Comparison of Retinal Transit Times and Retinal Blood Flow: A Study in Monkeys

Abstract

purpose. To determine the correlation between transit times of retinal blood flow calculated from fluorescein angiograms and retinal blood flow determined by the microsphere method.

methods. Two fluorescein angiograms were obtained in each eye of six monkeys, followed by determination of retinal blood flow with labeled microspheres. Angiograms in 10 eyes were analyzed for mean transit time (MTT) and arteriovenous passage time (AVP). MTT was determined in two ways: from dye curves reconstructed by extrapolation of semilogarithmic plots of the recorded curves (MTT_slope) and by a new technique based on an impulse–response analysis (MTT_ir).

results. Mean values (±SD) for retinal blood flow in 10 eyes were 23.2 ± 6.9 mg/min. Corresponding values for MTT_ir, MTT_slope, and AVP were 2.22 ± 0.38, 4.89 ± 5.89, and 1.08 ± 0.14 seconds. There was a weak, but not statistically significant, correlation between retinal blood flow and MTT_ir (r = −0.60, P = 0.06) but no useful correlation between retinal blood flow and either MTT_slope or AVP.

conclusions. The relationship between retinal blood flow and transit times determined from fluorescein angiograms is weak. Of the three transit parameters tested, MTT_ir, determined with the recently developed impulse–response technique, had the best correlation with retinal blood flow. Further studies are needed to determine the ability of these transit parameters to detect a change in retinal blood flow and the possibility that transit times may provide useful clinical information unrelated to absolute values of retinal blood flow.

It is possible that disturbed blood flow through the retina influences the development of several ocular diseases, including diabetes and glaucoma. This possibility cannot be investigated in animal experiments, and several clinical techniques have been introduced for noninvasive evaluation of retinal blood flow in the human eye. Some of these techniques are based on fluorescein angiography. Although mostly used for qualitative studies of the retinal vascular bed, fluorescein angiograms can also give quantitative information. Hickam and Frayser ¹ described a technique for determination of the mean retinal circulation time from fluorescein angiograms. The technique is based on the assumption that the gray-scale level in the angiogram is linearly related to the concentration of fluorescein in the retinal vessels. Densitometric analysis of frames from fluorescein angiograms or digitized video frames can then be used to obtain the fluorescence-intensity curve over retinal arteries and veins showing the change in fluorescein concentration with time. The mean retinal circulation time corresponds to the mean transit time (MTT). MTT can be determined from such dye curves. ^{2 3} It is defined as the vascular volume/flow through the vascular bed and is expressed in seconds. ³ It expresses the average time that a fluorescein molecule spends in the vascular bed, and it corresponds to the time necessary to renew all blood within the vascular bed. MTT is easily determined if the complete dye curves can be recorded. It is the difference in time coordinates of the centers of gravity of the extrapolated arterial and venous curves. However, in the retina, early recirculation makes it necessary to discard a large part of the downslope. The dye curve has to be reconstructed by conventional semilogarithmic extrapolation ² or by the method of Stow and Hetzel, ⁴ in which the whole curve up to the recirculation is used to find the log normal distribution function that optimally fits the first passage. These techniques can be used with reasonable precision if the shape of the bolus permits identification of the time when recirculation starts, but that is not always the case in clinical angiograms. The arteriovenous passage (AVP) time (i.e., the time between the first appearance of the dye in the artery and the vein ⁵ ) is therefore often used as a substitute. We have recently developed a new technique to determine MTT based on an impulse–response analysis that is less dependent on the shape of the bolus. ⁶ With this technique it is possible to analyze many clinical angiograms in which MTT cannot be determined with conventional techniques to study the retinal circulation in diseased eyes. To what extent such an analysis of transit times can provide information about retinal blood flow is not known, and therefore we examined these relationships by comparing MTT and AVP determined from fluorescein angiograms with retinal blood flow determined with radioactively labeled microspheres. 

Materials and Methods

Six young and healthy cynomolgus monkeys were used, four females and two males weighing between 3.7 and 8.6 kg. All investigations were in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the local ethics committee. Experimental glaucoma was induced in one eye by argon laser photocoagulation of the trabecular meshwork. ⁷ Approximately 4 years later, fluorescein angiograms were performed in both eyes, and retinal blood flow was determined. Anesthesia was induced with an intramuscular injection of ketamine (10 mg/kg body weight) and was maintained with an intravenous infusion of pentobarbital (30 mg/ml) at a rate of 12 to 30 μl/min. 

Before blood flow measurements, two fluorescein angiograms were performed in each eye. In some monkeys more than two angiograms were performed, but only the two first were analyzed and included in the study. The pupils were dilated with 0.5% tropicamide. Sodium fluorescein (0.1 ml of 12.5% or 25%) was injected into the tail vein. A scanning laser ophthalmoscope (SLO; Rodenstock, Munich, Germany) was used to record the angiogram with a 20° field centered on the optic disc. Laser and video gains were fixed, and the angiograms were recorded on a video recorder (7355 SVHS; Panasonic, Matsushita Electric Corporation, Osaka, Japan). One hundred twenty-eight images at a rate of 5 or 25 frames/sec were digitized to a matrix of 512 × 512 pixels, starting with three to five images before the dye appeared in the central retinal artery. Observing the video in a dark room to observe the first fluorescence identified the starting point for the sampling of the frames. The recorded dye curves verified that the baseline was included. Each image was aligned to one reference image from the sequence by an interactive alignment procedure to reduce influence of eye movements. The regions of interest (ROIs) were denoted on the reference image. The diameters of these areas were extended slightly over the diameters of the vessels (Fig. 1) . ROIs were placed near the optic disc over the superior and inferior temporal arteriovenous pairs, and dye dilution curves were recorded from the mean gray level of the ROIs. 

Ocular blood flow was determined with the labeled microsphere technique, as previously described. ⁷ The right brachial artery was cannulated for catheterization of the left ventricle of the heart for subsequent sphere injections. Two to 3 × 10⁶ labeled 15-μm spheres were injected intracardially. Blood was sampled from the tail artery during the sphere injection for 60 seconds at each blood flow determination. At the end of the last injection, the monkeys were killed with an intracardiac injection of KCl, and the eyes were dissected. Weight and radioactivity of the tissue and blood samples were used to calculate tissue flow. ⁸ A total of three intracardial injections were performed to determine ocular blood flow at various levels of intraocular pressure, but only the results of the first injection, administered shortly after the fluorescein angiograms, are reported. 

MTT of the retinal circulation was calculated from the dye curves in two ways: by conventional technique based on semilogarithmic extrapolation of the dye curves (MTT_slope) ² and by the recently developed impulse–response analysis (MTT_ir). ⁶ Both analyses require a complete dye curve, and for these analyses a sampling rate of 5 frames/sec is used. MTT_slope is the difference in time coordinates of the centers of gravity of the extrapolated arterial and venous curve. In the MTT_ir analysis the arterial concentration in each frame is regarded as a square-wave impulse (Fig. 2) and the resultant venous curve is the sum of identically shaped responses to a series of arterial square-wave impulses. An iterative curve-fitting procedure is used to find the impulse response that yields the simulated venous curve that is closest to the recorded venous curve. The MTT_ir is given by the time coordinate of its center of gravity. The procedure was somewhat modified from the original method. ⁶ The initial guesses of the points defining the early part of the impulse response (g) were chosen in a slightly different way: n was set to 13 and the impulse response was specified by five parameters, namely g ₁ = g ₂ = g ₃, g ₄ = g ₅ = g ₆, g ₇ = g ₈ = g ₉, g ₁₀ = g ₁₁ = g ₁₂, and \(\sqrt{{\tau}}\) (rather than onτ , avoiding some deadlock situations). τ is the time constant of the final exponential decline of the venous outflow curve. Forcing the g-values to be identical three and three in this way was motivated by the high noise in g, having approximately the same effect as average-smoothing g after calculation, and the simpler optimization worked faster on only five variables. 

AVP was also determined. To obtain data comparable with previous reports on AVP we used a sampling rate of 25 frames/sec for this analysis. It was calculated as the difference in time between the first appearance of fluorescein in the artery and the adjacent vein. The time when the fluorescein concentration reached 10% of its maximum was used to define the first appearance of dye in the vessels. MTT and AVP were determined for the upper and lower temporal vessels in each angiogram. The average of the values for upper and lower temporal vessels was used as an estimate of MTT and AVP for that particular angiogram. The values reported here are the means of two angiograms from each eye. 

Statistical Analysis

Standard linear regression was performed by computer (Statistica, ver. 5.5; Statsoft, Tulsa, OK). 

Results

The study was performed in monkeys with glaucomatous damage in one eye. Data for blood flow through the retina and other ocular tissues have been reported previously. ⁷ The purpose of the present study was to report the comparisons between retinal blood flow and transit values obtained from fluorescein angiograms. Such comparisons could be made in 10 eyes. Angiograms from two eyes could not be analyzed because of technical problems during the recording of the angiograms. 

Table 1 presents the values for retinal blood flow, MTT_ir, MTT_slope, and AVP for 10 eyes. Values for MTT_ir, MTT_slope, and AVP are the average from two angiograms. The mean (±SD) for retinal blood flow for all 10 eyes was 23.2 ± 6.9 mg/min. Corresponding values for MTT_ir, MTT_slope, and AVP were 2.22 ± 0.38, 4.89 ± 5.89, and 1.08 ± 0.14 seconds. 

The relationship between retinal blood flow and the various parameters obtained from the angiograms are shown in Figures 3 4 and 5 . A weak correlation was found between retinal blood flow and MTT_ir (r = −0.60, P = 0.06). No significant correlation was found between retinal blood flow and either MTT_slope or AVP. 

Discussion

The purpose of this study was to determine the relationship between retinal blood flow and retinal transit times by studying fluorescein angiograms. The results of the effect of changes in the ocular perfusion pressure on ocular blood flow have been reported elsewhere. ⁷ In this study, the relationship between retinal blood flow and retinal transit times is reported. 

A new method for determination of MTT was needed because of the problems involved in analyzing angiograms in which the shape of the dye curve does not permit an easy reconstruction of the complete curve. The technique for determining MTT_ir is objective and is based on an iterative curve-fitting procedure. ⁶ The technique for determining MTT_slope includes a subjective element. The operator has to select two end points of a straight line for extrapolation of the downslope of the dye curve. There is a tendency to include too much of the downslope, particularly from the venous curve, and the conventional technique has a tendency to overestimate the transit time. Thus, MTT_slope was always larger than MTT_ir. The MTT_ir technique is less dependent on the shape of the curve, and in our experience it permits analysis of most dye curves. The advantage of the MTT_ir technique over the conventional technique is evident from the SDs of MTT_ir and MTT_slope in the present study. For MTT_ir it was 17% of the mean compared with 120% of the mean for MTT_slope. As can be seen in Table 1 , this large SD occurred mainly because of one animal, in which MTT_slope provided a very large value, 21.39 seconds, because the dye curve was shaped in a manner that made it impossible to identify the time when recirculation began. However, such dye curves are not unusual when elderly patients are examined. If that outlier is excluded, the SD is 36% of the mean ± SD (3.05 ± 1.11, respectively), which is still a variation that is considerably larger than the variation found with MTT_ir. For AVP the variation was on the order of that for MTT_ir. Thus, it is clear that MTT_ir and AVP share the same advantage over MTT_slope: the possibility of analyzing transit times from clinical angiograms on which the time when recirculation begins cannot be identified. 

MTT is a well-defined parameter, and the study was performed to examine to what extent MTT can give information about ocular blood flow. Although MTT_ir has some theoretical advantages over MTT_slope and AVP, it was not clear whether these advantages result in a better correlation with retinal blood flow. Thus, all three measurements of transit time were determined in the present study and compared with retinal blood flow. Although the material is limited, it seems clear that the relationship between retinal blood flow and transit times is a weak one. Only between retinal blood flow and MTT_ir was there a clear tendency toward correlation, but the coefficient of correlation was only −0.60, indicating that other factors, such as vascular volume and measurement errors, have an effect on MTT_ir that is at least as great as that of retinal blood flow. This finding was not unexpected. A similar observation of the correlation between transit times and blood flow has been found in the cerebral circulation. ⁹ Furthermore, this study does not address the question of whether determination of MTT_ir is superior to determination of AVP for detecting a change in retinal blood flow. This is an important question and one that can be resolved by comparing the two techniques in human eyes in which changes in retinal blood flow are induced by changes in blood gas concentrations, for example. 

Thus, it is obvious that results obtained from analysis of fluorescein angiograms should be interpreted very cautiously in terms of blood flow. Nevertheless, that the correlation between transit times and retinal blood flow is weak does not exclude that they may be useful for clinical studies of retinal circulation. The results of the present study indicate that the MTT_ir technique, unlike the MTT_slope technique, permits determination of MTT also from angiograms with badly defined dye curves, which makes the MTT_ir technique useful also for studies on clinical material. A comparison of the respective value of MTT_ir and AVP in clinical studies cannot be made by the present study but should be determined in future studies with clinically obtained data. 

 Presented at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 1997.

 Supported by Grant K97-17X-12224-01A from The Swedish Medical Research Council and from Stiftelsen Kronprinsessan Margaretas Arbetsnämnd för synskadade.

 Submitted for publication October 22, 1999; revised June 2 and October 2, 2000; accepted October 16, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

One frame from a fluorescein angiogram showing the placements of ROIs over the upper and lower temporal vessels.

Figure 2.

 

Principle of the impulse–response analysis of the dye curves to obtain MTT_ir. The arterial concentration in each frame is regarded as a square-wave impulse, and the resultant simulated venous curve is the sum of identically shaped responses to the series of arterial square-wave impulses. An iterative curve-fitting procedure (right) is used to find the impulse response that yields the simulated venous curves that is closest to the recorded venous curve. The MTT_ir is given by the time coordinate of the center of gravity (left) of this optimal impulse response.

Table 1.

 

Retinal Blood Flow, MTT_ir, MTT_slope, and AVP in Two Eyes of Six Monkeys

Monkey	RBF (mg/min)			MTTir (sec)			MTTslope (sec)			AVP (sec)
		Control	Glaucoma	Control	Glaucoma	Control	Glaucoma	Control	Glaucoma
314	19.50	—	1.78	—	2.78	—	0.90	—
575	18.90	24.77	2.28	2.62	2.94	4.68	1.25	1.12
196	—	22.62	—	1.89	—	2.36	—	1.30
Z11	17.00	14.53	2.55	2.81	5.19	21.39	0.92	1.17
Z6	30.50	24.36	1.95	2.08	2.29	2.06	1.09	1.10
579	38.30	21.61	1.77	2.45	2.38	2.82	0.88	1.06
Mean	24.84	21.58	2.18	2.42	3.21	6.72	0.96	1.15

 RBF, retinal blood flow.

Figure 3.

 

Retinal blood flow versus MTT_ir, in 10 monkey eyes. MTT_ir data are the means of the average values for upper and lower temporal vessels calculated from two angiograms in each eye. The regression line was calculated with the least-squares method. MTT_ir = 2.98 − 0.04 × retinal blood flow (r = −0.60, P = 0.06).

Figure 4.

 

Retinal blood flow versus MTT_slope in 10 monkey eyes. MTT_slope = 14.76 − 0.43 × retinal blood flow (r = −0.50, P = 0.14).

Figure 5.

 

Retinal blood flow versus AVP in 10 monkey eyes. AVP = 1.23 − 0.007 × retinal blood flow (r = −0.32, P = 0.37).