INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

SPATIAL HETEROGENEITY OF MYOCARDIAL FLOW (2, 9, 15) and its marked stability over periods of several hours (10, 23, 33) have been well documented. Recently, it has been demonstrated that regional heterogeneity of O₂ consumption is matched by flow heterogeneity (13, 26, 34, 36). Such regional flow adjustments are made through arteriolar vasomotor tone, which is the essential determinant of flow distribution (1). Therefore, the pattern of flow heterogeneity will be regulated down to arteriolar-to-capillary levels; in other words, responding to disturbances in the balance between O₂ supply and consumption, the spatial pattern of flow distribution is possibly changeable at these microvessel levels. When the resultant flow distribution no longer fulfills the local O₂ supply-demand matching, it possibly leads to focal necrosis (38) or arrhythmia through the coexistence of metabolically disturbed cardiac cells and normal ones (5). Hence, the evaluation of flow pattern changes at a microvascular level after the various interventions is of great physiological and clinical importance with respect to the understanding of flow and metabolism matching, the nature of local ischemia, and the development of arrhythmias.

Desmethylimipramine (DMI), an ₂-adrenoceptor antagonist, is an ideal molecular tracer for microregional myocardial flows (24, 25). DMI is distributed to tissue in proportion to local flow, nearly completely extracted during a single pass, and stably deposited mainly at capillary-tissue units; therefore, local DMI depositions are proportional to the local flow. It has been shown that the radionuclide-labeled DMI deposition technique combined with quantitative autoradiography (27, 28, 35) provides a high-resolution imaging method for regional myocardial flow distributions. The regions resolved in this method are smaller than the regions supplied by single arterioles (20, 21) or vascular regulatory units (27). Before now, however, this method has not been extended to a two-time measurement in the same heart. In this study, double-tracer digital radiography with ¹²⁵I- and ³H-labeled DMI (IDMI and HDMI, respectively) was developed for a high-resolution, two-time measurement of myocardial flow distribution, e.g., before and after intervention. This method was applied to the study of microregional flow distribution in isolated rabbit hearts with the aim of evaluating 1) time-to-time flow variability in the same region or temporal flow stability and 2) the effect of microsphere embolization on the spatial pattern of flow distribution.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Double-Tracer Digitalradiography

Quantification of digital radiography for tissue tracer content. To confirm the linearity between tracer content in the myocardium and digital radiographic (DRG) density, we quantitated the tissue tracer content (in cpm/µg) by gamma-ray counting (IDMI) or liquid scintillation counting (HDMI) together with DRG measurements. Dog myocardial walls were homogenized together with IDMI, and several tissue pastes of different radioactivity were prepared. Fractions of these pastes were frozen in a 80°C freezer. The frozen pastes were divided into 10-µm-thick slices using Tissue-Tek tissue mount in a cold environment (25°C) of a cryostat microtome. The slices were put onto slide glasses and allowed to air-dry. The TR plate was then exposed to them for 2 days. The radioactive energy stored on the TR plate was scanned by the imaging plate reader (BAS2000, Fujix) and converted into DRG density in arbitrary units (AU) and with a resolution of 100 pixels/mm² through our image processing system (Macintosh G3 with an aid of MATLAB 5.2 software, MathWorks, Natick, MA). The density was extremely uniform in every slice. The mean density within each slice was then obtained after correcting for background density. During the DRG processing, other fractions of the remaining pastes were weighed and subjected to liquid scintillation counting. The first calibration line relating tissue IDMI content to DRG density, determined with the TR plate, was then obtained by linear fitting. The MS plate was also exposed to the same slices for 1 day, and, in like manner, the second calibration line relating tissue IDMI content to DRG density, determined with the MS plate, was obtained. Likewise, the third calibration line relating tissue HDMI content to DRG density determined with the TR plate was obtained. The MS plate could not detect HDMI at all because the sensitivity of the MS plate is too low to detect the extremely low radiant energy of ³H. The slices of tracer-containing myocardial paste were utilized later as the ³H- and ¹²⁵I-graded scales for the image subtraction processing.

Point-spread functions in IDMI digital radiography. The distribution of DRG density inevitably included blurring, characterized by point-spread functions (PSF), the one-half band width of which increases with radiant energy of the tracers. The blurring is negligible in DRG measurements of HDMI because of the tiny radiant energy of ³H; however, the effect of ¹²⁵I-radiation spread on density distribution of IDMI is measurable and dependent on the sensitivity of imaging plates. This IDMI blurring effect must be taken into account when conducting the image subtraction and comparing the IDMI and the HDMI distributions. To this end, four cylindrical IDMI point sources of <30-µm diameter and 10-µm height were made of silk threads soaked in a dilute solution of IDMI in advance. The PSF of the two plates (TR-PSF and MS-PSF) were determined by imaging the ¹²⁵I point sources using the TR and MS plates, respectively, with the assumption that the PSF is circular symmetric.

Image subtraction processing. In the quantitative double-tracer digital radiography, the IDMI- and HDMI-containing myocardial slices of 10-µm thickness were placed in contact with the TR plate for 2 days along with the above ³H- and ¹²⁵I-graded scales and then placed in contact with the MS plate along with the same graded scales for 1 day. The former resulted in the composite DRG density distribution of HDMI and IDMI, and the latter resulted in the DRG density distribution of IDMI only. In parallel with DRG measurements, the tissue tracer contents of the graded scales were measured for the preparation of calibration lines. Corrected for background density, the two DRG density distributions derived from each sample were so superimposed that their correlation coefficient (r) of regional DRG density reached the maximal value using the image processing system. The sample outlines identified by the two DRG density images coincided with each other nearly perfectly because DRG density in the region free from the tracer radiant energy was <5% of the mean DRG density of the tracers. The central 128 × 128-mm² portion, belonging to the same region of the myocardial tissue sample, was then easily drawn out from each digitalradiogram. The sectioned DRG density distributions of IDMI and IDMI plus HDMI were advanced to the following image subtraction processing (Fig. 1).

View larger version (29K): [in this window] [in a new window]   Fig. 1.   Scheme of sample preparation and double-tracer digital radiography using two imaging plates, BAS-MS2040 (MS) and BAS-TR2040 (TR). IDMI and HDMI, 125I- and 3H-labeled desmethylimipramine (DMI), respectively; MS-PSF and TR-PSF, point-spread functions (PSF) for IDMI digital radiography with the MS or TR plates, respectively; LV, left ventricular. Linear transformation from IDMI digital radiogram with the MS plate to that with the TR plate is done using calibration lines described in Fig. 2 (left). The operation × MS-PSF or × TR-PSF denotes convolving a digital radiogram with MS-PSF or TR-PSF, respectively.

Animal Studies

General experimental procedures. Japanese white male rabbits weighing 2.5-3.5 kg were anesthetized with intravenous administration of pentobarbital sodium (30 mg/kg). Heparin was administered intravenously (500 U · ml¹ · kg¹) to prevent coagulation. The hearts were excised quickly, and the aortas cannulated according to the Langendorff technique with Tyrode solution at 37°C. The perfusion of hearts was discontinued only for a few tens of seconds, and, meanwhile, the hearts did not stop beating. The perfusion medium, containing (in mM) 130 NaCl, 10 NaHCO₃, 5 CaCl₂ · 2H₂O, 4 KCl, 1 MgCl₂ · 6H₂O, 0.435 NaH₂PO₄ · H₂O, and 5.56 dextrose, was bubbled continuously with a mixture of 95% O₂-5% CO₂ and filtered through a 5-µm Millipore filter. The coronary sinus flow and thebesian flow into the right ventricle were collected via a cannula placed through the pulmonary artery, and an effluent flow was measured with an inline electromagnetic flowmeter. Aortic root pressure was measured with a pressure transducer connected to the aortic cannula.

Group I: study of myocardial tracer extraction and retention (n = 6). To ensure that HDMI and IDMI were highly extracted during a single passage through the myocardium, the indicator dilution study was done. A mixed bolus of HDMI (10 µCi) and IDMI (1 µCi) was injected with a glass syringe over a period of 2-3 s into the perfusion line at time (t) = 0 s. The dead space volume between the aortic root and the injection site was 3 ml, and the perfusion rate was 60 ml/min at the lowest, leading to the dead time of <3 s. The effluent was collected at 2-s intervals from t = 0 to 40 s and, subsequently, at 5-s intervals until 180 s. The hearts were then arrested by perfusing a saturated KCl solution and weighed.

(1)

Group II: study of methodological error (n = 8). To estimate the methodological error in the double-tracer digital radiography, the myocardial deposition patterns of simultaneously injected HDMI and IDMI were examined. After a mixed bolus injection of HDMI and IDMI over 2-3 s, we allowed the hearts to beat for 20 s, arrested them by perfusing a saturated KCl solution, and weighed them. The full wall thickness sample was excised from the equatorial portions of the left ventricle. After the papillary muscles were removed, the sample was sandwiched in aluminum sheets without compression and immediately put in a 80°C freezer. A frozen sample in a platelike shape was then divided into 10-µm-thick slices from subendocardium to subepicardium using a cryostat microtome. The subepicardial layer, including the large coronary vessels, and the subendocardial layer that were subjected to a warping effect due to an uneven endocardial surface were omitted from the study. The slices were put onto slide glasses and allowed to air-dry. Twelve slices from each heart were then subjected to the DRG processing.

Group III: study of temporal stability of spatial flow heterogeneity (n = 8). To evaluate the temporal stability of spatial flow distribution, IDMI was injected continuously over 90 s, and, subsequently, HDMI was injected over 2-3 s. IDMI was used for the continuous injection because it was reported that IDMI retention was much larger than HDMI retention (25). This matter was also confirmed in this study (group I). After the hearts were allowed to beat for 20 s, we applied the same procedure as described for group II.

Group IV: study of flow distribution with microembolism (n = 8). To evaluate the effect of multifocal embolism on the myocardial perfusion pattern, flow distributions before and after the microsphere injection were determined. IDMI, microspheres, and HDMI were injected sequentially; each injection was performed over 2-3 s at intervals of 20 s. The microsphere-injection solution contained 8 × 10⁴ nonradioactive microspheres (16.5 ± 0.1 µm in diameter, NEN Life Science Products) suspended in 0.2 ml of 0.01% Tween 80 in isotonic saline. Before the microsphere injections, we throughly dispersed the microspheres by vortex mixing. Microscopic observation of the sliced samples obtained later showed almost no aggregation of microspheres. The hearts were allowed to beat for 20 s after the HDMI injection, and the same procedure was then applied as described for group II except that 27 slices were studied from each heart.

Flow distribution analysis. The resultant DRG density distributions (128 × 128 mm² resolved into 100 × 100-µm² pixels) were reconstructed with a pixel size of 400 × 400 µm² (i.e., the order of magnitude of vascular regulatory units) to reduce random error incident to DRG measurements; thus the quantification was enhanced at the cost of spatial resolution. The reconstructed distributions were then normalized with their respective mean DRG densities. The similarity of those normalized distributions, i.e., relative flow distributions, was evaluated with a linear scatter plot of relative DRG density of IDMI versus HDMI in every myocardial region. The spatial flow heterogeneity was quantitated by the coefficients of variation (CV) of the tracer DRG density [CV (in percent) = SD/mean × 100]. The regional flow dispersion due to the methodological error and the stability of flow distribution were evaluated with the local density difference (in percent) given by the average difference of relative DRG density between IDMI and HDMI in one myocardial region (400 × 400 µm²) divided by their mean × 100. It should be noted that the focus of this study was the spatial pattern of flow distribution; that is, region-to-region flow variability was evaluated on a relative basis. Evaluation of absolute flow changes was beyond this study.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Calibration for Quantitative Digital Radiography

View larger version (18K): [in this window] [in a new window]   Fig. 2.   Example of calibration lines; relations between digital radiographic (DRG) density and tissue IDMI content with MS and TR plates (left) and between DRG density and tissue HDMI content with the TR plate (right). AU, arbitrary units.

Figure 3 shows the PSF for IDMI digital radiography. The DRG densities were normalized with the density at the location of an IDMI point source. The lines were obtained by fitting spline functions to the mean data from all digitalradiograms of point sources (n = 4). The shape of the PSF was assumed to be axial symmetry. The one-half band widths determined with the TR and MS plates were estimated to be 160 and 260 µm, respectively. The inset of Fig. 3 is the microscopic image of a point source embedded in Tissue-Tek tissue mount of 10 µm thickness.

View larger version (35K): [in this window] [in a new window]   Fig. 3.   Point-spread functions for IDMI digital radiography. Plots are means ± SD of all data from digital radiograms of IDMI point sources with MS and TR plates (n = 4). Lines were determined by spline function fitting. The one-half band widths were 160 µm with the TR plate and 260 µm with the MS plate. Inset, microscopic image of a point source in cross section.

Animal Studies

1

1

View larger version (17K): [in this window] [in a new window]   Fig. 4.   Myocardial retention residue function [R(t), where t is time] of IDMI and HDMI vs. time in isolated rabbit hearts after bolus injection (group I). Plots are means ± SD of all data (n = 6). Lines are the data traces for one experiment.

In group II, heart rate, mean perfusion pressure, and mean perfusion rate were 182 ± 20 beats/min, 80 ± 6 mmHg, and 8.8 ± 2.1 ml · min¹ · g¹, respectively. The tracer injection had no effect on these values. The representative results for one myocardial slice are shown in Fig. 5 (left), showing relative density distributions of IDMI and HDMI deposition and the scatter plot of relative densities of IDMI versus HDMI. The relative density of the tracers, which normalized with the mean density of tracer distribution, is hereafter referred to as density for the sake of convenience. The density was visualized in 256 levels of black and white gradations with a spatial resolution of 400 × 400-µm² pixels. Intensities are proportional to local flow. Apparently, the two tracer distributions are very similar to each other. The correlation coefficient, the slope of the regression line, CV values of IDMI and HDMI deposition densities, and the local density difference are summarized for each heart (means ± SD of 12 slices) in Table 1. As a whole (12 slices each from 8 hearts), these averaged 0.89 ± 0.03, 0.93 ± 0.08, 21.2 ± 4.0, 21.9 ± 3.8, and 7.9 ± 1.0%, respectively. Differences between CV values of IDMI and HDMI deposition were not significant.

View larger version (80K): [in this window] [in a new window]   Fig. 5.   Comparison of relative flow distribution based on IDMI and HDMI deposition densities. Top and middle: representative distributions of IDMI and HDMI deposition densities within a LV free wall layer in groups II-IV. Intensities are proportional to local flow; thus tracer density distribution represents relative flow distribution. Every image has the dimensions of 12.8 × 12.8 mm, resolved into 32 × 32 pixels (400 × 400-µm pixels). Bottom: linear scatter plots of IDMI vs. HDMI relative DRG densities associated with a pair of density distributions above. Solid lines and equations were obtained by the best fitting regression, the results of which are summarized in Table 1.

In group III, heart rate, mean perfusion pressure, and mean perfusion rate were 173 ± 22 beats/min, 81 ± 3 mmHg, and 8.8 ± 1.3 ml · min¹ · g¹, respectively. The representative results for one myocardial slice are shown in Fig. 5 (middle). The distributions of two tracers are also very similar to each other. The scatter of density plots is not discernibly different from that in group II. The correlation coefficient, the slope of the regression line, CV values of IDMI and HDMI deposition densities, and the local percent difference of density summarized for each heart (means ± SD of 12 slices) are shown in Table 1. As a whole (12 slices each from 8 hearts), these averaged 0.86 ± 0.03, 0.93 ± 0.10, 19.7 ± 3.3, 21.0 ± 3.7, and 8.7 ± 1.4%, respectively. Differences between CV values of IDMI and HDMI deposition were also not significant. Although correlation coefficient values were significantly lower than those of group II, the difference was slight, and distributions of two tracers correlated linearly with statistical significance. Moreover, differences of the local density difference between groups II and III were not significant; assuming the independent methodological and temporal contributions to the local variation of flow, the temporal component is estimated at only 3.6% [= (8.7²  7.9²)^1/2]. These indicate that regional flow is temporally stable over 90 s at least.

In group IV, mean perfusion pressure increased from 84 ± 4 to 86 ± 5 mmHg and mean perfusion rate decreased from 8.2 ± 1.4 to 7.7 ± 1.3 ml · min¹ · g¹ due to the microsphere embolization of 9 ± 2 spheres/mg (wet weight), resulting in the increase in coronary resistance by 8.6 ± 3.3%. Heart rate (195 ± 22 beats/min) was not affected by the embolization. The representative results for one myocardial slice are shown in Fig. 5 (right). The tracer distributions before and after the embolization have a remarkable pattern similarity; control high- or low-flow regions still received high or low flows even after the embolization. The scatter of density plots is slightly larger than those in groups II and III; however, the distributions of two tracers correlated significantly. The correlation coefficient, the slope of the regression line, CV values of IDMI and HDMI deposition densities, and the local density difference are summarized for each heart (means ± SD of 27 slices) in Table 1. As a whole (27 slices each from 8 hearts), these averaged 0.84 ± 0.06, 1.09 ± 0.13, 19.1 ± 4.5, 24.6 ± 5.4, and 10.8 ± 2.4%, respectively. Differences between CV values of IDMI and HDMI deposition were significant; CV, i.e., flow heterogeneity, increased after the embolization. Although correlation coefficient values were significantly lower than those of group II, distributions of two tracers still correlated linearly to the similar degree in group III. The slope of the regression line was significantly higher in group IV than in groups II and III; accordingly, regional high and low flows were more sharply contrasted due to the embolization. The local density difference was significantly higher in group IV than in groups II and III. Assuming the independent methodological, temporal, and microsphere-derivative contributions to the local variation of flow, the component ascribable to the microembolization is estimated at 6.4% [= (10.8²  7.9²  3.6²)^1/2].

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

The present study demonstrated that double-tracer digital radiography using IDMI and HDMI made it possible to measure the region-to-region myocardial flow variability on a relative basis at two different points of time, i.e., the spatial and temporal myocardial flow distribution with resolution of 400 × 400 µm². The regions resolved are comparable in size to vascular regulatory units. The application of this method to isolated rabbit hearts showed that the spatial pattern of flow distribution was fairly stable even at microvascular levels and that the microsphere embolization enhanced flow differences between control high- and control low-flow regions but rather kept the pattern of flow distribution almost as it was.

The microsphere method for intraorgan flow measurements has conflicting problems on statistical and physiological sides. A recent study (31) has demonstrated that even less than 400 spheres per piece are enough to quantitate the flow heterogeneity; nevertheless, a higher-resolution measurement requires a larger number of microspheres per piece. However, the larger the number of microspheres per piece is, the more regional flows are disturbed due to the embolization. In other words, the effect of embolization on flow is augmented as the piece of myocardium decreases. Accordingly, a fairly common practice of using 1-g-order pieces of the myocardium is a compromise, underestimating the true heterogeneity of flows because such pieces are not internally uniform. Although the microsphere method can provide the acceptable accuracy for flow distribution measurements when the regions resolved are diminished down to a 0.1-g piece, even so, the methodological error is two to four times that for the two differently labeled IDMI deposition technique (3, 4).

The delivery of microspheres to small myocardial regions is not governed by flow alone. There are systemic biases in the deposition of microspheres (6, 12). The larger bias, in accordance with the flow proportionality at equivalent binary branch points, is toward preferential or excessive deposition in region of higher flow; the small bias, which caused by preferential mobility of microspheres along straight vessels penetrating the epicardium toward the endocardium, is toward excessive deposition in the subendocardial versus subepicardial regions. This bias, though, will be greater in larger animals because the increased lengths of penetrating vessels offer a greater number of branchings where skimming or separation of spheres and fluid might occur. DMI deposition technique is free from artifacts due to these rheological or geometric effects on delivery of particles into the microvasculture.

The methodological error in the present digital radiography is attributable mainly to the spread nature of IDMI radiation. In fact, IDMI radiation is not perfectly equal in all directions, and the decay of IDMI density deviates somewhat from the assumed circular-symmetrical spread functions. Furthermore, IDMI digital radiography is under the influence of IDMI deposition over sample thickness because the IDMI radiation ranges over several hundred micrometers. In contrast, HDMI is detected exclusively in a sample surface attached to the imaging plate (i.e., the TR plate) because of the low radiant energy of ³H (its -particle path length is ~1 µm on average). However, the error arising from the radiant energy difference between two tracers would be small because the ventricular free wall was sliced most thinly to a 10-µm thickness, and, moreover, the thickness was quite reduced after drying. Decreases in tracer retention also lowered the accuracy of this method; in this study, however, the periods for IDMI and HDMI to be exposed to perfusate were 120 and 30 s at most, respectively. Washout effect over such short periods is negligible because the retention of the tracers is still nearly 95%.

Despite the evaluation with the considerably fine resolution, the temporal flow fluctuations observed over a 90-s period were almost negligible, as inferred from the temporal flow variation of 3.6% (group III). This indicates that the spatial flow distribution in the myocardium is quite stable over this time period and that microvascular vasomotion or arteriolar twinkling, which is considered to cause temporal flow fluctuations, barely contributes to the spatial flow heterogeneity, as reported previously (37). In this study, however, temporal flow fluctuations may be somewhat underestimated because in crystalloid-perfused Langendorff hearts, microvascular vasomotion might be attenuated (7, 30). The absence of blood cells may also reduce the temporal flow fluctuations. The cell distributions to vessels at branching with a low-flow velocity were reported to be highly time dependent (11), and, actually, the relative temporal fluctuations were greatest in low-flow regions, although diminished in high-flow regions (22).

The microsphere injection increased flow heterogeneity significantly (group IV). The CV value increased by >5% with the local flow variation of 6.4%. However, the regional correlation between normalized local flows before and after the microsphere injection was still significantly high, indicating that the patterns of flow distribution were well maintained and stable even under the microsphere microembolization. Furthermore, that the slope of the regression lines was greater in group IV than groups II and III shows a tendency of flows in control high- and low-flow regions to be more sharply contrasted. The mechanism responsible for this phenomenon remains unproven. However, we postulate a potential mechanism as follows. In control higher-flow regions, the microsphere embolization occurred with higher frequency, inducing greater release of vasodilator agents and, accordingly, greater hyperemic response in those regions. Thus, in high-flow regions, the local hyperemic response would overcome the loss of functioning microvessels; in low-flow regions, however, the latter would prevail against the former. Consequently, the relative flow disparity between high- and low-flow regions would be broadened after the embolization with a slight decrease in perfusion rate. A reduced systolic function in the ischemic regions and an enhanced systolic function in the nonischemic regions (18) may also contribute to the increased relative flow difference by attenuating and enhancing the flow impairment due to the microvessel blockades during systole, respectively.

The microsphere embolization consistently increased the heterogeneity of regional flow distribution; however, this embolizing effect on flow distribution weakens when the flow heterogeneity is evaluated with resolution much lower than the present one. That is, the local flow disturbance caused by the embolization is more dispersed and, accordingly, less discerned in larger sample regions. Actually, when the flow heterogeneity was evaluated with a pixel size of 1,600 × 1,600 µm², CV values before and after the embolization were 14.2 ± 4.0 and 17.7 ± 4.6, respectively, and the percent differences between CV values slightly but significantly decreased from 25 ± 12%, evaluated with a pixel size of 400 × 400 µm² to 22 ± 15%. These results affirm the previous studies (3, 4) leading to the conclusion that 16.5-µm microsphere deposition densities, if evaluated with 0.1- to 1-g-order myocardial pieces, provide a fairly good measure of regional myocardial flows.

In the present study, the microsphere embolization decreased perfusion rate, slightly increased perfusion pressure, and, accordingly, increased flow resistance (8.6%). However, in studies with in vivo hearts (14, 16, 17) or blood-perfused isolated hearts (29), such amounts of microspheres as used in this study increased coronary flow through massive release of adenosine or other endogenous vasodilators in the adjacent area of ischemic foci. Such hyperemic response of coronary flow is likely to occur only when coronary flow reserve is sufficiently remained. Actually, repetitive injections of microspheres increased coronary flow as long as the flow reserve was maintained at the control level with no embolization, but, decreased coronary flow after the flow reserve began to decrease progressively (17). That is, whether coronary flow increases or decreases during microsphere injections will depend on the balance between a coronary resistance decrease due to the remaining local hyperemic response and a coronary resistance increase due to the loss of functioning microvessels. In crystalloid-perfused rabbit hearts, however, coronary flow reserve was nearly exhausted (30) or almost a quarter of that in blood-perfused hearts (8). In the present Tyrode-perfused heart experiments, coronary flow reserve would also be reduced considerably, so that hyperemic response due to the embolization could not overcome the loss of functioning microvessels. Consequently, the microsphere embolization did not increase but rather decreased the total perfusion rate. With thoroughly preserved coronary reserve, the microsphere injection would increase the perfusion rate and enhance contrasts between flows in control high- and low-flow regions.

The method developed in this study can be applicable to hearts in situ; as a matter of fact, this method a promising method for in vivo measurements of myocardial flow distribution because of the lower washout effect on tracer escape. Myocardial perfusion rate is mostly lower in blood-perfused hearts than in crystalloid-perfused hearts, and, accordingly, the retention of both tracers will be larger in blood-perfused hearts. Blood perfusion is likely to be more effective in prolonging the HDMI retention. Indeed, HDMI is better extracted than IDMI in blood-perfused hearts, although IDMI was highly extracted in crystalloid-perfused hearts (25). DMI can be expected to be delivered to the tissue in proportion to local whole blood flow because it is carried by both plasma and erythrocytes (25). Errors due to hematocrit changes should be small, as shown by the same authors. DMI might have its own effect on coronary flow by diminishing ₂-adrenergic activity because cardiac receptor affinity is 10⁴ higher for DMI than for catecholamines (32). In the present study, however, there has been no change in cardiac performance and coronary flow during HDMI and IDMI injections either continuously over 90 s or through a bolus. Furthermore, in the previous in vivo studies (27, 28), no changes in both hemodynamics and cardiac performance were observed during HDMI injections even in the hypoxic state. In addition, it has been reported that ₂-adrenergic vasoconstriction was negligible in the normal heart both at rest and during exercise (7, 19). Thus ₂-antagonistic effects of DMI on regional flow will be minimal. Therefore, this double-tracer method using DMI will be suitable to the measurement of the spatial pattern of flow distribution in the heart or other organs where the extraction and retention of DMI is complete.

In conclusion, double-tracer digital radiography with HDMI and IDMI will be of great use in myocardial flow distribution analysis at microvascular bed levels under different pathophysiological conditions in the same heart. The application study of this method showed that the spatial pattern of myocardial flow distribution was temporally stable at a microvascular bed level even when subjected to the microsphere embolization. However, the embolization enhanced the relative flow differences between high- and low-flow regions and, accordingly, increased flow heterogeneity.