Adequacy of a Compartment Model for CMRO 2 Quantitation Using 15 O-Labeled Oxygen and PET: A Clearance Measurement of 15 O-Radioactivity Following Intracarotid Bolus Injection of 15 O-Labeled Oxyhemoglobin on Macaca Fascicularis

INTRODUCTION

Use of ¹⁵O-labeled molecular oxygen (¹⁵O₂) is an ideal molecule for tracing the kinetics of oxygen in vivo, and allows quantitative assessment of regional cerebral metabolic rate of oxygen (CMRO₂) in vivo. Several mathematical approaches have been developed to estimate functional images of CMRO₂ from positron emission tomography (PET) images acquired following continuous or short-period inhalation of ¹⁵O₂, together with additional assessments of regional cerebral blood flow (CBF) and regional cerebral blood volume. The calculation process is based on a model, in which the fraction of labeled oxygen that diffuses into brain tissue is rapidly reduced by cytochrome oxidase of the mitochondrial electron transfer system, immediately converted to ¹⁵O-labeled water (H₂₅O), and washed out according to CBF. It has therefore been assumed that the back diffusion of ¹⁵O₂ is negligibly small (Figure 1). It is also assumed that there is negligible amount of retention of ¹⁵O₂ in the cerebral tissue (Figure 1). These enable that the clearance after the ¹⁵O₂ transport from tissue to capillary bed can be compensated using CBF information obtained from an independent PET scan with H₂₅O. The practical procedures to estimate CMRO₂ from a series of PET scanning such as the steady-state technique^1–3, the three-step autoradiography ⁴ , and single-step approaches^5–7 are straightforward and have been successfully applied to a number of clinical studies.

OPEN IN VIEWER

Figure 1.

Tracer model commonly employed for describing radioactivity kinetics after ¹⁵O₂–hemoglobin (¹⁵O₂–Hb) injection. The model consists of a vascular compartment, a non-metabolized oxygen compartment, and a metabolized water compartment. Negligible amount of back diffusion in the form of ¹⁵O-labeled oxygen (¹⁵O₂) is attributed to the short life time of molecular oxygen in the cerebral tissue. No accumulation or retaining system exist for ¹⁵O₂ oxygen, neither for ¹⁵O-labeled water (H₂₅O). BBB, blood–brain barrier.

The clearance of ¹⁵O-radioactivity after the intracarotid bolus injection of ¹⁵O-labeled oxygen (¹⁵O₂) in the form of oxyhemoglobin has been first measured using a single gamma photon detector in humans. Slopes were claimed to be consistent between ¹⁵O₂–oxyhemoglobin (¹⁵O₂–Hb) and H₂₅O,^8,9 but a significant level of background due to the limited collimation of the detector was a limitation. Thus, another experiment was carried out recently by Seki et al¹⁰ using a beta ray-sensitive plastic scintillator directly attached to the surface of rat brain. They demonstrated that the clearance after the internal carotid bolus injection of ¹⁵O₂–Hb was considerably faster than that of H₂₅O, suggesting the presence of back diffusion of ¹⁵O₂ from brain tissue in addition to the back flux of H₂₅O. However, there could be several technical limitations, including the insufficient collimation of their detector system, and the relatively large volume of administrated ¹⁵O₂–Hb containing blood samples. It is, therefore, still an important issue to confirm whether or not the clearance slope of ¹⁵O-radioactivity after the internal carotid artery (ICA) bolus injection of ¹⁵O₂–Hb agrees with that of H₂₅O–saline.

In the present study, we intended to reevaluate the adequacy of the compartmental model commonly employed for ¹⁵O₂-inhalation PET studies using a bigger animal. Clearance of the radioactivity concentration in local brain regions were quantitatively assessed on the nonhuman primate of Macaca fascicularis after the ICA injection of ¹⁵O₂–Hb and H₂₅O, from sequentially acquired tomographic images using an advanced PET scanner. We then evaluated whether or not we are able to identify the significant difference in the clearance slopes between the ¹⁵O₂–Hb and H₂₅O samples.

MATERIALS AND METHODS

Animal Preparation

Anesthesia was induced with ketamine (10 mg/kg, intramuscularly) and maintained by intravenous propofol (4mg/kg per hour) and vecuronium (0.05 mg/kg per hour) during the experiment. Animals were intubated and their respiration was controlled by an anesthetic ventilator (Cato; Drager, Lubeck, Germany) providing a gas mixture of 24% O₂ and 76% N₂^6,7,11. A catheter was inserted into the ICA for intracarotid bolus injection of ¹⁵O₂–Hb, H₂₅O, and C¹⁵O–hemoglobin (C¹⁵O–Hb). A catheter was also inserted into the femoral artery for monitoring the blood pressure, heart rate, O₂ content, and the partial pressures of carbon dioxide level in arterial blood. Additional catheter was placed to the anterior tibial vein, and was utilized for infusing the anesthetic agents.

Preparation of ¹⁵O₂–Hemoglobin, C¹⁵O–Hemoglobin, and H₂₅O–Saline

¹⁵O-radioactivity was produced by the ¹⁴N(d,n)¹⁵O nuclear reaction with a 0.5% O₂/N₂ gas target using a CYPRYS-HM18 cyclotron (Sumitomo Heavy Industry, Tokyo, Japan). The bombarded target gas including the produced ¹⁵O₂ was transported to an in-house synthesizer system to achieve high radiochemical purity for each ¹⁵O₂ and C¹⁵O gases. The ¹⁵O₂–Hb and C¹⁵O– Hb were prepared using the labeling system developed by Magata et al¹². Briefly, a part of an infusion line kit (Terumo Corporation, Tokyo, Japan) and an artificial lung 18 cm in length (Senko Medical Instrument Manufacturing, Tokyo, Japan) were connected using silicone tubing to make a closed blood circulating system (see Figure 2). Venous blood collected from each monkey was infused into the system and was circulated by a peristaltic pump at a flow rate of 100 mL/minute, with the introduction of ¹⁵O₂ and/or C¹⁵O gases at ˜7,000 MBq/minute/500 mL into the artificial lung for 5 to 15minutes, producing the ¹⁵O₂–Hb and C¹⁵O–Hb samples of ˜60 MBq/mL. A coincidence detector block containing a pair of GSO scintillator ¹³ was implemented to monitor the radioactivity in the circulating blood. The ¹⁵O₂–Hb and C¹⁵O–Hb samples were adjusted to contain radioactivity of ˜17 MBq in 0.4mL at the time of injection into the ICA.

OPEN IN VIEWER

Figure 2.

Schematic diagram of a system to generate the blood samples containing ¹⁵O₂–hemoglobin (¹⁵O₂–Hb) and C¹⁵O– hemoglobin (C¹⁵O–Hb). The system is the same as the one described in Magata et al. ¹² An additional radioactivity detector was placed to monitor the radioactivity in the circulating blood sample.

The ¹⁵O₂ gas was also transferred to H₂₅O using a radioactive water generator ¹⁴ to produce the injectable saline containing H₂₅O. The H₂₅O–saline was diluted by the arterial blood of more than eight times volume, and had radioactivity of ˜20 MBq in 0.4mL at the time of bolus injection into the ICA.

Positron Emission Tomography Experiments

The PET scanner was ECAT HR, Siemens-CTI (Knoxville, TN, USA) containing BGO scintillator blocks, which can acquire the data either in two-dimensional or in three-dimensional modes. The scanner provides tomographic images for 47 slices for an imaging field of view of 55 cm in diameter and 15 cm in axial length. The intrinsic spatial resolution was 5.8 mm in full width at half maximum at the center of the field of view. Observed spatial resolution after applying the three-dimensional Gaussian post filter, the observed spatial resolution was ˜10 mm.

At least 1hour after the end of the preparation for catheter, we confirmed that the heart rate, the arterial blood pressure, and the arterial partial pressures of carbon dioxide and oxygen can be maintained constant, a series of PET scans were initiated. A 900-second transmission scan was first carried out using a rotating ⁶⁸Ge-⁶⁸Ga rod source for the attenuation correction. Each of ¹⁵O₂–Hb (˜17 MBq), H₂₅O–saline (20 MBq), and C¹⁵O–Hb (17 MBq) samples of 0.4 mL were injected in 3 to 4 seconds into either the right or the left ICA. Three sets of list mode acquisition were carried out in two-dimensional mode for 5.5 minutes, for each injection. The order was the same in all three experiments, because this is the easiest protocol in our radio synthesizing system.

Data Analysis

The list mode data were sorted to produce dynamic sinogram for each scan, with frame durations of 30 times 1 second, 15 times 2seconds, 12 times 5 seconds, and 18 times 10seconds, in total 5minutes. Images were then reconstructed using a filtered back projection technique, including corrections for dead time, the radioactive decay, detectors normalization, attenuation, and scatter, using the vendor software programs. Dynamic PET images following the intracarotid injection of H₂₅O were added over the initial 50seconds after the peak of the head curve. Five regions of interest were defined on this image for each experiment (see Figure 3), in the middle cerebral artery territory of the injected hemisphere on different slice levels, providing five sets of clearance curves for each injection. Angiography was carried out in order to confirm the location for the coronary arterial injection. It was also monitored that 0.4mL sample does not interfere the natural cerebral circulation. Each clearance curve after the ¹⁵O₂–Hb and H₂₅O–saline injection was fitted to a single exponential function during the first 5 to 40seconds (early), and 80 to 240 seconds (late) periods after the peak. Mean and s.d. were calculated for slope values for each of the early and late phases, and also for both ¹⁵O₂–Hb and H₂₅O–saline for each experiment. Significant difference in the slope parameters was identified between ¹⁵O₂–Hb and H₂₅O–saline, using a Student's t-test (paired one-tail, one-tail). Residues of the single exponential fit from the measured data were calculated, and the presence of the significant correlation was tested on the semi-logarithmic domain using Pearson's test. P < 0.05 was considered statistically significant throughout the study.

OPEN IN VIEWER

Figure 3.

(A) Regions of interest (ROIs) placed on each at the monkey experiments. Five ROIs were selected on the accumulated positron emission tomography (PET) images over the initial 50 seconds after the peak, on the middle cerebral artery territory of the injected hemisphere at different slice levels. In the PET images, radioactivity from the contralateral to the injected hemisphere was attributed to the collateral circulation though the Willis Ring. (B) An example image of angiography carried out prior to the carotid artery injection of ¹⁵O-labeled samples.

RESULTS

Positron emission tomography scans were successfully carried out on all animals for all injections, with the variation of partial pressures of carbon dioxide < 1mmHg, the systolic/diastolic blood pressure < 4mmHg, the heart rate < 5% throughout the study at each experiment. Typical coincidence counting rate of the whole PET gantry was ˜35, 60, and 30kcps at peak, corresponding to the ¹⁵O₂–Hb, H₂₅O–saline, and C¹⁵O–Hb injections, respectively. The dead time counting loss was < 5% at peak in all scans.

Figure 4 shows a typical example of the clearance curves obtained following the ¹⁵O₂–Hb, H₂₅O–saline, and C¹⁵O–Hb sample injections into one of the experiments in middle cerebral artery territory of the central slice. The clearance is rapid after the C¹⁵O–Hb sample injection, while they consisted of slower slopes with ¹⁵O₂–Hb and H₂₅O–saline injections. It can be seen that the clearance slopes after 5 seconds of the peak consisted of the early (faster) and delayed (slower) components in all experiments with the ¹⁵O₂–Hb and the H₂₅O–saline injections. A sharp peak was visible after the ¹⁵O₂–Hb blood sample injection at the beginning, but this was not seen after H₂₅O sample injection. It can also be seen that the clearance slopes were visually identical between the ¹⁵O₂–Hb and the H₂₅O–saline injections in both early (5 to 40 seconds) and late (70 to 240 seconds) phases. Results from the fitting of the clearance slopes for both the early (faster) and late (slower) slopes to the single exponential function are summarized in Table 1. Interslice variations of the slope values were 10% to 20% in each experiment, for both the early (fast) and the late (slower) slopes in the ¹⁵O₂–Hb and the H₂₅O–saline injections. Within these variations, no significant difference was detected between ¹⁵O₂–Hb and H₂₅O–saline injections in each of the three experiments for both the early (fast) and the late (slower) slopes. The residues of the fitting showed no significant correlation in relation to the time on the logarithmic domain, in the early and late phase slopes, both for the ¹⁵O₂–Hb and H₂₅O–saline injections in each of the three experiments.

OPEN IN VIEWER

Figure 4.

Typical results from the clearance measurement, with results of fitting by single exponential function for the early (5 to 40 seconds) and late (80 to 240 seconds) periods, indicated with blue and red lines, respectively. Clearance curves are similar between the two curves obtained from the internal carotid artery injection of ¹⁵O₂-hemoglobin (¹⁵O₂–Hb) and ¹⁵O-labeled water (H₂₅O). The clearance was faster for C¹⁵O–hemoglobin (C¹⁵O–Hb) injection. PET, positron emission tomography.

OPEN IN VIEWER

Table 1.

List of hemodynamic parameters and summary results of clearance slopes fitted the single exponential function to the first 5 to 40 seconds (early) and 80 to 240 seconds (late) phases after the peak of the clearance curves

	Weight (kg)	Age (year)	Hb (g/dL)	Systolic/diastolic blood pressure (mm Hg)	Heart rate (bpm)		PaCO2 (mmHg)	Slope (minute) (early)	Slope (minute) (late)
Monkey 1	7.5	3	13.5	110/60	136	15O2–Hb H25O	42.1 42.5	0.372 ± 0.078 0.374 ± 0.059	0.212 ± 0.070 0.208 ± 0.052
Monkey 2a	7.1	4	12.6	115/65	133	15O2–Hb H25O	42.6 43.4	0.468 ± 0.076 0.513 ± 0.045	0.265 ± 0.070 0.292 ± 0.014
Monkey 2b	7.1	4	12.9	132/90	131	15O2–Hb H25O	39.2 40.6	0.452 ± 0.038 0.475 ± 0.083	0.275 ± 0.017 0.296 ± 0.014
Average	7.23	3.7	13.0	119/71.6	133.3	15O2–Hb H25O	41.3 42.2	0.430 ± 0.076 0.454 ± 0.085	0.251 ±0.061 0.265 ± 0.051

HB, hemoglobin; H₂₅O, ¹⁵O-labeled water; ¹⁵O₂–Hb, ¹⁵O₂–hemoglobin; PaCO₂, partial pressures of carbon dioxide, ROI, region of interest. Clearance slope values are presented for mean and 1 s.d. for each experiment among five ROIs, and also for pooled data for five ROIs of total three experiments.

Figure 5 shows the Bland–Altman's plots demonstrating the difference in the clearance slope values being not significantly different between the ¹⁵O₂–Hb and H₂₅O–saline injection both in the early (faster) and late (slower) phases, at each of the three experiments. One s.d. values of the differences were 0.078 minutes and 0.041 minutes, and the averaged values between the ¹⁵O₂–Hb and H₂₅O–saline injections 0.442 minutes and 0.258 minutes corresponding to the early (faster) and late (slower) slopes, respectively, resulting in the ratio of these values being 17.6% and 15.9% for the entire analysis.

OPEN IN VIEWER

Figure 5.

Results from the Bland–Altman plot analysis obtained from the single exponential fitting analysis for each of the three experiments, and for all pooled data (All). The agreement of the two assessments showed uncertainty of ˜20% (1 s.d. of 0.078 and 0.041 at averaged slope value of ˜0.4 and 0.2 corresponding to the faster and slower components, respectively) for each of the two components. ¹⁵O₂–Hb, ¹⁵O₂– hemoglobin; H₂₅O, ¹⁵O-labeled water.

DISCUSSION

The clearance slopes of ¹⁵O-radioactivity following the slow bolus injection of ¹⁵O₂–Hb, H₂₅O–saline and C¹⁵O–Hb into the ICA were determined for local cerebral regions using sequential PET imaging in three independent experiments on two monkeys (M. fascicularis). The clearance curves consisted of faster (early) and slower (late) components after 4 seconds of the peak, corresponding to the gray matter and white matter tissues, respectively. The clearance curves were well reproduced by the single exponential functions for both the early (5 to 40 seconds) and late (80 to 240 seconds) periods, in which residuals showed no time dependency on the logarithmic domain in all fits. It should also be noted that the variation (1 s.d.) in the differences of the slope values showed no significant difference between the ¹⁵O₂–Hb and H₂₅O injections was for both the gray matter and the white matter components in each experiment. Variation of the differences was 17.6% and 15.9% relative to the mean slope values in the entire analysis, corresponding to the faster and slower components, respectively. These variations were associated with the uncertainty of the PET assessment, and also attributed to the variation of true CBF values at different slice levels. The agreement of the slope values found in this study between the ¹⁵O₂–Hb and H₂₅O injections supports the traditional hypothesis shown in Figure 1 on the ¹⁵O₂ kinetics in the cerebral tissue, namely the non-negligible amount of clearance from the brain in the form of ¹⁵O₂ to the capillary bed, and as well as the non-negligible amount of effective retention of ¹⁵O₂ molecule in the brain tissue.

Accuracy of the assessment is well supported in two-dimensional PET, by involving corrections for detector inhomogeneity, random coincidence, dead time, attenuation and scatter, and the well-established reconstruction procedures. Contamination of signals and hence, random coincidence events can be accurately corrected, and were only small in this study. The dead time count loss was only less than 5% at the peak of total counting rate of ˜60kcps over the whole field of view. Use of an artificial lung ¹² should have contributed to minimize the risk for the possible damage of the red cells and hemoglobin during the labeling procedures. The injected sample volumes of 0.4mL into an over a 3 to 4seconds period (namely 0.1 mL/second or 6mL/minute) was smaller than the natural cerebral perfusion of ˜20 mL/minute, as estimated by assuming the averaged regional perfusion of 0.3 mL/minute per gram, and the whole brain weight of 60g. Injection of the samples applied in this study thus unlikely causes disturbance in the cerebral circulation. It should also be noted that the contribution of recirculation of injected radioactivity can be considered only small, because only a small fraction of injected radioactivity can return the cerebral tissues. As the decrease of the brain radioactivity was only 2 to 3 times even at the end of the PET imaging, amount of the recirculating radioactivity should cause negligible effects. Despite the small number of the experiments, the results from this study suggest that the presence of significant difference in the clearance slopes between ¹⁵O₂–Hb and H₂₅O injections would be unlikely. Most of the previous approaches^{5–7,15–17} that quantitatively assessed CMRO₂ and CBF in various clinical settings, which stand for these assumptions, are thus considered to be adequate.

It has been believed that oxygen is excessively delivered into intact brain as compared with its utilization, resulting in significant amount of oxygen tension in the brain tissue ¹⁸ . This suggests that the back diffusion of radioactivity could exist in the form of ¹⁵O₂ from the tissue to the blood. However, the effective contribution of the back diffusion was not visible in the clearance slopes, namely no significant difference was seen in the slopes between the ¹⁵O₂–Hb and H₂₅O–saline injection in the present study. This is attributed to the short life time of oxygen molecule in the brain tissue associated with the high oxidative metabolism in the cerebral tissue. Our results are, however, controversial to those by Seki et al ¹⁰ , who claimed faster clearance after ¹⁵O₂–Hb than H₂₅O–saline administration. Exact reasons are not known, but could be attributed to a couple of methodological factors such as: (a) the observed time–activity curves could have been contaminated with backgrounds from surroundings particularly at the beginning after the ¹⁵O₂–Hb, (b) small size of the animal relative to the injecting volume of 30 to 50 μL in 2 to 3 seconds, corresponding to the perfusion of 0.3 to 0.75 mL/minute per gram, equivalent to the whole brain circulation of this animal, and (c) possible damage of hemoglobin during the bubbling procedures, possibly causing unexpected behavior in the kinetics. Exact reason is not known, and further evaluation is warranted.

There are limitations in this study: (a) Only three independent experiments were carried out on two monkeys. Limited sample size is clearly the weakness of this study. This was simply due to the difficulty of having more animals for this experiment. Given the reproducibility being 17.6% and 15.9% among PET slices corresponding to the faster and slower clearance slopes, however, additional experiments on different animals likely cannot detect significant differences between the ¹⁵O₂–Hb and H₂₅O–saline injections, even if it exists. (b) Positron emission tomography scans were carried out only at the baseline condition during the anesthesia, but not during hyperemia or other stressed conditions. This was due to the need for minimizing the possible damage to the animals, as they were scheduled to participate in another study of evaluating the revascularization therapeutic trials after embolic stroke. Repeating the same set of experiments in the stressed conditions is more challenging to maintain the arterial partial pressure of carbon dioxide, the heart rate, and the blood pressure. We were also concerned that no reason is clearly shown why the clearance slope could be altered after the physiological stimulation. It would however be of an interest to confirm the identical slope values also after the physiological stress is involved, in the future. (c) Due to the limited spatial resolution of the PET device employed in this study (˜10mm full width at half maximum), there must be significant contribution of heterogeneity of CBF and CMRO₂ within a selected ROI. This includes not only the gray matter and white matter heterogeneity, which could be observed as faster and slower clearance slopes in this study, but also in homogeneous distribution in a small local area within the selected ROI. (d) The clearance slopes ranging from 0.37 to 0.45 minutes for the faster component was smaller than the clearance slope-based CBF values reported previously with PET involving the partial volume correction on monkeys ¹⁹ . This could be attributed to different anesthesia protocols. In this experiment, anesthesia was maintained under the continuous intravenous infusion of propofol, while in the earlier study, the anesthesia was maintained by repetitive intramuscular injections of a mixture of xylazin and ketamine. Kaisti et al, ²⁰ demonstrated difference in the change of absolute CBF and CMRO₂ values in normal young adults dependent on the anesthesia. It was also demonstrated that both CBF and CMRO₂ can be reduced during the propofol administration, which might explain at least a part of the difference from the earlier publications. (e) There was significant amount of collateral circulation though the Willis Ring to the collateral hemisphere as can be seen in the accumulated images on Figure 3. With the same reason, injected ¹⁵O-radioactivity should be smeared by the collateral circulation. This may cause systematic bias in the difference between the ¹⁵O₂–Hb and H₂₅O– saline experiments, although this contribution is likely small enough. (f) The initial peak observed after the administration of ¹⁵O₂–Hb but not after H₂₅O–saline sample was attributed to the vascular radioactivity, which was not extracted into the cerebral tissue. The first-pass extraction fraction assessment from this portion as proposed by Raichle et al ⁹ could not be applied due to the slow bolus administration in this study. Smaller amount blood samples needs to be prepared, but the need for a higher radioactivity concentration requires increased production of ¹⁵O₂ radioactivity in the cyclotron target. Our cyclotron has not been prepared for that amount of production. Shorter temporal resolution less than 1-second interval images are limited in the out PET reconstructed images. Further efforts are needed in order to assess the first-pass extraction fraction on local brain regions from the sequential PET images.

In conclusion, dynamic PET imaging was performed to investigate differences in the clearance rates of cerebral ¹⁵O-radio-activity after bolus injections of ¹⁵O₂–Hb and H₂₅O into the ICA. Analysis of the bi-exponential time–activity curves suggested that a significant difference between the rates is unlikely for both the early and late stage data. Even though the number of experiments should be increased to improve the statistics, the results imply that the mathematical model presented in Figure 1 is adequate for quantitative assessment of CMRO₂ and may encourage further development of PET protocols to improve performance and logistics.