Transcutaneous Fluorescence Dilution Cardiac Output and Circulating Blood Volume during Hemorrhagic Hypovolemia

The study was approved by the Institutional Animal Care and Use Committee (University of Southern California, Los Angeles, California). Appropriate guidelines for the use of animals were observed throughout.

The animal preparation has been previously described in detail.6Briefly, seven adult New Zealand White rabbits were maintained anesthetized with isoflurane (concentration: 1.5% during surgery, 0.8–1.0% during measurements), paralyzed with pancuronium bromide (0.1 mg · kg⁻¹· h⁻¹intravenous), and mechanically ventilated with pure oxygen through a tracheostomy. End-tidal partial pressure of carbon dioxide (Pco²) monitored with a capnometer (model 254; Datex, Andover, MA) was maintained at 32–36 mmHg. An 18-gauge catheter was inserted in the left axillary artery for continuous blood pressure monitoring and blood sampling. A venous catheter placed in the left axillary vein was used for fluid injections. A 4-French thermodilution balloon catheter (AI-07044; Arrow, Reading, PA) was inserted into the right femoral vein and guided until the thermistor at the catheter tip was inside the main pulmonary artery. Correct placement of the catheter was verified visually through a median sternotomy. The thermodilution catheter was connected to a cardiac output computer (Sat 2; Baxter, Irvine, CA) to measure thermodilution cardiac output (CO^TD). Core temperature monitored on the cardiac output computer was maintained at 40°C with heat lamps. After the surgery, the animal was turned to the prone position. The right ear was secured with adhesive tape on top of a custom-designed heater shaped to accommodate a rabbit’s ear and was heated to 40°–42°C to induce maximal vasodilation of the ear vasculature.19 

Near-infrared light from a 782-nm laser diode (SRT-F785S; Micro Laser Systems, Garden Grove, CA) modulated at 7.7 kHz was guided to the blood vessels in the midline of the heated ear (fig. 1) through an excitation optic fiber (400-μm diameter) at the center of a bifurcated multifiber optic probe (R400.7; Ocean Optics, Dunedin, FL). The optic probe was held flush in gentle contact with the skin above the central ear artery with an articulated manipulator and stayed in the same position during the entire study. Six detection fibers (400-μm diameter) surrounding the excitation fiber in the probe forwarded the detected fluorescence light to an 830-nm interferential filter (079-2230; OptoSigma, Santa Ana, CA) and a photomultiplier tube (H7732-10; Hamamatsu, Bridgewater, NJ). The photomultiplier output was amplified and demodulated with a lock-in amplifier (SR 830; Stanford Research Systems, Sunnyvale, CA), which also generated the modulation signal for the laser diode output. The wavelengths of excitation (782 nm) and detection (830 nm) were selected to maximize the intensity of the ICG fluorescence.20,21 

A four-channel analog/digital converter module (Powerlab/4SP; AD Instruments, Colorado Springs, CO) provided continuous recording and display of the arterial blood pressure, heart rate, expired carbon dioxide concentration, and the fluorescence and thermodilution curves. A computer program written in LabWindows CVI (National Instruments, Austin, TX) processed the optical signal for on-line estimation of fluorescence dilution cardiac output (CO^ICG).

The animal preparation was allowed to stabilize for 15–20 min to let the ear vasculature reach a stable, fully vasodilated state. The transcutaneous fluorescence intensity measured at the level of the ear was calibrated as a function of circulating blood ICG concentration as described previously.6Briefly, a standard calibration curve was established to relate blood ICG concentration (0–2.25 μg/ml) to the fluorescence of ICG in the animal’s blood measured in a calibration cell (fig. 1). The standard calibration curve was approximated with a quadratic polynomial equation whose second-order term was less than 0 to reflect the progressively reduced fluorescence yield as the ICG concentration increased.20,22Thereafter, a 1,000-μg dose of ICG (1 ml of 1 mg/ml in 5% dextrose solution) was rapidly infused through the distal port of the thermodilution catheter. Five arterial blood samples (1.5 ml) were withdrawn between 1 and 5 min after the ICG infusion when the ICG was homogenously mixed in the circulating blood while the transcutaneous fluorescence intensity was recorded. The blood samples were placed in the calibration cell, and their fluorescence was measured. The relation between transcutaneous fluorescence intensity and ICG fluorescence in blood was linear and passed through the origin of the axes, yielding a proportionality factor between ICG fluorescence measured in vivo  and in vitro  fluorescence in the cell. Multiplication of the quadratic equation by this proportionality factor produced the equation that was used to convert the transcutaneous ICG fluorescence signal in terms of ICG concentration in circulating blood.

Five measurements of the baseline cardiac output were obtained with the thermodilution and fluorescence dilution techniques. For each measurement, 1.5 ml iced solution containing 45 μg ICG (0.03 mg/ml in 5% dextrose solution) was bolus injected through the proximal port of the thermodilution catheter while the ventilator was stopped at end-expiration. Analysis of the ICG dilution trace also yielded circulating blood volume (BV^ICG) as described below. We waited 3–5 min between measurements for the transcutaneous ICG fluorescence signal to return to 0.

After the baseline readings, blood (30–50 ml) was withdrawn from the arterial catheter in a heparinized syringe until exhaled Pco²began to decrease to 28–29 mmHg, indicating that cardiac output had diminished. Ventilation was readjusted to normocapnia. Then, calibration of the ICG transcutaneous fluorescence as a function of blood ICG concentration was repeated to determine whether the calibration factor had changed with hypovolemia. Thereafter, five measurements of CO^ICG, CO^TD, and BV^ICGwere obtained in the hypovolemia condition. Last, the blood was reinfused in the animal through the venous catheter. After a 5-min equilibration period, the calibration factor was measured a third time, after which cardiac output and circulating blood volume were again measured in quintuplicate.

The capability of the ICG fluorescence dilution technique to estimate circulating blood volume (BV^ICG) was tested in four additional rabbits in which BV^ICGwas compared to estimates derived from the Evans Blue dilution technique (BV^EB).13,18These animals, prepared and instrumented as the other preparations, were only studied in baseline conditions. Three 1.5-ml injections of chilled ICG solution (0.03 mg/ml ICG in 5% dextrose solution), equivalent to approximately 13 μg/kg, were performed for triplicate measurement of CO^ICG, CO^TD, and BV^ICG. After return of the ICG signal to baseline, a 1.5-ml arterial blood sample was drawn for measurement of central blood hematocrit and to serve as a blank for spectrophotometric measurement (Spectronic 20; Bausch & Lomb, Rochester, NY) of the plasma optical density needed for estimation of BV^EB. A dose of 6 mg Evans Blue dye (1 ml of 6 mg/ml solution in 0.9% saline) was bolus injected intravenously. The 1-ml syringe containing the Evans Blue dye was flushed five times with saline within 20 s after the first dose injection. Blood samples (1.5 ml) were drawn at 90 s, 3 min, and 5 min after the Evans Blue injection and centrifuged (3,000 rpm) for preparation of 0.5-ml plasma samples. The samples were diluted with 2.5 ml distilled water before measuring their optical density at 620 nm relative to that of the blank plasma sample. To rule out major shifts of the cardiac output and blood volume associated with the Evans Blue injection and blood withdrawals, three additional ICG injections were performed after the last blood sample was obtained for repeated calculation of BV^ICG.

The transcutaneous fluorescence data were converted to blood ICG concentrations using the calibration technique described previously.6The algorithm used in conventional cardiac output computers to process thermodilution curves23was applied to the ICG dilution traces to derive CO^ICG, as previously reported.6All values of CO^ICGand CO^TDwere computed from single injections, as opposed to the average of replicated injections as is common in clinical practice.

After the initial ICG concentration peak, which corresponds to the first passage of the indicator bolus at the site of measurement, the intravascular ICG concentration trace typically exhibits a slow decay as the liver extracts the injected dye from circulating blood (fig. 2). The slow decay phase can be represented by a decreasing exponential function when the amount of ICG is small (<< 1 mg/kg) and the dye is assumed to be distributed in a single vascular compartment cleared through a first-order process.24Back-extrapolation of the slow decay phase to the instant of ICG injection (time = 0) yields the theoretical concentration ([ICG]⁰) that would be obtained if the dye was homogeneously mixed in the entire circulating intravascular volume. Circulating blood volume BV^ICGis calculated as the ratio of the mass of the injected ICG divided by [ICG]⁰.12 

An algorithm for computing [ICG]⁰was devised to dynamically determine the starting point of the exponential fit (T^exp), the duration of the transcutaneous fluorescence signal on which to perform the fit, and the time of injection (T⁰) needed to back-extrapolate the fit and estimate [ICG]⁰(fig. 2). To choose instant T^exp, the average duration of one circulation of the dye around the vascular network (ΔT^circ) was estimated as the time interval between the first-pass concentration peak and the peak of the recirculation hump. Data from our previous study6showed that duration ΔT^circwas approximately 12 s in rabbits. We assumed that homogeneous mixing of the ICG would be observed after a number of circulations (N^circ), such that the exponential approximation could start N^circ×ΔT^circs after the appearance of the dye. The concentration trace reached a stable decay trend 40–60 s after the injection of dye, such that parameter N^circwas set at 3. We verified that use of N^circ= 3, 4, or 5 produced comparable estimates of BV^ICG.

The ICG concentration returned to near the baseline 0 level at 120–180 s after dye injection. To avoid the low fluorescence signal at the trailing end of the trace, the exponential fit was applied to a segment of data lasting 45 s. Last, the instant of dye injection (T⁰) was assumed to occur at a time point that preceded the first-pass peak by half the circulation time (1/2ΔT^circ). This assumption considered that half the circulation time ΔT^circwas required for the dye to travel from the site of injection near the right atrium to the small systemic blood vessels in the ear, where the fluorescence detection occurred.

The plasma concentration at the instant of Evans Blue injection ([EB]⁰) was estimated using standard methods by back-extrapolation of the experimental plasma concentrations of Evans Blue.13BV^EBwas calculated with the following formula:

where m^EBis the injected mass (6 mg) of Evans Blue dye, hematocrit is the central hematocrit, and constants 0.96 and 0.9 represent correction factors for the plasma volume trapped between erythrocytes after centrifugation and for the microvascular hematocrit/central hematocrit ratio.13,25 

Estimates of CO^ICG, CO^TD, and BV^ICGwere analyzed with repeated-measures analysis of variance to investigate the effect of the hypovolemia and reinfusion conditions on these variables. For each animal and each condition, the five measurements obtained in that condition were arranged in random order before performing the analysis of variance to preserve the information related to intraanimal variability, while avoiding any effect of the temporal order in which the five measurements were obtained. Different randomizations were used for reordering CO^ICG, CO^TD, and BV^ICG. The Bonferroni correction was used in post hoc  pairwise comparisons of the values obtained in the three experimental conditions.

The relation between CO^ICGand CO^TDwas analyzed by linear regression analysis. Agreement between the two techniques was investigated using the approach of Sakka et al. ,1which is better suited to experimental designs with multiple measurements obtained in the same subject than the more conventional Bland-Altman technique.26Specifically, the difference CO^ICGminus CO^TDwas approximated with a general linear model comprising a constant term (measuring bias), a subject term (measuring interanimal variability), a condition term (measuring variability between experimental conditions), and an error term (measuring intraanimal variability). Only the main effects were considered. The general linear model analysis of the difference CO^ICGminus CO^TDwas also performed after combining the data obtained in the three experimental conditions, i.e. , considering only the bias, subject, and error terms. For visual purposes, the results of the analysis were represented with a modified Bland-Altman plot26comprising two 95% bands. The narrower band, which reflected the difference between cardiac output measurement techniques for the average of repeated measurements in an animal, was calculated using the interanimal root mean square difference. The wider band reflecting the difference between techniques for single measurements in an animal was calculated using the total (interanimal + intraanimal) root mean square difference.

The algorithm for dynamic analysis of ICG dilution traces and estimation of BV^ICGwas validated in the four test animals in which the average of six values of BV^ICGwere compared to the Evans Blue blood volume BV^EBwith a paired t  test. Values are reported as mean (± SE). In all analyses, statistical significance was set at P < 0.05.

Estimates of CO^ICGand CO^TDaveraged 455 ± 16 and 450 ± 13 ml/min, respectively, in baseline conditions. Cardiac output decreased markedly during hypovolemia (CO^ICG= 323 ± 15 ml/min; CO^TD= 330 ± 13 ml/min) and recovered partially, but not entirely, after blood reinfusion (CO^ICG= 426 ± 16 ml/min; CO^TD= 414 ± 11 ml/min). Similarly, mean arterial pressure decreased from 57 ± 1 mmHg at baseline to 36 ± 1 mmHg during hypovolemia. Mean arterial blood pressure was to 48 ± 2 mmHg after the blood was reinfused.

Cardiac output CO^ICGwas linearly related to CO^TD(fig. 3). The slope of the linear relation (1.13 ± 0.05) was slightly but significantly greater than 1.0, whereas the ordinate (−50 ± 19 ml/min) was less than 0. The variability of the differences between cardiac output estimates obtained with the ICG and thermodilution techniques primarily resided in interanimal differences (table 1). Furthermore, near-identical partitioning of this variability was obtained when the effect of the experimental condition was included in the analysis and when the condition term was not considered. Accordingly, in the former analysis, the condition term, which would have suggested an effect of the experimental condition on the variability between measurement techniques, was not significant (P = 0.05). The average bias between techniques was also not significant (P = 0.12). The total root mean square difference between cardiac output estimates obtained with the two techniques (45 ml/min), which represented all sources of error and bias, was approximately 10% of the average baseline cardiac output.

Cardiac output CO^ICGwas more likely to be less than CO^TDwhen the mean cardiac output was in the low range (< 350 ml/min). Accordingly, the density of points below the zero line was higher toward the left side of the horizontal axis on the modified Bland-Altman plot (fig. 4), in agreement with the greater than 1 slope of the linear regression line between CO^ICGand CO^TD.

In the validation group of animals, the average BV^ICG(205 ± 10 ml) was not different from the circulating blood volume obtained with the Evans Blue technique (206 ± 6 ml). The root mean square difference between blood volume measurements obtained with the two techniques was 8 ml. The measurement of BV^ICGwas reproducible, with a coefficient of variation between repeated measurements in the same animal of 4% (range, 2.3–5.7%).

In the main experimental group, baseline BV^ICGaveraged 204 ± 5 ml or 59 ± 7 ml/kg (range, 51–68 ml/kg), slightly above the reported total blood volume values for the rabbit (54 ± 4 ml/kg).27BV^ICGdecreased significantly during hypovolemia (174 ± 8 ml/min) and returned to near the baseline level after blood reinfusion (191 ± 6 ml). The blood volume decrease varied between animals during the hypovolemic phase for comparable amounts of withdrawn blood (table 2).

The main findings of this study are as follows: (1) The transcutaneous fluorescence dilution technique can be applied in a hemorrhagic hypovolemia model and yields estimates of CO^ICGthat are comparable to those measured with the thermodilution technique. (2) BV^ICGestimated from the analysis of the slow transcutaneous fluorescence decay is nearly identical to blood volume measured with the reference Evans Blue technique and reflects blood volume changes in an acute bleeding–reinfusion model.

The fluorescence dilution technique yielded estimates of CO^ICGthat were strongly correlated with and nearly equal to estimates of CO^TDduring normovolemic baseline conditions, during hemorrhagic hypovolemia, and after blood reinfusion. Small systematic differences between the estimates produced by the two techniques remained relatively constant within each animal and independent of the volume status. Thus, hypovolemia and the ensuing decrease in peripheral perfusion pressure did not prevent the fluorescence dilution technique from yielding accurate estimates of circulating blood ICG concentration and, therefore, calculation of cardiac output.

Blood flow in the rabbit ear is characterized by an active thermovasodilatory mechanism that is modulated by the baroreflex.19,28The rabbit ear provides a close model to the human cutaneous circulation in that skin blood flow is reduced in response to moderate reductions of cardiac output or blood pressure.10,28In an awake rabbit model of severe hemorrhagic hypovolemia, previous work has shown that when cardiac output and blood volume decreased by approximately 50%, skin perfusion diminished by 72%, and the resistance of the skin vascular bed increased markedly.8One aim of the current work was to determine whether the fluorescence dilution technique, which requires a stable peripheral perfusion at the site of the optical probe, would be compromised during hemorrhage. In the animals we used, the volume bled was approximately 17% of the initial total blood volume, which resulted in a 27% reduction of cardiac output. Such conditions, although associated with some peripheral vasoconstriction, were insufficient to preempt transcutaneous detection of circulating ICG fluorescence. This may have been in part because the local vasoconstrictive response to hemorrhage was blunted by vasodilatory heating of the measurement site on the ear19and by anesthesia. Pilot studies in our laboratory (data not shown) have suggested that more severe bleeding (> 30% of the total blood volume) resulted in marked peripheral vasoconstriction that nearly suppressed ear blood flow and could not be counteracted by local heating.

The agreement between cardiac output estimates produced by the fluorescence dilution and thermodilution techniques was likely improved by recalibrating the transcutaneous fluorescence intensity measurements in terms of blood ICG concentration for each experimental condition. However, the relative change of the calibration factor6between experimental conditions was small (average, +8%; range, −8% to +21%) relative to baseline values. This change may have originated from small local blood flow alterations or from changes in the hematocrit or blood composition that are often encountered in hemorrhagic hypovolemia models.12,29Because the estimated cardiac output is inversely proportional to the calibration factor, keeping the same calibration factor would have mildly affected (approximately −8%) the fluorescence dilution measurements of cardiac output and circulating blood volume.

The current study did not investigate the effect of vasopressors whose vasoconstrictive effect on the peripheral circulation could greatly reduce the amount of circulating ICG detectable with a fluorescence dilution probe placed on the skin surface. We also did not test the influence of aggressive fluid resuscitation with crystalloid solution, which could change the binding of ICG to more diluted plasma proteins and modify the optical absorption of incident light and emitted fluorescence by erythrocytes.21Such alterations could affect the calibration of the fluorescence dilution signal in terms of circulating ICG concentration and lead to discrepancies between measurements of CO^ICGand CO^TD.

The linear regression analysis between CO^ICGand CO^TDhad a slope of greater than 1 and an intercept of less than 0 while concurrently a slight upward trend was observed in the modified Bland-Altman plot. These results suggested a tendency for the thermodilution and dye dilution techniques to produce diverging estimates when cardiac output decreased. In such conditions, the thermodilution technique tends to overestimate cardiac output because thermal losses around the catheter increase when venous flow is slower, resulting in an artificially reduced area under the temperature deflection curve.30States of low cardiac output also result in recirculation of the dye through rapid circulatory pathways, with a concomitant slowing of the blood flow in slow circulatory pathways such as the skin. The net effect artificially increases the area under the delayed dye dilution first-pass curve, leading to an underestimation of cardiac output.30,31In the animals we used, ΔT^circincreased significantly from 11.7 ± 0.5 s to 13.4 ± 0.7 s in response to hemorrhage, which was evidence that the circulation of the dye was on average slower when cardiac output was reduced. Our results suggest that measurement of the ICG dilution curve by transcutaneous fluorescent dye detection follows the previously described trend for increasing discrepancy between dye dilution and other measurement techniques at low cardiac outputs.31 

The original Bland-Altman analysis requires that the difference between techniques being compared does not show a systematic trend over the range of the measurement.12,26This condition was not satisfied in our study as shown by the greater than 1 slope of the regression line between CO^ICGand CO^TD. Although the conventional Bland-Altman analysis was not used on our data, the modified Bland-Altman plot provided a useful illustration to distinguish between single-measurement variability and variability for the averages of repeated measurements.1The modified plot also helped in recognizing systematic trends between the two cardiac output measurement techniques.

The difference between CO^ICGand CO^TDvaried from animal to animal but remained relatively independent of the experimental condition. Systematic differences between animals may have had multiple origins, including experimental error in the calibration parameters of the fluorescence dilution technique6and systematic errors related to the thermodilution technique. In a few animals, we noticed that thermodilution cardiac output was overestimated when the tip of the thermodilution catheter was inserted deep in the pulmonary artery, possibly because the thermistor was in contact with the arterial wall. We cannot exclude that the tip of the thermodilution catheter would have been in close proximity to the vessel wall in a few animals leading to a systematic error in the thermodilution measurements. Variability of the difference between CO^ICGand CO^TDon repeated trials in an animal estimated from the intraanimal root mean square difference (approximately 27 ml/min; table 1) was between 6 and 8% of the cardiac output, depending on the experimental condition. Comparable levels of agreement have been reported between conventional dye dilution, thermodilution, and the Fick method.31This observation suggests that the precision of the transcutaneous fluorescence dilution technique relative to standard techniques for cardiac output measurement was in the same range as that of conventional dye dilution.

Measurement of the circulating blood volume based on the analysis of the slow concentration decay of bolus-injected ICG has been validated in humans13,15,16and experimental animals.12,14Our study differed from previous reports in two ways. First, the ICG concentration was calculated based on the fluorescence intensity of the circulating dye measured transcutaneously with an external probe. A substantial advantage of this sensing modality over optical absorption measurements is that living tissues do not fluoresce at the near-infrared wavelengths of peak ICG fluorescence, such that there is practically no background signal in the absence of ICG.6,32Second, previous studies used ICG doses of approximately 200–300 μg/kg, possibly to create a substantial light absorption difference over that of blood hemoglobin, whereas we injected approximately 13 μg/kg ICG. Use of the lower ICG dose was possible because we measured ICG fluorescence rather than absorption and because we used a noise-resistant optical detection scheme centered on a lock-in amplifier. The low dose of ICG quickly distributed in a rapidly circulating blood compartment, as suggested by the fact that we obtained a close match between BV^ICGand the reference Evans Blue blood volume BV^EB, even when the exponential approximation of the ICG decay curve started soon (approximately 35 s) after the first pass peak of the dye curve. In both animals12and humans,15ICG can still be detected for longer than 15 min after injection of a 200- to 300-μg/kg dose, whereas the ICG dose of the current study was almost completely extracted by the liver from circulating blood after 3 min. The low amount of ICG used in our preparations conveniently allowed for CO^ICGand BV^ICGto be frequently measured without interference from the previous dye injections. Using low ICG doses would be advantageous in situations that require frequent assessment of cardiac output and blood volume.

The baseline circulating blood volume in our animals, 59 ± 7 ml/kg, slightly exceeded total blood volume estimates for the rabbit,27possibly because of the repeated 1.5-ml injections of dextrose associated with the cardiac output and blood volume measurements. The actively circulating blood volume is thought to represent approximately 90% of the total blood volume, except in severe hypovolemia, where a larger percentage of the blood volume may reside in slow exchanging compartments.12In our animals, the difference between baseline BV^ICGand bleeding BV^ICG(approximately 30 ml) was comparable to the volume bled (35 ml) only on the average. In most animals, BV^ICGdecreased more after bleeding than was accounted for by the volume bled. The difference likely reflects a shift in the blood volume partition from fast to slow exchanging compartments. In addition, BV^ICGremained slightly below its baseline value after blood was reinfused in the animal, in part to reflect this shift, as well as to account for blood removal (approximately 10 ml) associated with the recalibration of the fluorescence signals and for fluid losses in the extravascular space.12,29Despite these factors, circulating blood volume measured with the fluorescence dilution technique provided a convenient estimate of the circulating volume available for perfusion of vital organs, such as the liver.29 

The pulmonary artery catheter used for clinical measurement of thermodilution cardiac output has been associated in a small number of cases with severe complications related to positioning or maintenance of the catheter, including severe arrhythmias, pulmonary embolism, and catheter-related sepsis.33–35Such complications and the invasiveness of the pulmonary artery catheter have motivated abundant research into alternative and less invasive techniques for measuring cardiac output.4,5Assessment of the circulating blood volume provides additional information related to cardiac preload not available with the thermodilution technique and poorly reflected by traditional central pressure monitoring.17,36If the fluorescence dilution technique can be successfully applied to a practical clinical device, it could allow for simultaneous monitoring of cardiac output and blood volume with minimal invasiveness. Knowing these variables at the same time could assist clinicians in identifying root causes of low cardiac output in surgical and critically ill patients and in deciding on a therapeutic course.7,37The fluorescence dilution technique as implemented in this study requires much smaller doses of ICG than comparable dye dilution techniques that use optical absorption to measure the ICG concentration.5This could be advantageous for rapid monitoring of therapeutic intervention, especially in patients with compromised hepatic function, who metabolize the dye at a slower rate than healthy individuals do.

Pilot experiments in rabbits have shown that circulating ICG is clearly detected by fluorescence dilution from the ear tissue away from the central vessels and from the nasal mucosa just above the upper lip and that the optical signal can be calibrated in terms of circulating ICG concentration to compute cardiac output (personal verbal communication, J.-M. I. M., September 2004). Circulating ICG is detectable in humans by monitoring optical absorption at the level of the finger and the nasal wing,1,5which suggests that transcutaneous detection of ICG fluorescence is also possible. Future studies will aim at validating the fluorescence dilution technique in human subjects and at identifying accessible measurement sites where adequate perfusion can be maintained for measuring transcutaneous fluorescence during surgery and critical illness.

The applicability of the fluorescence dilution technique would be enhanced if the ICG were injected through a peripheral vein catheter. Central injections of the dye were used in this study to facilitate the comparison between fluorescence dilution and the standard thermodilution cardiac output measurements. Peripheral and central injections of the indicator have been reported to yield identical estimates of the cardiac output in humans38and in experimental animals.39 

The authors thank Susan McCord, B.S. (Director of Software Development), and Sumit Yadav, B.S. (Programmer), from the Alfred E. Mann Institute for Biomedical Engineering at the University of Southern California, Los Angeles, California, who assisted in the development of the software for on-line estimation of the fluorescence dilution cardiac output.