FIG. 5. CBF, OEF, CMR02, pH, and K1 of pa­ tient no. 1 at OM + 81 mm level, acute stroke 7 h after onset. Each image is displayed at an arbitrary gray scale range. The patient's left is on the right. The injured area in the left frontoparietal lobes shows low CBF, high OEF, low CMR02 (misery perfusion), acidotic pH" and low K1 compared to the contralater­ al side.

J

Cereb Blood Flow Metab, Vol. 9, No.6, 1989

M. SENDA ET AL.

868

""'

QIJ 0 0

"-c

·s

"--

6 7

N u

�:

7

4

S ..-... 0 Il:: :::E

5

2

�5 5

5() 8646

777 5.:> 7� 6 1 'l7 f3 5

97

7

66 5

3 7

8

5

6

O ��-L����-L-L����-L� 80 o 40 60 20 CBF

(ml/min/lOOg)

FIG. 6. Plots of regional CBF, CMR02, and pH, within the injured gray matter in patient no. 5. The number stands for the pH, score keyed to Table 5. Each number represents the value of a four-pixel box within the ROI. The crossbar shows CBF and OEF of the corresponding contralateral ROI, with the average value at the center and the standard deviation shown as the half length of the bar. The average pH, score of the contralateral area was 6.

eBF were compared among subpopulations, no sig­ nificant difference was observed between injured and contralateral areas or between gray and white matter, suggesting that IIeo2 extraction is not strongly influenced by gray-white difference or by the pathophysiological changes encountered in stroke. The absence of significant correlation between k/

eBF ratio and OEF in spite of the presence of strong correlation between eBF and OEF suggests that e02 extraction is independent of regional oxy­ gen extraction or, in other words, the balance of oxygen demand and supply, further supporting the validity of the IIeo2 tracer kinetic model. Relationship between pHt, CBF, OEF, and CMR02

The effective pH (pHt) measured with IIeo2 and PET, in theory, is sensitive to intracellular pH (pH), extracellular pH (pHE)' extracellular water volume fraction (XE), and tissue water content (Vo) (Buxton et al., 1 987). The pHt will increase when either pH" XE, or Vo increases, and will decrease when any of these parameters decrease. The pattern of changes in ischemic brain injury is likely to vary significantly from patient to patient, but the following events are expected to occur in the acute ischemic injury : (I) anaerobic metabolism leading to production of lactic acid and acidosis; (2) destruction of cell membrane integrity leading to nerve cell loss, damaging both neurons and glia; and (3) brain edema. In the acute phase, cell swelling occurs due to hypoxic cytotoxicity. Later, extracel­ lular water increases with increased permeability or destruction of the blood-brain barrier. Of the three noted above, anaerobic metabolism is thought to occur first and would decrease pHt. Brain edema, producing an increase in Vo, and cell destruction leading to an increase in XE would occur later. These changes may produce conflicting effects, making it difficult to predict the change in pHt in stroke, but alterations in pHt may in some stages serve as a marker of tissue acid-base status or cell membrane integrity.

FIG. 7. CBF, O EF, CMR02, pH" and K

1 images of OM + 44 mm of patient no. 6, with subacute stroke 5 days after onset. Each image is displayed at an arbitrary gray scale range. The patient's left is on the right. The injured area shows high CBF, low OEF (luxury perfusion), and al­ kalotic pH,.

J

Cereb Blood Flow Metab, Vol. 9, No.6, 1989

BRAIN pH MAPPING WITH lleo2 AND PET

869

A

B

1F �7��

60 ,-...

�

--

� r.:l 0

If?}768

40

1

56 6 8 8

7 20

o

o

20

CBF

40

*

60

6

60

ft$§g5 �::P

6 8 7

80

(rnl/min/100g)

\ �6

8B 8

7 899

100

o

7

6 6

20

120

6

6

40

0

20

CBr

40

7

8

8

80

60

(rnl/rnin/100g)

7 7

7 7 99 B8

100

120

(A) and white matter (8) in patient no. 6. The number stands for the pH, score keyed in Table 5. Each number represents the value of a four-pixel box within the ROI. The crossbar shows CBF and OEF of the corresponding contralateral ROI, with the average value at the center and the standard deviation FIG. 8. Plots of regional CBF, OEF, and pH, within the injured gray matter

shown as the half length of the bar. The average pH, score of the contralateral area was

Figure 9 represents a simulation of the effects of these factors on pH! in a two-compartment tissue consisting of an extracellular and intracellular space. Those curves were derived from the follow­ ing equations based on the llC02 model (Buxton et ai., 1 987) : 1 + IO(pHav -pK') L = VD = 1 + 1 0(pHa-pK')

+ IO(pHt - pK') + 1 0(pHa-pK') (6)

pHav = 10g[XEIOpHE + ( 1 - XE)IOpHq

(7)

where pHav represents "average" tissue pH. As­ suming pHE = 7.40, pH! was calculated as a func­ tion of pHI for various values of VD and XE, illus­ trating how the pH of intra- and extracellular com­ partments are weighted to make pH!, the measured pH value. The simulation indicates that when the extracellular space increases to a certain level (X E > 0.6), pH! will remain elevated (>7.05) even if intracellular pH drops to a substantial acidosis (pH of 5.5). This is because pH! (strictly, pHaJ is nei­ ther the volume-weighted average of pH in each compartment nor that of hydrogen ion concentra­ tion [H+ ], but is the volume-weighted average of the reciprocal of hydrogen ion concentration (l/[H+ ]). Therefore, when the extracellular space is large, pH! is mainly determined by the volume and pH of the extracellular space, whereas when extra­ cellular space is small, pHt is sensitive to intracel­ lular pH. The confounding effect of XE on pHI es-

6

(gray) and

4.5

(white).

timation was also noted by Kobatake et ai. ( 1 984) 4 regarding their e C]DMO autoradiographic tech­ nique. Figure 9 also indicates that increase of tissue water content (VD) from the normal value (0.77) may induce a relatively smaller increase in pHt in all situations, the maximum increase being 0. 1 5 units in pH(' We obtained the normal pHt value of 7.04 (gray) and 6.9 1 (white) from three normal subjects. If XE = 0.2, pHE = 7.40, and VD 0.77 (gray) and 0.67 (white) (Syrota et ai., 1 983), these values yield an intracellular pH of 7.08 (gray) and 6.99 (white). The ANOV A results showed a similar difference in pHt (0.093 units) between undamaged gray and white matter. This value is significantly larger than 0.068 (p < 0.0 1), the expected difference induced by whether VD = 0.77 or 0.67 in normal conditions, suggesting that the observed gray-white difference can be in part, but not all, explained by the differ­ ence in water content. Our results indicated that five acute stroke pa­ tients (nos. 1-5, group I) within 2 days of onset all had areas of misery perfusion (low CBF, high OEF) surrounding the core lesions; and in the misery per­ fusion areas, three (nos. 1, 2, and 4) had signifi­ cantly acidotic pHt (0.06-0.08 units lower than con­ tralateral), while the other two cases (nos. 3 and 5) had pHt comparable to that of contralateral regions (Table 3 and Fig. 5). When patient no. 5 was exam­ ined box by box, those pixels with normal CBF and CMR02 had acidotic pHt whereas those with low =

J

Cereb Blood Flow Metab, Vol. 9, No.6, 1989

M. SENDA ET AL.

870

7.4

---�--�r

-+-

7.2 -

�

.5 ==

7.0

Co

� .... ...

u

�

6.8

X.=0.4

VD= 1.0 0.92 0.85 0.77

x.

=

VD=

rill

6

0.2 1.0 0.92 0.85 0.77

6A,-+------;,---..,--r---r---J--+_ 6.2 5A 5.8 6.6 7.0 7.4 Intracellular pH

(pHI)

FIG. 9. Simulations about the dependence of effective pH (pH,) on intracellular pH (pHI), extracellular volume ratio (XE), and

tissue water content (Vo). Assuming extracellular pH (pHE) of 7040, pH, was described as a function of pHI for Vo = 0.77, 0.85, 0.92, and 1.0 and XE = 0.2, 004, and 0.6. The horizontal broken line represents the normal pH, value in the gray matter ( = 7.04) obtained in the present study, which, together with normal values of other parameters in the literature (XE = 0.2, pHE = 7040, and Vo = 0.77), corresponds to a normal pHI of 7.08 (vertical line).

CBF and CMR02 had alkalotic pHI' averaging out the difference for the entire ROI (Fig. 6). These findings indicate that at least a part of the misery perfusion areas encountered in the acute stroke population have acidotic pHt. Our simulation study in Fig. 9 indicates that un­ der normal extracellular pH (pHE 7.40) and nor­ mal tissue water content (VE 0.77), an acidotic pHt of 0.08 units lower than normal, for example, can be observed in moderate intracellular acidosis (pHI 6.95) if extracellular volume is normal (XE 0.2). However, if a fraction of cells is destroyed so that XE goes up to 0.4, an acidotic pHt indicates severely acidotic intracellular pH (pHI 6.5); and if the tissue has developed edema, as expected in the natural course of stroke, so that VD increased to 0.85, the intracellular space should be further aci­ dotic (pHI 6.0). If 50% of the cells are destroyed (XE 0.6), acidotic pHt cannot be observed with this method. There is a possibility that the extracel­ lular space may be as acidotic as intracellular, but this situation is unlikely to occur as long as there is some blood flow. Also, studies using microelec­ trodes suggest an active mechanism to maintain =

=

=

=

=

=

=

J

Cereb Blood Flow Metab, Vol. 9, No. 6, 1989

pHE in the case of intracellular acidosis (Kraig et al., 1 985; Harris et al., 1 987). Therefore, acidotic pHt in acute stroke may suggest intact cell mem­ branes with intracellular acidosis induced by lactic acid production. All group II stroke patients (nos. 6-8) showed luxury perfusion (high CBF, low OEF) and alkalotic pHt (Tables 3 and 4, Figs. 7 and 8). From our simulations, it is evident that this increase in pHt cannot be explained by the increase in tissue water content alone. There must be an increase ei­ ther in extracellular space or intracellular pH, or in both. The simulation also indicates that when 50% of the cells are destroyed and exposed to extracel­ lular pH so that XE increases to 0.6, pHt becomes elevated no matter how far pHI decreases in the remaining intact cells. Although the physiological nature of lUXury perfusion is not completely clear, it is considered to reflect nerve cell death, release of vasoactive agents, vasomotor paralysis, inflamma­ tory cell infiltration, and neovascularization, and is associated with destruction of blood-brain barrier and increased permeability. In this phase of the stroke, the nerve cells are lost and edema occurs, increasing both XE and VD• The pHE is considered

BRAIN pH MAPPING WITH

to be close to plasma pH because sufficient ex­ change of molecules takes place between plasma and extravascular extracellular space and produced acids are washed out rapidly. Also, pHI in the lux­ ury perfusion stage may reflect that of inflammatory cells and not of nerve cells, as suggested by Syrota et al. ( 1 985), who measured extracellular water vol­ ume and intracellular pH using 76 BrNH 4 and [ l l C]DMO and reported increase in both parame­ ters. Therefore, alkalotic pHt observed in areas of luxury perfusion suggests destruction of a signifi­ cant fraction, perhaps > 50%, of nerve tissue. Patient no. 5 had alkalotic pHt in those pixels that had low CBF and eMR02 (Fig. 6). The decrease in cell metabolism may be accompanied by a decrease in intact cell volume. The resulting increase in XE may explain increased pHt in those pixels. This sug­ gests the possibility that cell loss or at least mem­ brane destruction may occur in the acute phase of stroke, in which OEF is still increased (misery per­ fusion stage), and could be observed as an increase in pHt. Our data on cerebral pH in acute and subacute ' stroke are consistent with other authors results l l with [ C]DMO method. Decreased CBF and aci­ dotic pH were observed in eight stroke cases within 48 h of onset (Hakim et al. , 1 987), and alkalotic pH with increased eBF and decreased OEF were re­ ported in subacute cases ( 1 0--3 4 days after onset) (Syrota et aI., 1 983, 1 985) . The ischemia and TIA cases (group III) did not show any apparent difference in pHt between in­ jured and contralateral cortex. We can speculate that the cell membranes in those ischemic areas are not destroyed as much as the subacute stroke pa­ tients. No further speCUlation is possible from the present data, since normal pHt may suggest normal pHI or acidotic pHI canceled by a mild increase in

XE · Our data showed a negative correlation between pHt and OEF (r 0. 78) and a positive correlation + 0. 82) in the injured between pHt and CBF (r cortex of the entire population studied. When prin­ cipal component analysis was applied to the distri­ bution, the data were distributed along a single line in the three-dimensional space of CBF, OEF, and pHt (the first component explained 89% of the total variation) (Fig. 4). Since the studied population in­ cludes a wide spectrum of ischemic disorders, it may be difficult to formulate the finding. However, the negative correlation between pHt and OEF agrees with the observations of Syrota et ai. ( 1 985) and Hakim et ai. ( 1 987) that showed negative cor­ relation between cerebral pH measured by [ l l C]DMO and OEF. Their data did not show any =

-

11

CO2 AND PET

871

significant relationship between pH and CBF while our data showed a positive correlation. This may be attributed to the difference in the studied popula­ tion, especially the period from the onset. Some sources of errors in calculating pHt should also be mentioned here. The activity within the ce­ rebral blood volume was corrected in this work as­ suming that the blood uniformly occupies 4% of the tissue volume. If the true regional CBV is larger, pHt is overestimated. The regional eBV is known to be different between gray and white matter in the normal state and varies in pathological conditions (Lammertsma et aI., 1 983) . However, this is likely to be a small effect compared to statistical error, be­ cause, according to our simulation, a deviation of ± 2% from the assumed value (i.e., eBV 2 or 6%) induces an error of 0 . 020 units in pHt estimation in a typical condition. Another problem is that a se­ vere lack of flow will result in an "apparent" pHt acidosis because little l l e02 is delivered. Also, if part of the volume is no-flow space (e. g., CSF space), pHt is underestimated due to partial volume effect. The data presented in this paper and the earlier data by Syrota et ai. ( 1 983, 1 985) and Hakim et ai. ( 1 987) form a consistent picture with a general pat­ tern of events following stroke injury. Although the number of cases is small, our preliminary results suggest a hypothesis about the predictive ability of pH measurements with respect to the tissue state in the natural course of stroke : If pHt of injured tissue is lower than the contralateral tissue, the cell is ischemic with anaerobic metabolism, the cell mem­ brane if intact, and the tissue damage may be re­ versible. On the other hand, if pHt is increased, cell destruction has already occurred and the tissue is irreversibly injured. To test this hypothesis, many more observations are needed to compare directly the changes in pHt in evolving stroke cases with the long-term assessment of the viability of irijured tis­ sue . =

=

Acknowledgment: We thank John Bradshaw, Larry Beagle, and William Buciliewicz for their invaluable tech­ nical support. We also wish to acknowledge the cooper­ ation of Drs. Kistler and Roper as well as the medical staff in performing the studies reported here. This re­ search was supported in part by NINCDS Stroke Center grant number NS 10828.

REFERENCES Alpert NM, Eriksson L, Chang J Y , Bergstrom M , Litton J E , Correia JA, Bohm C , Ackerman RH, Taveras JM ( 1 984) Strategy for the measurement of regional cerebral blood flow using short-lived tracers and emission tomography. J Cereb Blood Flow Metab 4: 28-34 Anderson RE , Sundt TM ( 1 983) Brain pH in focal cerebral isch-

J

Cereb Blood Flow Metab, Vol. 9, No. 6, 1989

M. SENDA ET AL.

872

emia and the protective effects of barbiturate anesthesia. J Cereb Blood Flow Metab 3 : 493-497 Baron JC,Bousser MG,Rey A,Guillard A,Comar 0, Castaigne P ( 1 9 8 1 ) Reversal of focal " misery-perfusion syndrome" by extra-intracranial arterial bypass in hemodynamic cerebral ischemia. Stroke 1 2 : 454-459 Bergstrom M,Litton J,Eriksson L,Bohm C,Blomquist G ( 1 982) Determination of object contour from projections for atten­ uation correction in cranial positron emission tomography . J Comput Assist Tomogr 6 : 365-372 Brooks OJ,Lammertsma AA,Beaney RP,Leenders KL,Buck­ ingham PO,Marshall J,Jones T ( 1 984) Measurement of re­ gional cerebral pH in human subjects using continuous in­ halation of l l C02 and positron emis sion tomography . J Cereb Blood Flow Metab 4: 458-465 Brooks OJ. Beaney RP, Thomas DGT, Marshall J, Jones T ( 1 986) Studies on regional cerebral pH in patients with ce­ rebral tumors using continuous inhalation of I I C02 and po­ sition emission tomography . J Cereb Blood Flow Metab 6 : 529-535 Buxton RB,Wechsler LR,Alpert NM,Ackerman RH,Elmaleh DR,Correia JA ( 1 984) Measurement of brain pH using I I C02 and positron emis sion tomography . J Cereb Blood Flow Metab 4 : 8- 1 6 Buxton R B ,Alpert N M ,Babikian V,Weise S,Correia JA,Ack­ erman RH ( 1 987) Evaluation of the "C02 positron emission tomographic method for measuring brain pH. I. pH changes measured in states of altered Pco 2 • J Cereb Blood Flow Metab 7: 709-71 9 DeSalles AAF, Kontos HA, Becker DP, Yang MS, Ward JD, Moulton R,Gruemer HD, Lutz H, Maset AL, Jenkins L, Marmarou A,Muizelaar P ( 1 986) Prognostic significance of ventricular CSF lactic acidosis in severe head injury . J Neu­ rosurg 65 : 6 1 5-624 DeSalles AAF,Kontos HA,Ward JD,Marmarou A, Becker DP ( 1 987) Brain tissue pH in severely head-injured patients : a report of three case s . Neurosurgery 20: 297-30 1 Eichling 10,Raichle ME,Grubb RL,Ter-Pogossian M M ( 1 974) Evidence of the limitations of water as a freely diffusible tracer in brain of the rhesus monkey. Circ Res 3 5 : 3 58-364 Frackowiak RSJ,Lenzi GL,Jones T,Heather JD ( 1 980) Quan­ titative measurement of regional cerebral blood flow and oxygen metabolism in man using 150 and positron emis sion tomography: theory,procedure and normal values . J Com­ put Assist Tomogr 4: 727-736 Freund RJ, Littell RC, Spector P ( 1 986) SAS system for linear models, Cary, NC, SAS Institute Ginsberg MD,Welsh FA,Budd WW ( 1 980) Deleterious effect of glucose pretreatment on recovery from diffuse cerebral isch­ emia in the cat . I. Local cerebral blood flow and glucose utilization. Stroke 1 1 : 347-354 Hakim AM ( 1 986) Cerebral acidosis in focal ischemia: II. Nimo­ dipine and verapamil normalize cerebral pH following mid­ dle cerebral artery occlusion in the rat . J Cereb Blood Flow Metab 6 : 676-683 Hakim AM, Pokrupa RP, Villanueva J, Diksic M, Evans AC, Thompson CJ,Meyer E,Yamamoto YL,Feindel WH ( 1 987) The effect of spontaneous reperfusion on metabolic function in early human cerebral infarct s . Ann NeuroI 2 1 : 279-289 Harris RI, Richards PG, Symon L, Habib AHA, Rosenstein J ( 1 987) pH,K + ,and Po2extracellular space during ischaemia of primate cerebral cortex . J Cereb Blood Flow Metab 7 : 599-604

Herscovitch P,Markham J,Raichle ME ( 1 983) Brain blood flow measured with intravenous H2 150. I. Theory and error anal­ ysis. J Nucl Med 24: 782-789 Kearfott KJ, Junck L, Rottenberg DA ( 1 983) C- I I dimethylox­ azolidinedione (DMO) : biodistribution, radiation absorbed dose, and potential for PET measurement of regional brain p H : concise communication. J Nucl Med 24: 805-8 1 1 Kety SS ( 1 960) Measurement of local blood flow by the exchange of an inert, diffusible substance . Methods Med Res 8 : 228236

J

Cereb Blood Flow Me/ab, Vol. 9, No.6, 1989

Kim SH,Handa H,Ishikawa M,Hirai 0,Yoshida S,Imadaka K ( 1 985) Brain tissue acidosis and changes of energy metabo­ lism in mild incomplete ischemia-topographical study . J Cereb Blood Flow Metab 5 : 432-438 Kobatake K, Sako K, Izawa M, Yamamoto YL, Hakim AM ( 1 984) Autoradiographic determination of brain pH following middle cerebral artery occlusion in the rat . Stroke 1 5 : 540547 Koeppe RA,Holden JE,Ip WR ( 1 985) Performance comparison of parameter estimation techniques for the quantitation of local cerebral blood flow by dynamic positron computed to­ mography . J Cereb Blood Flow Metab 5 : 222-234 Kogure K,B usto R,Schwartzman RJ,Scheinberg P ( 1 980) The dissociation of cerebral blood flow, metabolism, and func­ tion in the early stages of developing cerebral infarction. Ann Neurol 8 : 278-290 Kraig RP,Pulsinelli WA,Plum F ( 1 985) Hydrogen ion buffering during complete brain ischemia. Brain Res 3 42 : 28 1 -290 Lammertsma AA,Jones T,Frackowiak RSJ,Lenzi GL ( 1 98 1 ) A theoretical study of the steady-state model for measuring regional cerebral blood flow and oxygen utilization using oxygen- I 5 . J Comput Assist Tomogr 5 : 544-550 Lammertsma AA, Wise RJS,Heather JD,Gibbs JM, Leenders KL,Frackowiak RSJ, Rhodes CG,Jones T ( 1 983) Correc­ tion for the presence of intravascular oxygen- 1 5 in the steady-state technique for measuring regional oxygen ex­ traction ratio in the brain: 2. Results in normal subjects and brain tumour and stroke patients . J Cereb Blood Flow Metab 3 : 425-43 1 Litton J,Bergstrom M,Eriksson L,Bohm C,Blomquist G ( 1 984) Performance study of the PC-384 PET camera for the brain. J Comput Assist Tomogr 8 : 74-87

Mabe H,Blomquist P,Siesjo BK ( 1 983) Intracellular pH in the brain following transient ischemia. J Cereb Blood Flow Metab 3 : 1 09-1 1 4 Meyer FB, Anderson RE, Sundt TM, Yaksh T L ( 1 986) Intra­ cellular brain pH, indicator tissue perfusion,electroenceph­ alography,and histology in severe and moderate focal cor­ tical ischemia in the rabbit . J Cereb Blood Flow Metab 6 : 7 1 78 Myers RE, Yamaguchi S ( 1 977) Nervous system effects of car­ diac arrest in monkeys. A rch Neural 34: 65-74 Paschen W, Djuricic B, Mies G, Schmidt-Kastner R, Linn F ( 1 987) Lactate and pH in the brain: association and dissoci­ ation in different pathophysiological states . J Neurochem 48 : 1 54- 1 59 Petito CK, Kraig RP, Pulsinelli WA ( 1 987) Light and electron microscopic evaluation of hydrogen ion-induced brain ne­ crosi s . J Cereb Blood Flow Metab 7: 625-632 Pulsinelli WA, Waldman S,Rawlinson 0, Plum F ( 1 982) Mod­ erate hyperglycemia augments ischemic brain damage : a neuropathologic study in the rat . Neurology 3 2 : 1 239- 1 246 Raichle ME, Eichling JO, Straatmann MG, Welch MJ, Larson KB, TerPogossian MM ( 1 976) Blood-brain barrier perme­ ability of " C-Iabeled alcohols and ISO-labeled water. A m J PhysioI 230: 543-552

Raichle ME, Grubb RLJ, Higgins C S ( 1 979) Measurement of brain tissue carbon dioxide content in vivo by emission to­ mography . Brain Res 1 66 : 4 1 3-4 1 7 Renkin E M ( 1 959) Transport of potassium-42 from blood to tis­ sue in isolated mammalian skeletal muscle s . A m J Physiol 1 97: 1 205- 1 2 1 0 Rosner MJ, Becker DP ( 1 984) Experimental brain injury : suc­ cessful therapy with the weak base,tromethamine. With an overview of CNS acidosis . J Neurosurg 60 : 9 6 1 -97 1 Rottenberg DA, Ginos JZ, Kearfott KJ, Junck LR, Dhawan V ( 1 983) Determination of regional cerebral acid-base status using " C-dimethyloxazolidinedione and dynamic positron emission tomography. J Cereb Blood Flow Metab 3 (suppl I ) : S 1 50-S 1 5 1 Senda M , Buxton RB, Alpert NM, Correia JA, Mackay BC, Weise SB, Ackerman RH ( 1 988) The 150 steady-state

BRAIN pH MAPPING WITH lIeD2 AND PET method : correction for variation in arterial concentration. J Cereb Blood Flow Metab 8 : 6 8 1 --690 Smith ML, von Hanwehr R, Siesj6 BK ( 1 986) Changes in extra­

and intracellular pH in the brain during and following isch­ emia in hyperglycemic and moderately hypoglycemic rat s . J Cereb Blood Flow Metab 6: 574-583

Syrota A, Castaing M, Rougemont D, Berridge M, Baron JC, Bousser MG, Pocidalo 11 ( 1 983) Tissue acid-base balance and oxygen metabolism in human cerebral infarction studied with positron emission tomography . Ann NeuroI 1 4 : 4 1 9-428 Syrota A , Samson Y, Boullais C, Wajnberg P , Loc ' h C, Crouzel C, Maziere B, Sous saline F, Baron JC ( 1 985) Tomographic

873

mapping of brain intracellular pH and extracellular water space in stroke patients . J Cereb Blood Flow Metab 5 : 3 5 8368

Welsh FA, Ginsberg MD, Rieder W, Budd WW ( 1 980) Delete­ rious effect of glucose pretreatment on recovery from diffuse cerebral ischemia in the cat . II. Regional metabolic levels . Stroke 1 1 : 355-363

Yoshida S, Busto R , Martinez E, Scheinberg P, Ginsberg MD ( 1 985) Regional brain energy metabolism after complete ver­ sus incomplete ischemia in the rat in the absence of severe lactic acidosis . J Cereb Blood Flow Metab 5 : 490--5 0 1

J

Cereb Blood Flow Metab, Vol. 9, No.6, 1989