Kinetic Analysis of ¹²⁵I-Iodorotenone as a Deposited Myocardial Flow Tracer: Comparison with ^99mTc-Sestamibi

Preparation for Experiment

All procedures were performed in accordance with institutional guidelines for animal research. Preparation of isovolumic, retrograde, RBC- and albumin-perfused rabbit hearts (n = 25) was similar to that reported previously (6,14). Hearts were obtained from male New Zealand rabbits (R & R Rabbitry, Stanwood, WA) weighing approximately 4 kg. The perfusate buffer was a modified Tyrode’s solution containing oxygenated bovine erythrocytes and 22 g/L bovine serum albumin (Fraction V, fatty-acid free; Boehringer Mannheim Corp., Indianapolis, IN). The specific electrolyte concentrations (in mmol/L) were 110 for NaCl, 2.5 for CaCl₂, 6 for KCl, 1 for MgCl₂, 0.435 for NaH₂PO₄, and 28 for NaHCO₃. Glucose concentration was 5 mmol/L and sodium pyruvate was 2 mmol/L. The mean (±SD) pH value was 7.41 ± 0.04, and the partial pressure of oxygen was 316 ± 90 mm Hg. The concentration of RBCs in the perfusate buffer was adjusted to a hematocrit level of 17%–20%.

A fluid-filled latex balloon connected to a Gould-Statham P23ID pressure transducer (Gould, Oxnard, CA) was inserted into the left ventricle. The balloon was inflated to maintain diastolic pressure at 6–10 mm Hg. A coronary venous sampling catheter and needle thermistor (Omega Engineering, Stamford, CT) were inserted into the right ventricle. The hearts were paced at a rate of 180 bpm. Temperature was maintained between 36°C and 38°C with a water-jacketed heating coil and heart chamber. Coronary flow was held constant with a peristaltic pump (Rainin Instruments, Woburn, MA). The flow rate for control hearts was approximately 1.5 mL/min/g of left ventricular wet weight.

After the heart was prepared, an equilibration period of at least 15 min preceded all experimental interventions. During the equilibration period, developed pressure (peak systolic minus diastolic) was stable and averaged 85 ± 15 mm Hg (minimum developed pressure, 60 mm Hg). After equilibration, myocardial perfusion was gradually changed to the experimental flow rate and subsequently held constant by the perfusion pump. Twenty-five hearts were studied, with each heart evaluated at only one flow rate, which ranged from 0.5 to 3.7 mL/min/g of left ventricular wet weight (on the basis of previous work, flow rates greater than 0.5 mL/min/g of left ventricular wet weight are not ischemic (14)). Ten minutes after equilibration at the experimental flow rate, a mixed isotope bolus consisting of ¹³¹I-albumin, ¹²⁵I-iodorotenone, and ^99mTc-sestamibi (7.4–14.8 kBq of each isotope in 0.2 mL perfusate buffer) was injected just above the aortic cannula. Venous sampling was performed as previously described (6).

Radiopharmaceuticals

^99mTc-sestamibi was purchased from Syncor International (Berkeley, CA). Bovine serum albumin was labeled with ¹³¹I (DuPont NEN Research Products, Boston, MA) using a protein iodination technique based on the IODO-GEN reagent (Pierce Chemical Co., Rockford, IL) (18). ¹²⁵I-iodorotenone was prepared as described by Enas et al. (8).

RBC Uptake and Albumin Binding Studies

The availability of free ¹²⁵I-iodorotenone to the myocardium was found by determining the binding of ¹²⁵I-iodorotenone to bovine erythrocytes in the presence and absence of bovine serum albumin in three experiments. All binding experiments included ^99mTc-sestamibi for comparison.

One set of tubes containing the radiotracers, albumin (2.2%), and washed RBCs (21%) in perfusate buffer was incubated at room temperature for periods of 30 s to 35 min, followed by centrifugation to pellet the cells. A second set of tubes without albumin was also incubated. The relative uptake into the cells was then determined by counting the RBC pellets and the supernatants (with or without albumin). Some of the albumin-and-buffer supernatants were further treated with trichloroacetic acid to assess the degree of association between ¹²⁵I-iodorotenone and albumin.

The rate of release of ¹²⁵I-iodorotenone by RBCs to albumin-containing buffer was also studied in one experiment. RBCs suspended in buffer without albumin were loaded with ¹²⁵I-iodorotenone and ^99mTc-sestamibi by a 15-min incubation followed by centrifugation. The RBC pellet was then resuspended in albumin and buffer, and aliquots were removed at times varying from 1 to 30 min. These aliquots were centrifuged, and the radioactivity was counted.

Data Acquisition and Data Analysis

Venous samples and aliquots of a dilution of the mixed isotope injection solution were counted on a gamma counter as previously described (6). Myocardial extraction, washout, retention, and uptake of ¹²⁵I-iodorotenone and ^99mTc-sestamibi were assessed from the measured venous activity. The venous activity was expressed as a fractional venous appearance rate, h(t), computed by the formula: 1 where F denotes the blood flow (mL/s), C_i(t) denotes the venous sample activity (cps/g), and Q₀ denotes the injected activity (cps).

The multiple-indicator dilution technique and spectral analysis (17) were used to assess blood–tissue exchange of ¹²⁵I-iodorotenone and ^99mTc-sestamibi. The fundamental assumption of the multiple-indicator dilution technique is that the intravascular reference tracer (¹³¹I-albumin) accurately measures intravascular perfusion tracer transport and dispersion. Therefore, any differences between the ¹³¹I-albumin venous concentration curve and the curves for ¹²⁵I-iodorotenone and ^99mTc-sestamibi reflect perfusion tracer transit time delays caused by movement into and out of the extravascular space. Mathematically, a linear-systems approach is used in which it is assumed that no interaction occurs between molecules of the perfusate or labeled tracers and that the uptake of any one molecule does not influence the uptake of any other. Using this formalism, each perfusion tracer in the venous output of the heart is modeled as a convolution of the appearance of the intravascular reference tracer with a unit impulse response function that characterizes the extravascular kinetics of the perfusion tracer.

Spectral analysis uses the differences in fractional venous appearance curves for albumin and perfusion tracer (Fig. 1) to quantify extravascular transit time delays for ¹²⁵I-iodorotenone and ^99mTc-sestamibi by producing a spectrum of kinetic components that describe the unit impulse response function of each flow tracer. The impulse response function is the diffusible tracer venous concentration curve that would be observed after an idealized bolus without circulatory tracer dispersion or transport delay.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Fractional venous appearance rate, h(s⁻¹), of ¹²⁵I-iodorotenone, ^99mTc-sestamibi, and ¹³¹I-albumin as function of venous collection time from experiment with flow rate of 1.3 mL/min/g of left ventricular wet weight.

Spectral analysis computes the impulse response by deconvolving the venous concentration curves for the perfusion tracer by the curve for the intravascular reference tracer (¹³¹I-albumin). A nonnegative weighted least squares algorithm was used to model the diffusible tracer fractional venous appearance rate, h_D(t), as having one component that behaves like the reference tracer, h_R(t), along with delayed components, each of which can be represented as the convolution of the reference tracer curve with a decaying exponential: 2 where the unit impulse response function is: 3 and δ(t) is the Dirac delta function yielding the component that behaves like the reference tracer. This undelayed component contains the fraction c₀ of injected activity that never escaped into the extravascular space. Extrapolating the spectral model to infinite time, the jth delayed component contains a fraction c_j of the injected activity that has a mean transit delay time of t_j caused by bidirectional diffusion between vasculature and cell.

A total of 100 nonnegative components were used, with exponential time constants ranging from 1 s to 190 min equally spaced on a logarithmic scale. Because of the relatively short time frame of the experiments (26 min) compared with the mean transit time of the slowest possible spectral component (190 min), the extrapolated spectral component fractions were not constrained to add up to 1. Given m positive delayed components indexed according to increasing mean transit delay time, the first m − 1 components are termed intermediate components and the mth component is termed the slow component. Using this kinetic model, the extraction fraction (D₀) for the diffusible tracer is: 4 and the combined fraction and combined mean transit delay time for the intermediate components are: 5.

The deconvolved net tracer retention, E_net(t), at time t is: 6Net tissue tracer uptake, U(t), provides a measure of tissue tracer content and is computed as the product of tracer delivery by way of myocardial blood flow and net tracer retention: 7The deconvolved fractional escape rate, FER(t), a measure of tracer washout, is calculated as the ratio of the delayed components of the impulse response and the deconvolved net tracer retention, for t > 0: 8

Statistics

Data are expressed as mean ± SD. Statistical analyses were performed using the MATLAB Statistics Toolbox (MathWorks, Natick, MA). Regression lines were obtained using the unweighted least squares method. A t test was used to test hypotheses about the slopes and y-intercepts of individual regression lines. The slopes of regression lines were compared using Welch’s procedure (19). Paired comparisons of kinetic parameters of ¹²⁵I-iodorotenone and ^99mTc-sestamibi were made using both a paired t test and a nonparametric Wilcoxon signed rank test. In all cases, the two tests yielded similar results with respect to the statistical significance of the differences, and the larger of the two values of P is reported. P < 0.05 was considered statistically significant.

RBC and Albumin Binding

¹²⁵I-iodorotenone was observed to bind strongly to bovine RBCs and albumin. In the absence of albumin, an average of 97.1% (±1.1%) of the ¹²⁵I-iodorotenone was associated with the RBC pellet. This binding reached a steady state almost immediately: the 30-s incubation showed 98% uptake and remained constant for 30 min. In the presence of albumin, the RBC–to–¹²⁵I-iodorotenone binding decreased to an average of 53.1% (±0.65%). Again, steady state was reached within 30 s. Precipitation of albumin with trichloroacetic acid indicated that at least 81% (±6.7%) of the activity that was not associated with blood was instead associated with the albumin.

When RBCs loaded with ¹²⁵I-iodorotenone were exposed to albumin and buffer, they immediately released some of the radiotracer into the solution; that is, at 1 min approximately 35% of the radiotracer was released. This release increased slowly for 30 min, at which time the partitioning between the RBCs and the albumin and buffer approached that seen in the initial binding experiments (55% in the RBCs).

^99mTc-sestamibi behaved differently. The level of RBC binding was independent of the presence of albumin. The 30-s incubations showed that 25.4% (±2.9%) of the ^99mTc-sestamibi was bound to the RBCs, and this level increased linearly to 40.6% (±1.3%) at 30 min.

Myocardial Tracer Distribution

Figure 1 shows an example of the distribution of transit times (expressed as fractional venous appearance rates) of the intravascular tracer, ¹³¹I-albumin, and the two perfusion tracers, ¹²⁵I-iodorotenone and ^99mTc-sestamibi. Although the tracer is introduced as a compact arterial bolus, tracer transport through the myocardium resulted in considerable temporal dispersion of all three tracers in the venous effluent. For the intravascular tracer, the fractional venous appearance rate reflected the distribution of transit times through the myocardial vasculature as well as the arterial inflow and venous outflow tubing. For the two perfusion tracers, some tracer molecules failed to escape the capillary and exhibited the same transit time distribution as the intravascular tracer. However, other tracer molecules diffused out of the capillary into the tissue, where they remained trapped or reentered the vascular space later. Qualitatively, the effect of perfusion tracer escape into the extravascular space was reflected in the lower fractional venous appearance rates of ¹²⁵I-iodorotenone and ^99mTc-sestamibi relative to ¹³¹I-albumin during the initial peak of the venous outflow curves. Subsequent reentry of previously extracted tracer was seen during the later portions of the curves when the fractional venous appearance rates for the two perfusion tracers exceeded the rate for the intravascular tracer.

Spectral Analysis

The spectral kinetic components that describe the impulse response function for the ¹²⁵I-iodorotenone and ^99mTc-sestamibi venous concentration curves from the experiment shown in Figure 1 are illustrated in Figures 2A and B. The modeled curves (Figs. 2C and D) were obtained by convolving the measured ¹³¹I-albumin intravascular reference tracer curve with the impulse responses given by the spectral models. For this example, the ¹²⁵I-iodorotenone impulse response (Eq. 3) had an undelayed component that contained a fraction c₀ = 0.15 of the injected activity and four delayed components (Fig. 2A). The slowest component contained a fraction c₄ = 0.63 of the injected activity and had a mean transit delay time of t₄ = 67 min. The three intermediate components (Eq. 5) contained a combined fraction c_1,2,3 = 0.20 of the injected activity and had a combined mean transit delay time of t_1,2,3 = 3.9 min. By comparison, the ^99mTc-sestamibi impulse response had an undelayed component that contained a fraction c₀ = 0.51 of the injected activity and three delayed components (Fig. 2B). The slowest component contained a fraction c₃ = 0.22 of the injected activity and had a mean transit delay time of t₃ = 39 min. The two intermediate components contained a combined fraction c_1,2 = 0.06 of the injected activity and had a combined mean transit delay time of t_1,2 = 35 s.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Spectral model components for ¹²⁵I-iodorotenone (A) and ^99mTc-sestamibi (B) and fitted concentration curves for entire experiment (C) and first minute (D). Lines through ¹²⁵I-iodorotenone and ^99mTc-sestamibi data points are fitted curves. Data are from experiment shown in Figure 1. • = ¹²⁵I-iodorotenone; ○ = ^99mTc-sestamibi; × = ¹³¹I-albumin.

The spectral model parameters exhibited relatively large variability when the 25 individual hearts were compared (Table 1). The number of delayed components, m, in the impulse responses ranged from 2 to 6 for both ¹²⁵I-iodorotenone and ^99mTc-sestamibi. For each experiment, the number of ¹²⁵I-iodorotenone components was greater than or equal to the number of ^99mTc-sestamibi components. On average, the ¹²⁵I-iodorotenone models contained 4.3 delayed components and the ^99mTc-sestamibi models contained 3.2 delayed components. In three fourths of the experiments, the ¹²⁵I-iodorotenone slow-component fraction, c_m, was estimated to be at least 0.51, with a mean transit delay time, t_m, of at least 45 min. The comparable c_m and t_m estimates for ^99mTc-sestamibi were at least 0.17 and at least 26 min, respectively. In 10 experiments, the ¹²⁵I-iodorotenone slow component was assigned a mean transit delay time of 190 min, the upper limit of the spectrum. This also occurred for ^99mTc-sestamibi in five experiments. The combined fraction for the ¹²⁵I-iodorotenone intermediate components, c_{1,…,m − 1}, was not affected by flow (P > 0.1), whereas the combined mean transit delay time, t_{1,…,m − 1}, decreased significantly with flow (P < 0.02). The combined fraction and mean transit time for the ^99mTc-sestamibi intermediate components were not affected by flow (P > 0.2).

The consistency of the modeled curve estimates with the measured data was assessed for each experiment. The convolution of the ¹³¹I-albumin reference curves with the impulse responses given by the spectral models was compared with the observed fractional venous appearance curves for the diffusible tracer. For the 25 experiments, the average root mean square differences, normalized by the root mean square values of the measured curve samples and expressed as percentages, were 3.4% ± 2.0% for ¹²⁵I-rotenone and 3.7% ± 1.5% for ^99mTc-sestamibi. These results indicate that the models fit the data well throughout the 26-min time frame of the experiments despite the variability in the spectral model kinetic parameters. Similarly, estimates of extraction fraction (Eq. 4), net retention (Eq. 6), and washout (Eq. 8) obtained by evaluating the impulse response function (Eq. 3) and its time integral also appeared robust.

¹²⁵I-Iodorotenone and ^99mTc-Sestamibi Extraction

Figure 3 shows the initial extraction fraction values (Eq. 4) as a function of myocardial blood flow for all 25 experiments. Despite the tight binding of ¹²⁵I-iodorotenone to RBCs and albumin, the average initial extraction fraction for ¹²⁵I-iodorotenone was almost twice that of ^99mTc-sestamibi (0.84 ± 0.05 vs. 0.48 ± 0.10, respectively, P < 0.001). The initial extraction fraction declined for both tracers as flow rate increased. However, the rate of decline with increasing flow was greater for ^99mTc-sestamibi than for ¹²⁵I-iodorotenone (P < 0.001).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Initial extraction fraction, D₀, of ¹²⁵I-iodorotenone and ^99mTc-sestamibi as function of blood flow. Data were pooled from 25 experiments, with each data point representing a single experiment. Equations for linear fits to data points are also shown. • = ¹²⁵I-iodorotenone; ○ = ^99mTc-sestamibi.

¹²⁵I-Iodorotenone and ^99mTc-Sestamibi Washout

Washout, measured as FER(t) (Eq. 8), for ¹²⁵I-iodorotenone and ^99mTc-sestamibi was evaluated at multiple times (15 and 30 s and 1, 2, 4, 8, 16, and 26 min) after tracer introduction (Fig. 4). The FER for ^99mTc-sestamibi was significantly faster than ¹²⁵I-iodorotenone at 15 and 30 s (P < 0.01) and insignificantly faster at 1 min (P < 0.3). From 2 to 8 min, the FER was significantly faster for iodorotenone than for ^99mTc-sestamibi (P < 0.04). At 16 and 26 min, no significant differences in FER were seen for the two tracers.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

FER of ¹²⁵I-iodorotenone and ^99mTc-sestamibi as function of time. Data were pooled from 25 experiments, with each data point representing a single time point. • = ¹²⁵I-iodorotenone; ○ = ^99mTc-sestamibi.

The effect of flow on the washout of these two tracers was also evaluated at each of the times listed above. For ¹²⁵I-iodorotenone, increasing flow rate significantly accelerated the rate of tracer washout at all eight times evaluated (P < 0.02). In contrast, a significant flow effect on ^99mTc-sestamibi washout was observed only during the three earliest times (15 and 30 s and 1 min) (P < 0.05).

¹²⁵I-Iodorotenone and ^99mTc-Sestamibi Retention

Net retention of ¹²⁵I-iodorotenone was significantly greater than ^99mTc-sestamibi retention at 1 min (0.77 ± 0.08 vs. 0.41 ± 0.11, respectively; P < 0.001) and 26 min (0.46 ± 0.13 vs. 0.27 ± 0.11, respectively; P < 0.001) after tracer injection. Figure 5 shows the retention (Eq. 6) of ¹²⁵I-iodorotenone and ^99mTc-sestamibi at 1 min and 26 min as a function of flow. Also, increasing flow rate had a greater effect on ^99mTc-sestamibi retention than on ¹²⁵I-iodorotenone retention at 1 min (P < 0.04). However, at 26 min, increased flow had a similar effect on the retention of both tracers (P < 0.4).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

Deconvolved net tracer retention, E_net(t), of ¹²⁵I-iodorotenone and ^99mTc-sestamibi 1 min (A) and 26 min (B) after tracer injection as function of blood flow. Linear regression equations for 1-min data are y = 0.889 + −0.068x, r = 0.773, for ¹²⁵I-iodorotenone and y = 0.59 + −0.10x, r = 0.83, for ^99mTc-sestamibi. Linear regression equations for data in (B) are y = 0.66 + −0.11x, r = 0.75, for ¹²⁵I-iodorotenone and y = 0.43 + −0.09x, r = 0.72, for ^99mTc-sestamibi. Data were pooled from 25 experiments, with each data point representing a separate experiment. • = ¹²⁵I-iodorotenone; ○ = ^99mTc-sestamibi.

¹²⁵I-Iodorotenone and ^99mTc-Sestamibi Uptake

Figure 6 shows the maximum and 26-min net uptake values as a function of flow (Eq. 7). The maximum net uptake corresponds to the product of flow and initial extraction fraction, whereas the 26-min value corresponds to the product of flow and deconvolved net tracer retention at 26 min. For both the initial and the late values, ¹²⁵I-iodorotenone uptake significantly exceeded ^99mTc-sestamibi uptake (P < 0.001), and the rate of increase in uptake with increasing flow was significantly greater for ¹²⁵I-iodorotenone than for ^99mTc-sestamibi (P < 0.001). These data indicate that ¹²⁵I-iodorotenone is superior to ^99mTc-sestamibi as a flow tracer for at least the first 26 min after injection in the isolated rabbit heart.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 6.

Net uptake, U(t), of ¹²⁵I-iodorotenone and ^99mTc-sestamibi as function of blood flow: maximum uptake (A) and uptake at 26 min (B). Linear regression equations for data in (A) are y = 0.08 + 0.78x, r = 0.99, for ¹²⁵I-iodorotenone and y = 0.26 + 0.29x, r = 0.94, for ^99mTc-sestamibi. Linear regression equations for data in (B) are y = 0.29 + 0.26x, r = 0.86, for ¹²⁵I-iodorotenone and y = 0.25 + 0.09x, r = 0.51, for ^99mTc-sestamibi. Line of identity is also shown for reference. Data were pooled from 25 experiments. • = ¹²⁵I-iodorotenone; ○ = ^99mTc-sestamibi.

Comparison of ¹²⁵I-Iodorotenone with Other Flow Tracers

Three general classes of deposited myocardial perfusion tracers exist: potassium analogs, cationic lipophilic tracers, and neutral lipophilic tracers. Comparing ¹²⁵I-iodorotenone with the potassium analogs, its mean extraction (0.84) is in the range reported for ²⁰¹Tl and ⁸²Rb (0.67–0.88) (4,6,21,22). In contrast, mean extraction is considerably greater for ¹²⁵I-iodorotenone than for the two cationic lipophilic tracers, sestamibi and Q12 (0.26–0.55), that have been evaluated with the multiple-indicator dilution technique (3,4,7). In this study, net retention was greater for ¹²⁵I-iodorotenone than for ^99mTc-sestamibi during the 26 min evaluated. Because ^99mTc-sestamibi has been reported to be better retained than ²⁰¹Tl and ⁸²Rb (3,4,6), it follows that net retention is also superior for ¹²⁵I-iodorotenone than for the two potassium analogs. Therefore, in the isolated rabbit heart, ¹²⁵I-iodorotenone combines the desirable extraction characteristics of the potassium analogs with superior net retention.

Extraction and retention of ¹²⁵I-iodorotenone was excellent despite avid binding to RBCs and albumin. RBC and albumin binding are features common to all neutral lipophilic tracers that have been evaluated in the heart. In addition, binding to blood components appears to play an important role in the ability of these agents to trace myocardial perfusion. ⁶²Cu-pyruvaldehyde-bis(N4-methylthiosemicarbazone) and ^99mTc-teboroxime have inadequate extraction and retention kinetics in humans despite initially promising results in experimental animals (5,23–26). For both tracers, RBC or albumin binding is thought to be responsible for the disappointing results in patients (24,27). [bis(N-ethoxy, N-ethyl dithiocarbamato)nitrido ^99mTc] (TcN-NOET) is a neutral, lipophilic perfusion tracer that is undergoing clinical evaluation (28). This agent is the first ^99mTc flow tracer that redistributes like thallium. Although the effect of RBC and albumin binding on the accuracy of this perfusion tracer is not yet clear, the possibility has been proposed that reversible blood component binding may provide an intravascular pool of tracer that could equilibrate with underperfused myocardium in delayed images and explain the redistribution of TcN-NOET (29). Iododesmethylimipramine was initially proposed as a molecular microsphere on the basis of its nearly complete extraction and retention in buffer-perfused hearts (30). However, when the tracer was evaluated in the presence of blood, significant tracer washout occurred because of binding to RBCs and albumin, thus reducing accuracy as a perfusion tracer. Extrapolation of the results of these neutral, lipophilic perfusion tracers to ¹²⁵I-iodorotenone suggests that, although bovine RBCs and albumin had at most a modest effect on extraction and retention in the isolated rabbit heart, blood component binding could play a more important role in other species, including humans.

Spectral Analysis

In a previous report evaluating ⁸⁶Rb and ²⁰¹Tl kinetics, we used a nonparametric deconvolution procedure to estimate the impulse response function that was nonnegative and monotonically decreasing (6). The nonparametric deconvolution used in that report did not yield an analytic expression for the impulse response function. In addition, the data were resampled into equal time bins to expedite convolution. In this study, in contrast, spectral analysis provided an analytic expression for the impulse response function, and the time intervals were allowed to vary according to the sampling protocol. Because of these differences in data analysis, direct comparison of the current spectral analysis to the previous deconvolution procedure is of limited value.

The spectral-model component fractions and mean transit delay times were poorly reproducible when comparing hearts. Three potential explanations exist. First, the 26-min time frame of the experiments may not have provided adequate data to precisely resolve the slow-component fraction and mean transit delay time, particularly at low flows when the ¹³¹I-albumin intravascular reference tracer had increased time dispersion. Second, use of a decaying exponential basis function may be inadequate to accurately describe the kinetics of myocardial tracer distribution. Third, flow affected the extraction, washout, and retention of both tracers. Because these parameters were computed from the impulse response function derived from spectral analysis, the fivefold variation in flow rate may have adversely affected the reproducibility of the spectral model kinetic components.

Despite the variability, convolution of the ¹³¹I-albumin curves with the computed impulse response functions fit the experimentally measured diffusible tracer curves well. The impulse response function derived from spectral analysis has two important advantages. First, it is independent of the form of the input function and would be equally valid in the in vivo rabbit heart after intravenous tracer injection and in the in vitro rabbit heart. Second, the impulse response function reflects only extravascular tracer kinetics, providing a clear separation of intravascular tracer distribution from penetration into tissue.