Histochemical Correlates of ¹⁵O-Water–Perfusable Tissue Fraction in Experimental Canine Studies of Old Myocardial Infarction

Estimation of Functional Images

The radioactivity concentration in a small element of a myocardial region after administration of H₂¹⁵O has been expressed by assuming a single-tissue compartment model, as described previously (8,12,13). This formulation includes 3 parameters: regional MBF, PTF, and arterial blood volume (V_a). The adoption of PTF has implications for the interpretation of MBF, such that the measurement exclusively represents the mean regional MBF to that mass of tissue within the ROI that is capable of rapidly exchanging water, that is, the mass of tissue defined by PTF. This MBF, because corresponding to blood flow only to perfusable tissue, is free from the so-called partial-volume effect. “MBF_p” is the term adopted to describe this parameter, which has units of mL/min/g of perfusable tissue. Another blood flow is defined and used in this article as a product of MBF_p and PTF. This blood flow represents the average to the volume of the ROI and has units of mL/min/mL of ROI. “MBF_t” is the term adopted for this blood flow.

MBF_t, PTF, and V_a can be estimated with H₂¹⁵O and PET using a single-compartment kinetic model. The model equation has been fitted to the PET tissue and blood–activity curves by a nonlinear least squares fitting analysis (NLLSF) (8–10,12,13). In this study, an alternative approach—the weighted integration technique developed by Alpert et al. (14) and Koeppe et al. (15)—was applied so as to generate functional images of PTF and MBF_t (not MBF_p) simultaneously. The weighted integration method essentially provides the same results as NLLSF but requires a much shorter time for computation. Two time-dependent functions of unity and time were used, as suggested by Alpert et al. (14), and the parametric images were calculated using a look-up-table procedure. In this calculation, V_a was estimated from the C¹⁵O blood volume and from the extravascular tissue density images, with an assumed venous fraction in the myocardium (0.10 mL/g), as reported previously (8,16,17). The extravascular tissue density images were calculated from the C¹⁵O blood volume and from the transmission datasets (8).

The arterial input function was obtained from the left ventricular time–activity curve, and a previously validated method (13) was used to correct for incomplete recovery of the left ventricular region and spillover from surrounding tissue. The PTI (9,10) was then calculated for a given ROI as: Eq. 1

Animal Preparation

The subjects consisted of 12 mongrel dogs (body weight range, 10–20 kg; age range, 2–3 y) with OMI. All experiments were performed according to a protocol approved by the Committee for Ethics and Clinical Research of the Research Institute for Brain and Blood Vessels, Akita, Japan. During open-chest surgery, the main trunk of the left anterior descending branch of the left coronary artery was occluded in each dog. After at least a month of recovery (35–45 d), a series of echocardiographic, electrocardiographic, and PET studies was performed.

After an overnight fast, the dogs were sedated with 4 mg acetopromazine by intramuscular injection. Anesthesia was induced with intravenous thiopental sodium (25 mg/kg). The animals were intubated and mechanically ventilated with a mixture of oxygen, nitrous oxide, and air. Anesthesia was maintained by inhalation of 0.5%–1% halothane. Of the 12 PET studies, 8 were of the open chest, and adhesions of the myocardium to the chest wall and lungs were carefully removed before the scan.

Catheters were placed in the femoral artery to monitor arterial blood pressure and H₂¹⁵O input and in the jugular vein to infuse H₂¹⁵O. Electrocardiography findings and arterial blood gases were monitored throughout the procedure.

Echocardiography

Standard 2-dimensional echocardiography was performed by an experienced echocardiography physician on the same day before the PET study to evaluate wall motion (dyskinesis, akinesis, hypokinesis, and normokinesis) in the septum and in the anterior, lateral, and inferior walls.

PET Procedures

PET was performed using a Headtome-III tomograph (Shimadzu, Kyoto, Japan) (18), which enabled data acquisition in 5 planes simultaneously over an axial field of view of 7.5 cm. All emission and transmission data were reconstructed using a Butterworth filter to provide an in-plane spatial resolution of 10.0 mm full width at half maximum for both emission and transmission scans at the center of the field of view. The axial resolution was 13 mm full width at half maximum at the center of the field of view. The slice interval of this scanner was 15 mm. Data were acquired at 2 bed positions to provide adequate sampling for 10 slices over the axial field of view.

A 20-min transmission scan was acquired at each bed position. The blood-pool images were obtained at each of the 2 bed positions (a 2-min acquisition at each position) after inhalation of 3.7 GBq C¹⁵O (19). Arterial blood samples were taken every minute during scanning, and the radioactivity concentration in the whole blood was measured using a NaI well counter that was cross-calibrated with the PET scanner.

After 15 min to allow for decay of ¹⁵O-radioactivity to background levels, 370 MBq H₂¹⁵O was infused intravenously during 2 min, and then 2 dynamic scans were obtained at each of the 2 bed positions. The 20-frame, 6-min dynamic PET scan was started when radioactivity appeared in the right ventricle and consisted of 6 frames of 5 s each, 6 frames of 15 s each, and 8 frames of 30 s each (20). Between these 2 scans, radiolabeled microspheres (¹⁴¹Ce, 3 MBq) were injected into the left ventricle (n = 8 studies) for measuring regional MBF ex vivo.

An additional static scan was initiated 10 min after injection of a 3.7-GBq bolus of H₂¹⁵O at the 2 bed positions to measure the H₂¹⁵O equilibrium distribution in the body (both in the blood and in the myocardium). Because of the short half-life of ¹⁵O, the activity injected was considerably high, and this method is not practical for clinical human studies.

Histochemical Analysis

After the PET study, the dogs were killed with concentrated potassium chloride. Guided by the laser positioning lights attached to the PET device, we marked the imaging positions directly on the myocardium for the open-chest dogs and on the chest for the closed-chest dogs. The heart was then excised and sliced at a thickness of 7.5 mm to correspond to each PET acquisition plane. The myocardial slices were stained with triphenyltetrazolium chloride (TTC) and then by the elastica Masson (EM) and hematoxylin-eosin (HE) procedures.

Comparison with H₂¹⁵O Equilibrium Image

After injection of a large bolus of H₂¹⁵O, equilibrium is reached when the distribution of H₂¹⁵O no longer reflects flow. More than 10 min are required to reach this point. Thus, PTF was calculated from a single static equilibrium image (equilibrium method) measured at more than 10 min. First, this image was normalized by calibrating the left ventricular region to the blood density (1.06 g/mL). The C¹⁵O blood-pool image (V_B) was then subtracted from the normalized image to provide a distribution of the perfusable tissue. The water content in the blood was assumed to be 0.86 g water/g blood, and that in the myocardium was assumed to be 0.78 g water/g tissue (11); thus, the resultant image was divided by a factor of 0.91 (0.78/0.86) to provide the quantitative image of PTF in g tissue/mL. PTF values calculated from the dynamic data were then compared with those obtained by the equilibrium method.

Comparison with Microsphere MBF

Estimated MBF (MBF_p) values were compared with microsphere estimates for 8 studies. Five small myocardial samples were taken from the inferolateral wall to determine the regional blood flow with the radiolabeled microsphere, and the MBF_p values were compared with the microsphere flow values. Samples were not obtained from anteroseptal regions so as not to disturb the histochemical analysis around the infarcted zone.

Comparison with Histochemical Results

Regional values of PTF, PTI, and MBF were compared with histochemical and wall motion findings for all ROIs; histochemical infarct surface area was compared with defect surface area on PTF and MBF images; and the fraction of infarcted myocardium in a central slice determined by PTI was compared with that determined by histochemical analysis.

Statistical Analysis

All data were presented as mean ± 1 SD. Specific differences were detected by either a paired or an unpaired Student t test. P < 0.05 was considered statistically significant.

Identification of Infarcted Tissue by PTF and PTI

H₂¹⁵O is a diffusible and chemically inert tracer (21). The early distribution after the administration of H₂¹⁵O reflects regional blood flow (22–24). On the other hand, rapid clearance from high-flow areas and further uptake into low-flow areas cause the H₂¹⁵O distribution, after a sufficient long delay, to be proportional to the distribution of perfusable tissue. According to our simulation study, the H₂¹⁵O distribution is expected to reach equilibrium 10 min after the bolus injection, and this technique is probably the simplest for visualization of PTF. However, because of the need for a large injected activity owing to the short half-life of ¹⁵O (e.g., 3.7 GBq for a 10- to 20-kg animal in this study), application of this technique to clinical studies is not practical. The present approach, which estimates the quantitative PTF (in addition to the MBF_t) from the early dynamic dataset, requires only a tenth of the activity (370 MBq) and provides values that essentially agree with those obtained by the equilibrium technique (Fig. 2).

According to an early work in Europe (25), the radiation dose (whole-body effective dose) for this protocol can be estimated as 3–6 mSv, including C¹⁵O and H₂¹⁵O scans with an administration dose of 3.7 GBq and 370 MBq, respectively. This level of radiation is smaller than that for the typical combined protocol of ¹³N-ammonia and FDG, that is, 5–7 mSv with administration of 740 MBq ¹³N-ammonia and 185 MBq FDG.

This study also showed that the estimated PTF and PTI values were significantly smaller in infarcted regions and correlated with the degree of infarction; that is, regions with transmural infarction had significantly lower PTF (and PTI) values than did regions with nontransmural infarction. In contrast, in asynergic but noninfarcted septal segments, both PTF and PTI were preserved and not significantly different from normal control segments, whereas both MBF_p and MBF_t were reduced. Thus, PTF and PTI values can identify infarcted regions and differentiate infarcted areas from ischemic but noninfarcted areas. In addition, the distribution of tissue necrosis identified by histochemical staining (TTC, EM, and HE) corresponded well visually with the defect in PTF images. Quantitatively, the PTF defect surface area, using a cutoff of 50% of the normal PTF value, agreed well with the histochemically defined infarct area. The average PTI (ratio) for a given slice should correspond to the fraction of histochemically defined nonscar tissue in the slice. We found a good correlation between PTI and the fraction of noninfarcted tissue (Fig. 6), although a small but significant difference was present. This small overestimation of PTI can be explained by the exclusion of only V_a in the calculation of PTF, whereas the calculation of D_ev subtracts both arterial and venous blood volumes from the tissue signal (8–10). This factor causes PTI to be overestimated by approximately 10% in normal myocardium (8), a value that is consistent with our observation.

Thus, by directly comparing PTF and PTI with histochemical findings, we have shown that PTF and PTI can estimate the extent of infarction and the proportion of nonscar tissue in a given area of myocardium in OMI. Hypocontractile ischemic segments that have preserved PTF (and PTI) with a threshold level given in previous articles (8–10) should therefore be viable and can potentially benefit from revascularization.

Validity of Mathematic Procedures

Weighted integration is a convenient method for calculating functional images. Although the H₂¹⁵O methodology requires image manipulation, including blood volume subtraction using the C¹⁵O dataset, the MBF_t and PTF images can be calculated automatically and require only selection of an ROI in the left ventricle. Earlier studies found that this technique provided quantitative values that were consistent with those obtained by the conventional NLLSF approach (14,15,26), and our study confirms this finding. Good agreement between MBF_p and the in vitro microsphere flow measurement in the lateral wall further suggests the validity of this mathematic approach, although a potential limitation in the left ventricular injection of labeled microspheres still exists (i.e., incomplete mixing of the spheres). However, the NLLSF method yielded PTF and MBF_t values in the infarcted segment that were approximately 10% and 15% greater, respectively, than were those obtained with the weighted integration method. Although the validity of MBF quantitation in areas of reduced MBF remains unknown, particularly in OMI segments, this small discrepancy can, at least partly, be explained by the overestimation in NLLSF caused by the limited count statistics in the infarcted region (27). Simultaneous determination of 3 parameters—MBF_t, PTF, and V_a—using NLLSF in areas of low count statistics (i.e., reduced flow or necrotic tissue) was shown to cause not only increased statistical uncertainty but also a systematic bias (overestimation) in both MBF_t and PTF. Suggestions to minimize this systematic bias have included fitting only 2 parameters—MBF_t and PTF—with a measured (fixed) V_a (27); improving the image reconstruction algorithm, as has successfully been shown by Katoh et al. (28); or improving the count statistics of time–activity data using larger ROIs (13).

Recently, Gerber et al. (29) showed that the regional MBF values determined by H₂¹⁵O were significantly greater than those determined by ¹³N-ammonia in infarct regions. Those investigators claimed that the former methodology is not preferred over the latter. Although correct values are not known for that comparison, the discrepancy between the 2 methods may be caused by the limited count statistics in the regions with infarction. Systematic overestimation can also result if the noninvasive input function is not corrected for the limited recovery of the left ventricular (or left atrial) chamber, as has been shown (13). Further investigation is required.

Effects of Flow Heterogeneity

In the penumbra of the myocardial infarction, the microscopic flow distribution is unlikely to have the homogeneity assumed by the model; in theory, systematic errors in PTF as well as in MBF can result (30). Using simulations, Herrero et al. (31) showed that PTF (and hence PTI) could be underestimated by 25% at maximum for tissues containing a mixture of 4 different flows with a minimum flow level greater than 0.10 mL/min/g. Iida et al. (32), in a simulation of dynamic H₂¹⁵O studies in the brain, showed that the blood–brain partition coefficient of water (equivalent to PTF in the heart) could be systematically underestimated by 20% because of flow heterogeneity in regions containing a mixture of gray matter and white matter. However, Herrero et al. also showed that, for mixtures of extremely low-flow tissue (e.g., MBF < 0.01 mL/min/g), which should correspond to scar tissue, the estimated MBF and PTF well represent the values for only the high-flow tissue. Thus, PTF well represents the fraction of nonscar tissue around the myocardial infarction, and MBF should correspond to the flow for nonscar tissue. The work by Herrero et al. suggested that perfusion was low enough in the scar tissue to be neglected in the PTF and MBF calculations. In addition, we observed that NLLSF analysis of averaged ROI regions provided greater PTF values than did pixel-by-pixel calculation. This difference could not be explained by flow heterogeneity. Thus, flow heterogeneity is unlikely to be a significant source of error in practical applications.

Herrero et al. (31) also showed that H₂¹⁵O can diffuse into necrotic infarcted tissue within a PET study period. Diffusion into areas of scarring may explain the slightly but significantly higher PTF values obtained by the longer equilibrium method. Diffusion, along with the limited resolution of the PET scan and partial-volume effects, may contribute to the fact that PTF and PTI did not approach zero in scar tissue. The exact magnitude by which PTF (and PTI) are overestimated in infarcted tissue is unknown, and further systematic study is required.

Study Limitations

This study was designed to validate the underlying hypothesis that PTF and PTI indicate the physical extent of myocardial infarction, in a global sense. The accuracy of this approach in assessing myocardial viability needs to be further studied in a much larger patient population (9,10). Further study is also needed to compare this approach with other independent methods, such as FDG imaging, in clinical settings.

PTI and histochemical staining were compared in open-chest dogs, to avoid the confounding effect of chest-wall artifacts. PTI measurements may be unreliable in the anterior wall in some clinical studies, because spillover from the chest wall in the transmission study can cause systematic overestimation of D_ev and, hence, underestimation of PTI for part of the anterior wall. In this study, anterior wall D_ev in closed-chest dogs was 0.73 ± 0.06 g/mL (n = 4), compared with 0.48 ± 0.05 g/mL (n = 8) in open-chest dogs, resulting in a significant (∼30%) underestimation of PTI in the anterior wall of the closed-chest dogs. This effect is expected to be more pronounced in dysfunctional and dilated hearts. Pericardial fat accumulation around the inferolateral wall may also overestimate D_ev and, hence, underestimate PTI, particularly in obese patients. On the other hand, a study showed that calculated PTF values were relatively homogeneous throughout the left ventricular wall in healthy volunteers, provided that the ROI selection was standardized (20). In addition, recovery of the contractile function of the wall after revascularization was well predicted not only by PTI but also by PTF alone (33). Thus, although PTI is largely unaffected by partial-volume effects caused by limited resolution of the PET scanner and wall motion, the use of PTF, rather than PTI, may be recommended in clinical studies.

Movement of a patient during a study is an important source of error. Transmission scans both at the beginning and at the end of the study are recommended, and significant movement between the 2 transmission scans has to be corrected or the study has to be repeated. Also recommended, as shown by Hermansen (34), is a methodology that does not require separate blood-pool scanning using ¹⁵O-carbon monoxide. Further investigation would be of value.