Absolute Quantification of Myocardial Blood Flow with H₂¹⁵O and 3-Dimensional PET: An Experimental Validation

Animal Preparation and Surgery

The principles of laboratory animal care and use in accordance with the Animals (Scientific Procedures) Act of 1986 were followed. After a 12-h fast, 9 large white pigs (35 ± 4 kg) were premedicated with intramuscular azaperone (2 mg · kg⁻¹). Anesthesia was induced with thiopentone (25 mg · kg⁻¹) and maintained with propofol infusion (9–20 mg · kg⁻¹ · h⁻¹). The animals were artificially ventilated (15 mL · min⁻¹ · kg⁻¹) with room air and supplemented with oxygen, as required, to maintain blood partial pressure of O₂, partial pressure of CO₂, and pH within physiologic limits.

Teflon catheters (DuPont, Wilmington, DE) were implanted in the femoral artery and vein, through an incision in the groin, to measure aortic pressure, sample blood, and administer radioactive tracers. The heart was exposed through an incision in the fourth left intercostal space. A solid-state pressure gauge (Konigsberg Instruments, Pasadena, CA) was inserted into the left ventricle (LV). The circumflex (n = 4) or the left anterior descending (n = 5) coronary arteries were dissected 2 cm from their origin, and a transit time flow meter (Triton Technology, San Diego, CA) and a hydraulic occluder were implanted. Ultrasonic crystals (Triton Technology) were placed into the inner third of the LV wall in the center of the area to be subtended by the stenosed artery to monitor circumferential fiber shortening. A catheter was inserted in the left atrial appendage for radioactive microsphere injection. The chest was then closed in layers after restoring negative pressure.

Mean arterial, LV peak systolic, and end-diastolic pressures, LV dP/dt (first derivative of LV pressure with respect to time, to assess LV contractility), segmental shortening, coronary blood flow, and 3 electrocardiograph leads were continuously monitored and recorded on a multichannel recorder (DISS, Pinckney, MI). Average values for individual parameters were calculated from digitized data sampled at the time of PET data acquisition.

Experimental Protocol

After baseline hemodynamic measurements, the occluder was inflated slowly through a motor-driven syringe to reduce mean coronary blood flow by ∼ 50%. Once the desired flow rate was achieved, flow was kept stable by adjusting the pressure in the hydraulic occluder for the duration of the 2 PET H₂¹⁵O acquisitions (i.e., under resting and stress [dipyridamole] conditions). MBF was also measured using 15-μm plastic microspheres labeled with either ⁵⁷Co (rest) or ⁴⁶Sc (stress) to high specific radioactivity (NEN Research Products, Boston, MA). The microspheres were suspended in 10% dextran. To ensure a homogeneous suspension for injection, each vial was spun in a vortex until the microspheres were mixed in the solution. Finally, before injection, microspheres were sonicated in warm water (∼40°C) for 15–30 min. Approximately 1.6 × 10⁶ microspheres (0.5–2.5 MBq in 0.5–1.5 mL) were injected into the left atrium over 4 s. A reference blood sample was withdrawn from the femoral artery over 2 min at a rate of 7.5 mL · min^{− 1}, starting 10 s before the injection of microspheres.

PET Scanning

Scanning was performed with an ECAT EXACT3D (8). The acquisition system of this scanner has a flexible design that can record data in frame and list mode. List-mode acquisition was used in this study because it represents a significant feature of the tomograph, allowing efficiency of data storage and high temporal sampling with flexible frame rebinning posthoc. Emission scanning was performed with an energy window of 350–650 keV. Transmission scanning was performed with a single photon point source (150 MBq ¹³⁷Cs; γ-energy, 0.663 MeV; half-life, 30.2 y) contained in a small pellet that was driven in a fluid-filled steel tube wound into a helix and positioned just inside the detector ring (6,10).

Animals were positioned in the scanner lying on their right side, and a transmission scan was recorded for 10 min. To image the blood pool, C¹⁵O (1.5 MBq · mL⁻¹) was administered through the endotracheal tube for 4 min at a rate of 500 mL · min⁻¹. A list-mode PET scan and serial blood sampling (3 mL arterial blood every minute) started at the end of C¹⁵O inhalation and lasted for 18 min.

Subsequently, MBF was measured using H₂¹⁵O (110 MBq) injected intravenously over 20 s at an infusion rate of 10 mL · min⁻¹. During the 2 PET MBF measurements, radioactive microspheres were injected 40 s after the start of H₂¹⁵O infusion. The total duration of the H₂¹⁵O scan was 10 min. It is important to note that after the administration of C¹⁵O and H₂¹⁵O, list-mode scan data were collected for extended periods to optimize the start of framing relative to the arrival of radioactivity in the FOV and to allow for the possible lengthening of the framing period. This use of what was designated previously as the decay period did not affect the total duration of the study and resulted in a negligible increase in the amount of disk space required. MBF was measured twice in each animal: at baseline with the occluder inflated and after administration of dipyridamole (0.56 mg · kg⁻¹) infused intravenously over 4 min. During dipyridamole-induced hyperemia, the delivery of H₂¹⁵O was timed to start at 5 min from the beginning of dipyridamole infusion. A second C¹⁵O scan was repeated after the second H₂¹⁵O scan (n = 7).

PET Data Analysis

All emission data were acquired in list mode and then sorted into sinogram data off-line on an Ultra 10 workstation (Sun Microsystems, Mountain View, CA). For each list-mode dataset, a total counting rate curve was generated. This curve defines the global tomograph response to the arrival and passage of radioactivity through the FOV and was used to define the frame durations into which the list-mode data were rebinned. The maximum ring difference used was 40 (acceptance angle, ±13.6°), with a span of 9, which results in 559 sinograms per frame, each the sum of 4 or 5 adjacent individual sinograms except for those at the edge of the FOV. Averaging of 2 adjacent views in the sinogram (i.e., mashing) was performed to further reduce data volumes. This process causes only a very slight loss of transaxial resolution (8). Each sinogram consists of 288 projection bins × 144 projection views. The total frame size is 46.4 MB.

Each of the H₂¹⁵O list-mode acquisitions was sorted into 27 frames, a background frame, followed by a 20-s frame to allow for the circulation of the activity, followed by 14 × 5-s, 3 × 10-s, 4 × 20-s, and 4 × 30-s frames to accurately describe the kinetics of the radioactive water in the body. The background frame had a variable length from the start of acquisition to the arrival of activity in the FOV, minus 20 s. The arrival time of the activity in the FOV was determined from the total counting rate curve.

A static C¹⁵O emission image was generated after the integration of the list-mode data for a period of 480 s. The start time for the integration occurred either at 1 min after the end of the C¹⁵O inhalation, thus allowing for a period of equilibration of radioactivity within the cardiovascular system, or, if occurring later, at 10 s after the point at which true coincidence counts began to fall after saturation on those occasions when the early accumulation of scan data was counting-rate limited in the acquisition hardware.

Myocardial Factor Images and MBF Calculation.

Myocardial images were generated directly from the dynamic H₂¹⁵O scans using a factor analysis technique (18). The main goal was to provide vascular and extravascular myocardial images from the dynamic H₂¹⁵O scan that had an optimal signal-to-noise ratio to be suitable for ROI definition. Briefly, factor sinograms were generated by means of linear dimension reduction of the dynamic FORE sinograms (16). The procedure for the creation of the factor sinograms requires estimates of vascular (right and left heart) and myocardial tissue time-activity curves: the variate and covariate factors. In contrast with the original process, in which lung regions were drawn manually to provide all 3 time-activity curves (18), cluster analysis was applied to the dynamic FORE images to automate and optimize this step. Cluster analysis was used to segment the dynamic data to identify regions with similar time-activity curves (19). Three clusters were chosen, representing mixed venous (right ventricular [RV] chamber), arterial (left atrium and LV chamber), and myocardial tissue regions. For each cluster, a probability matrix (cluster image: Figs. 1A–C) was generated together with the associated time-activity curves (Fig. 1, right). A cluster image thus represents a set of regional probability values that describe the degree of association of a given voxel with the cluster time-activity curves. The resulting 3 cluster time-activity curves were subsequently entered into the factor analysis to provide the 2 factor sinograms. Factor images describing tissue and blood distributions were generated by iterative reconstruction (Fig. 2) (18).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Cluster probability images of transaxial planes and corresponding time-activity curves. (A) Venous cluster (right ventricular cavity). (B) Arterial cluster (left ventricular cavity. (C) Myocardial tissue cluster.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Short-axis images obtained from representative study show blood pool (A) measured with C¹⁵O, which labels erythrocytes through formation of carboxyhemoglobin, and distribution of H₂¹⁵O separated in blood (B) and myocardial tissue (C) component. B and C are calculated by means of factor analysis. RV = right ventricle; LV = left ventricle.

Factor images were resliced into 5-mm-thick short-axis images (voxel size, 1.12 × 1.12 × 5.0 mm³) in an orientation perpendicular to the long axis of the LV. This transformation matrix was also used to reslice all transaxial images, including the dynamic water images and the blood volume images.

Starting from the apex, 12 consecutive short-axis planes (each 2 planes corresponding to 1 slice of the anatomic sections of the heart used for the microsphere analysis) were determined visually for ROI definition. ROIs were placed manually on the factor images by defining the inner and outer LV wall and radially dividing the myocardium into 4 segments for the apical planes and 6 segments for the middle and basal planes. The anterior interventricular groove, the point identifying the meridian plane of the anatomic cut, was defined as the separation line between septal and anterior regions. All segments were then grouped into 16 anatomic regions and projected onto the dynamic 3D water image to extract tissue time-activity curves.

Small circular ROIs (∼60 voxels; radius, 5 mm each) were placed in the left atrium on as many consecutive planes as possible (3–6 planes) to obtain the arterial input function. Similarly, a mixed venous blood time-activity curve was defined by placing ROIs in the RV cavity. A global blood volume recovery factor was then determined for the left atrial region by transposing the left atrial ROIs to the blood volume image determined using C¹⁵O.

Arterial, venous, and tissue time-activity curves were fitted to a single-compartment model to give values of regional and global MBF (mL · L⁻¹ · min^{− 1}) and perfusable tissue fraction ([PTF] mL_{exchangeable tissue} · mL_ROI⁻¹) as described (17,20). The PTF-to-ATF ratio, which has previously been termed perfusable tissue index ([PTI] mL_{exchangeable tissue} · mL_{anatomic tissue}⁻¹), represents the proportion of extravascular tissue perfusable by H₂¹⁵O. For comparison with microsphere flow data, MBF PET results were divided by the density of myocardial tissue (1.04 g · mL⁻¹) to express MBF in mL · g⁻¹ · min⁻¹.

Calculation of MBF with Radiolabeled Microspheres

After PET scanning, the pigs were killed with an overdose of barbiturates. The heart was excised and cut into 6 or seven 10-mm-thick slices perpendicular to the long axis of the LV. Each slice was further cut into 8 equally sized wedges using the interventricular groove as a marker point for the meridian cut between the anterior and septal region. Tissue and arterial blood samples were counted in an automated γ-counter (CompuGamma 1282; Wallac Oy, Turku, Finland), and blood flow per gram of myocardium was calculated by dividing the sample radioactivity by sample weight using a standard reference technique (21). To allow comparison with regional PET flow values (22), microsphere MBF data were grouped into 16 standard regions (anterior, lateral, posterior, and septum for the apex; anterior, lateral, inferior, posterior, posterior septum, and inferior septum for the middle and the base of the heart).

Statistics

Results are expressed as mean ± SD. Hemodynamic measurements, PTF and PTI, were compared by means of ANOVA and post hoc Fisher protected least-significant difference test. The 2 series, PET and microsphere MBF and their difference, were tested for homoscedasticity by means of the F test for equality of variances. Linear least-square regression was performed with microsphere MBF as the independent variable against PET MBF to test the deviation from the line of identity. The mean difference and the average between paired measurements (microsphere and PET MBF) were calculated to test the agreement between the 2 measurements according to Bland and Altman (23). A value of P < 0.05 was considered statistically significant.

Hemodynamics

The baseline hemodynamic variables and their changes are given in Table 1. Coronary blood flow and segmental shortening were reduced by ∼50% in the ischemic area without significant changes in systemic blood pressure or global LV contractility (LV dP/dt). The systemic vasodilation induced by dipyridamole significantly reduced aortic systolic and diastolic blood pressure. The flow in the constricted coronary artery was slightly but not significantly increased, as expected, after dipyridamole infusion.

Image Analysis

Representative short-axis factor images of blood pool and myocardium from 1 study are shown in Figures 2B and 2C. The factor images are of excellent quality, and the factor images of the blood pool compare well with the corresponding blood-pool images obtained after inhalation of C¹⁵O (Fig. 2A).

Left Atrial Blood Volume Recovery Factors

One baseline blood volume measurement was unobtainable because of a failure in the data transfer of the associated transmission scan. Left atrial recovery values for the remaining 8 studies were found to be high (0.84 ± 0.06 mL_BLOOD · mL_ROI⁻¹). However, the interstudy variation in recovery factors (range, 0.78–0.92 mL_BLOOD · mL_ROI⁻¹) was substantially greater than the differences obtained for the repeated measurements made in 6 studies (typically 3%; arithmetic mean, − 0.8%; range, −6.6% to 3.6%).

MBF, PTF, and PTI

A total of 256 paired microsphere and PET measurements were obtained in 8 pigs, with microsphere flows ranging from 0.05 to 4.40 mL · g⁻¹ · min⁻¹. Three scans were excluded from the direct comparison because of technical delays between the PET image acquisition and the injection of microspheres. In 1 pig, an additional hyperemic scan obtained during intravenous adenosine infusion (140 mg · kg⁻¹ · min⁻¹) was included in the comparison.

The F test performed on the 2 series of measurements showed equality of variances. The slope of the scatter diagram in Figure 3A (y = 0.15 + 0.97x; r = 0.87; r² = 0.76) was close to unity and the intercept was close to zero, thus indicating good correlation between the gold standard microsphere MBF and PET MBF over a broad range of values. The Bland-Altman plot of the differences between the 2 measurements against the average value is shown in Figure 3B. The mean difference between microsphere and PET MBF was −0.11 ± 0.36 mL · g⁻¹ · min⁻¹ with 95% of the differences lying between the limits of agreement of −0.83 and 0.61 mL · g⁻¹ · min⁻¹.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Regression analysis (A) and Bland-Altman plot (B) of variation of MBF obtained from microspheres and PET in all ROIs. Dashed lines = 95% confidence limits for slope; Diff = difference.

The best agreement between the 2 methods was observed under resting conditions (Fig. 4A) with a mean difference of −0.09 ± 0.22 mL · g⁻¹ · min⁻¹ and limits of agreement of −0.53 and 0.35 mL · g⁻¹ · min⁻¹. During hyperemia, in spite of a similar difference of −0.13 ± 0.45 mL · g⁻¹ · min⁻¹, the limits of agreement were broader: −1.03 and 0.77 mL · g⁻¹ · min⁻¹ (Fig. 4B). The agreement between microsphere and PET MBF in the 2 different ischemic territories during hyperemia appeared to be more precise in the circumflex coronary artery (mean difference, −0.04 ± 0.39 mL · g⁻¹ · min⁻¹; limits of agreement, −0.82 and 0.74 mL · g⁻¹ · min^{− 1}; SE, 0.08) (Fig. 5A) than in the left anterior descending coronary artery territory (mean difference, −0.15 ± 0.53 mL · g^{− 1} · min⁻¹; limits of agreement, −1.21 and 0.91 mL · g⁻¹ · min⁻¹; SE, 0.094) (Fig. 5B), even though no significant difference was detected.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

Bland-Altman plots show agreement of MBF obtained from microspheres and PET under resting conditions (A) and during hyperemia induced with dipyridamole (B). Diff = difference.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

Bland-Altman plots show agreement of MBF obtained from microspheres and PET during hyperemia induced with dipyridamole. Data are presented for group in which left circumflex coronary artery was occluded (A) and for group in which left anterior descending coronary artery was occluded (B). Diff = difference.

Table 2 summarizes the comparison of the accuracy of the measurement of MBF in the various regions of the LV. The apical and posterior regions show a greater variance in the differences between PET and microsphere measurements.

The values of PTF and PTI in control and ischemic tissue at rest (PTF in mL_{exchangeable tissue} · mL_ROI⁻¹: control, 0.66 ± 0.10; ischemic, 0.66 ± 0.11; PTI in mL_{exchangeable tissue} · mL_{anatomic tissue}⁻¹: control, 0.93 ± 0.13; ischemic, 0.94 ± 0.12) and during dipyridamole-induced hyperemia (PTF in mL_{exchangeable tissue} · mL_ROI⁻¹: control, 0.69 ± 0.10; ischemic, 0.59 ± 0.15; PTI in mL_{exchangeable tissue} · mL_{anatomic tissue}⁻¹: control, 0.95 ± 0.17; ischemic, 0.83 ± 0.20) are illustrated in Figure 6. It is worth noting that PTF and PTI are reduced significantly during hyperemia in the territories subtended by the stenotic coronary arteries.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 6.

Histogram shows comparison of PTF (A) and PTI (B) in normally perfused (control) and ischemic myocardium. Values of PTF and PTI are reduced during dipyridamole-induced stress (hyperemia) *P < 0.05.