2.1.

Study Design

In blood, fluorescence intensity was measured in vitro as well as in vivo as a function of the fluorescein concentration. The CI curve in vitro was obtained by measuring the fluorescence intensities (in arbitrary units) of several known sodium fluorescein concentrations in sampled blood. In vivo, the actual concentration in the retinal arteries was unknown. However, by using three different injection doses with a fixed ratio of 1:10:20, the ratio of the peak concentrations in the retinal arteries was known (1:10:20). Due to the difference in optics, the three peak intensities measured in vivo differed from the intensities measured in vitro by an unknown scaling factor. On a log I vs log C scale, the graph of the in vivo points was shifted horizontally and vertically until the in vivo points coincided with the in vitro calibration curve as well as possible, i.e., by minimizing χ². The in vivo concentration values and the intensity scaling factor were derived from the horizontal and vertical shifts, respectively. The values obtained were verified by an independent experiment and two calculations.

2.2.

Instrument

For fluorescein angiography and for the in vitro measurements we used an in-house-designed and -built confocal scanning laser ophthalmoscope (SLO). In contrast with an earlier version of this instrument,⁸ an aspherical front lens of 28 Diopters was used to obtain a larger field of view (40°×23°). The position of the lens could also be adjusted to correct ametropia. A special feature of this SLO was the absence of an overlap of outgoing and incoming light, thus reducing stray light. Figure 1 shows the pupil configuration [Figure 1(A)] and a schematic of the in- and outgoing beams [Figure 1(B)].

Figure 1

Pupil configuration of the SLO designed to reduce ocular stray light: frontal (A) and sagittal views (B). Optical arrangement for in vitro measurements; the 25 Diopter lens and the cuvette have replaced the eye (C).

For excitation of the fluorescein an argon laser beam of 488 nm was used, which scanned a retinal area of 40°×23° (refresh rate of 50 Hz). By means of a confocal retinal aperture in the optics of the SLO, a retinal detection area of 1.35° centered around this laser spot, scanning the retinal area concurrently, was defined to reduce the level of stray light. The diameter of this retinal detection area was much larger than the diameter of the focus of the laser beam (0.06°) to ensure that the detection area fully covered the illumination area even when their centers did not coincide due to residual refractive errors. Before reaching the photomultiplier tube, the returning, descanned light from the fundus passed an OG515 high pass filter to select the green fluorescent light and to block reflected excitation light. The intensity of the output video signal is proportional to the amount of fluorescence light detected by the photomultiplier. The video amplifier was DC coupled to avoid dynamically induced changes in the level.

During image acquisition there were 488 nm flashes (40 ms) at a maximum rate of 0.9 Hz. In between these flashes only a 635 nm red cross was shown for eye fixation.

Retinal exposure limits for thermal and blue-light photochemical retinal hazards were calculated using the revised guidelines of the International Commission on Non-Ionizing Radiation Protection.⁹ The power of the 488 nm laser beam was 2.1 mW in front of the front lens. Even in the case of breakdown of the scanner, the exposure limit for thermal damage was three orders of magnitude higher than exposure from our laser. Considering the exposure limits for photochemical damage, the retinal irradiance of 2.69 mW/cm² allowed a safe exposure time of over half an hour, much longer than the total exposure time per subject (25 s at most).

A bite board with a dental compound and two forehead rests ensured proper fixation of the subject’s head.

Following each angiogram, a blue reflection image of a plate covered with white reflectance paint (Eastman Kodak Company^®) was obtained for calibration of the laser intensity and for the off-line correction of the slight inhomogeneity of the fundus’ illumination (a gradual drop in intensity towards the borders of the image). This white reflectance paint had near unity diffuse reflection of light in a broad spectral range. The correction required for the inhomogeneous fundus illumination was mainly in the horizontal direction. Thirty pixels from the left border of the image and 20 pixels from the right border the pixel intensity was 50 of the maximum (correction factor=2). The area of analysis was always located more centrally in the image.

2.3.

Image Processing

Off-line analysis of the recorded images of 512×512 pixels consisted of several steps. Firstly, the inhomogeneity of the fundus illumination was corrected using the image obtained from the white plate described above. After this image was corrected for being plane instead of concave, and smoothed to remove small structural details in the paint layer, all angiographic images were divided by this image. Second, the minor translational and rotational eye movements during image acquisition were compensated for by translating and rotating each image until two designated vessel crossings coincided with those in a bright reference image taken near the arterial bolus peak. Images taken during major eye movement were rejected. Third, the pixel gray values were corrected for the variation in gain levels during the angiogram. Finally, a straight segment of a retinal artery with a diameter of at least 50 μm was selected to obtain the arterial time–intensity curves. Sixteen adjacent line profiles were selected perpendicular to the vessel axis. Their mean intensity profile was displayed on the monitor. From that, the pixel value in the vessel’s center was designated to constitute one point of the arterial time–intensity curve.

2.4.

Participants

Eleven healthy volunteers (2 males and 9 females, age 19–26 years, mean age 21.8 years) underwent fluorescein angiography. The Medical Ethical Committee of the University Medical Center Utrecht approved the procedures that were in accordance with the Declaration of Helsinki. Written informed consent was obtained from each volunteer after an explanation of the procedure. None of the subjects had ophthalmic or cardiovascular complaints, nor contraindications against fluorescein injection. Their pupils were dilated with two drops of 0.5 tropicamide and two drops of 5 phenylephrine. A cannula (Vasculon, 20 gauge) was inserted into the antecubital vein; 20 mL blood was drawn into ethylene-diamine-tetraacetic-acid (EDTA)-filled tubes. The waist of the SLO scan was adjusted in the upper part of the pupil’s plane in such a way that at least 1 mm separated both the entrance and exit pupil from the border of the iris [Figure 1(A)]. The fundus image was optimized for sharpness and vignetting on a television monitor by adjusting the position of the front lens. Patients were instructed to maintain eye fixation and a stable head position during the angiography.

2.5.

In Vivo Measurements

Three quantities of fluorescein were injected: first 18 mg/m² body surface,¹⁰ after about 100 s followed by 180 mg/m² body surface. Approximately 11 min after the first injection a final injection of 360 mg/m² body surface was given (Figure 2 depicts the three intra-arterial fluorescence peaks resulting from these three injections). To achieve the same bolus shape the volumes injected were kept constant: the highest dose was prepared using the 20 fluorescein solution. The lower doses were prepared by diluting 20 fluorescein solution in 5 glucose to obtain the same volume as the volume of the third bolus.

Figure 2

Time course of the fluorescence intensity measured in a large retinal artery with three subsequent intravenous injections of fluorescein (subject No. 6). The line connects the data points. The arrows beneath the vertical axis indicate the times of injection.

To obtain reproducible compact bolus profiles, a standardized means of injection was essential. In earlier angiograms we obtained severely disturbed inflow curves resulting from fluorescein that was retained in the arm after injection.¹¹ To avoid such irregularities we standardized the procedures. The fluorescein was injected as fast as possible (within 1 s) to obtain a compact and reproducible bolus configuration. All injections were given by the same experimenter. The injections were followed by a 10 mL flush of 5 glucose. The bolus profile was further improved by (1) warming the arm with an infrared lamp prior to the injections to improve circulation in the arm, (2) applying a tourniquet just distal to the injection site to prevent retrograde flow of fluorescein towards the hand during injection, and (3) asking the subject to make a fist 15 times immediately after the injection and release of the tourniquet to enhance circulation in the hand and thus the venous flow. Several minutes after the second and third injections all subjects were asked to move their arm up- and downward to check that no additional inflow of fluorescein occurred several seconds later due to residual fluorescein in the arm.

Image acquisition started 7 s after the injection of the first bolus and continued for about 25 min, interrupted by two breaks about 5 min after both the second and the third injections. During the first bolus passage the image acquisition rate was set to maximum (one image per 1.1 s).

2.6.

In Vitro Measurements

Immediately after fluorescein angiography the venous blood samples were oxygenated by air exposure to achieve (arterial-like) oxygen saturation levels of at least 90: the blood was gently shaken in wide containers to enlarge the contact surface between the blood and air. Measurements showed that the oxygen saturation of the blood after this procedure was at least 90. For calibration purposes, sodium fluorescein concentrations of 0.02, 0.06, 0.10, 0.20, 0.30, 0.50 and 1.0 mg/mL were prepared in 1 mL autologous blood collected prior to the injection of fluorescein. The required fluorescein solutions had been prepared by diluting 20 fluorescein in phosphate buffer ( pH 7.38). By using autologous blood to obtain a calibration curve the same hemoglobin concentration as that in the eye was ensured. The dilution factor by adding fluorescein (20 μL) to the blood (1000 μL) was not significant.

The fluorescence intensities of the blood samples were measured in an experimental setup, shown in Figure 1(C). A 25 Diopter lens was used to focus the laser beam onto the surface of the blood in a 10 mm sq cuvette. This lens was placed in the waist of the scan beam, similar to in the human pupil plane. Measurements in a 10 mm thick blood layer in vitro will result in similar intensities to those that would have been measured in a layer of the same thickness as a large retinal artery (about 100 μm), since 90 of the total fluorescence originates from the superficial 50 μm of the blood column.² The blood sample was kept in motion by a rod that rotated at the bottom of the cuvette to prevent aggregation of the red blood cells and the formation of plasma gaps between the aggregates that might allow a contribution from deeper layers, resulting in higher fluorescence intensities.⁴ The value of the background pixel (mainly an electronic offset and only a small amount of stray light) measured in the blood sample that contained no fluorescein was subtracted from all the intensities measured. All concentrations were prepared and measured in duplicate at room temperature, within a few hours after the vein was punctured. For each concentration, the four measured intensities were averaged. Previous experiments had shown that the fluorescence intensities in blood samples remained constant for at least 2.5 h.

2.7.

Analysis

We have used an ad hoc function to describe the relation between the log dye concentration and log fluorescence intensity within the physiological range of concentrations (below the peak of the curve of about 1–2 mg/mL)^{2 4} and at the beginning of the descending part of the curve observed in the in vitro experiments (Appendix A).

To compare the in vivo data with the in vitro data we used a fit procedure. First, the fit was performed on the data sets of each subject separately. The three peak intensities measured in vivo had to be scaled by a factor to be determined by the fit procedure to the intensities obtained in the in vitro curve (Figure 3, open symbols). In addition, the concentrations during the angiogram were unknown. However, the ratios between the peak concentrations were known and could be used in the fit: the strict injection protocol assured that the shape of the three peaks was similar, with an amplitude ratio of 1:10:20. When using the ratio of the three peak intensities, the residual fluorescence values prior to injection of the second and the third bolus had to be taken into account: the residual fluorescence intensities were converted into concentration values using the function given in Appendix A, and then subtracted from the peak concentrations. Thus, on a logarithmic scale the three in vivo data points had to be shifted both vertically and horizontally to fit the in vitro data. This fit was optimized by minimizing χ², in which the standard deviation was considered to be equal for all data points (Figure 3).¹²

Figure 3

Concentration–intensity curves; comparison between in vitro (•) and in vivo (⋄) data in 11 subjects. The in vitro data points were fitted by the function given in Appendix A (smooth line). The in vivo data points were matched to the in vitro curve by minimizing χ². Both the fluorescence intensities measured in vitro and in vivo are expressed in the same units ( AU=arbitrary units=pixel gray values corrected for the gain), but log(k _fit ) was added to the in vivo intensities (see the note at the top of Table 1).

After transformation to a linear scale, the fit yielded both the first peak concentration and an intensity scaling factor (k _fit ):

Figure 4

Mean concentration–intensity curve measured in vitro (•), fitted by the function given in Appendix A (smooth line) together with the mean in vivo data points that were fitted (⋄). Vertical error bars represent the standard error of the mean, but especially in case of the in vitro data, these are too small to be visible. The ellipses represent the 68 confidence two-dimensional intervals. The broken line shows the result of the fit of the physiological description of the saturation of the CI curve to the data (see Appendix A).

2.8.

Verification of the Fitted Results

The fit yielded both the first peak concentration and an intensity scaling factor (k _fit ). The plausibility of the values was verified in three ways: (1) calculation of the cardiac output from the first peak concentration, (2) blood sampling after the third fluorescein injection to measure the intensity scaling factor (k _{blood sample} ) and (3) optical calculation of the intensity scaling factor (k _optical ).

(1) The peak concentrations should be the result of physiological cardiac output. Cardiac output (CO) (L/min) is equivalent to

(2) To check the intensity scaling factor obtained from the fit (k _fit ), this factor was also obtained from an oxygenated blood sample that was drawn between 10 and 13 min after injection of the third bolus from the contralateral antecubital vein in an EDTA-filled tube (k _{blood sample} ). In a pilot experiment (three angiograms) we have verified that at this time of blood sampling the mixing of fluorescein over the body has been completed by drawing and measuring multiple blood samples between 4 and 26 min after the third injection. This is in agreement with the conclusion of Lund-Andersen et al.;,¹³ who estimated that the net extraction of fluorescein in peripheral capillaries is minimal 4 min after injection.

At least five fundus images were taken immediately before and after this blood sampling. To calculate the in vivo intensity at the time of blood sampling (I_in vivo) the data points before and after blood sampling were interpolated by fitting the decay curve with an exponential function. The intensity scaling factor k was

(3) The intensity scaling factor was also calculated on the basis of comparing in vivo and in vitro optical pathways (k _optical ). These calculations are summarized in Appendix B.

3.1.

In Vivo Measurements

A typical example of a time–intensity curve is shown in Fig. 2. Three main peaks are visible as a result of the three injections of fluorescein. Due to the shape of the compact bolus, recirculation of the bolus is visible as a small peak about 15 s after the main peak. The effect of quenching and self-absorption can be observed in the peak intensities: the third peak is only slightly higher than the second peak, although the dose injected was twice as high as that in the second injection.

3.2.

In Vitro Measurements and Analysis

The concentration–intensity curves determined in vitro in all 11 subjects are depicted in Fig. 3 (closed symbols). Also in Figure 3 the three in vivo data points are plotted (open symbols): the three fitted log peak concentrations constitute the x values. The y values of the three in vivo data points (intensities that are comparable with the in vitro intensities) are derived by summation of the log fitted intensity scaling factor [log(k _fit )] and the log intensities measured in the retinal artery. Figure 4 displays the mean results. In Figure 4 the vertical error bars show the standard errors of the mean. The error bars of most of the data points are smaller than the symbols. The 68 confidence intervals of the first peak concentration and the intensity scaling factor (k _fit ) were determined by calculating (χ _min ²+1) contours.¹² The confidence interval of the intensity scaling factor (k _fit ) was converted into intensity values. The two-dimensional confidence intervals of the in vivo data points are shown as ellipses surrounding the data points (Figure 4). The in vivo data points could be brought to an accurate match with the in vitro curve. The rather small confidence intervals of the fitted parameters (the first peak concentration and the intensity scaling factor, Figure 4) ensure that estimation of the parameters is satisfactory.

The peak intensities of the first bolus are listed in the second column of Table 1, and the peak concentrations following the first injection derived from the fit in the third column. The mean dose given was 33.3 mg (a range of 30.1–37.8 mg), resulting in a mean peak concentration of 0.049 mg/mL [±0.005 standard deviation (SD)]. The intensity scaling factors that result from the fit are listed in the fourth column of Table 1. The mean is 0.41±0.03.

Table 1

3.3.

Verification of the Fitted Results

To verify the peak concentrations obtained from the fit, the cardiac output was calculated. According to Eq. (2), using the mean value of the dose and the first peak concentration, the mean cardiac output would be about 7.6 L/min. This value is within the physiological range.¹⁴

The intensity scaling factor according to the fit (k _fit ) was compared with the results of the two other methods. The intensity scaling factors resulting from the blood samples after the third injection (k _{blood sample} ) are shown in the fifth column, the mean k _{blood sample} was 0.44±0.12. Furthermore, from the optical calculations an intensity scaling factor (k _optical ) of about 0.54 was derived (Appendix B). These three independent determinations of k yield values ( 0.41±0.03, 0.44±0.12 , and 0.54±0.14 ) that are not significantly different.