Blood Velocity ond Volumetric Flow Rote in Humon Retinal Vessels

Transcription

1 Blood Velocity ond Volumetric Flow Rote in Humon Retinal Vessels Charles E. Riva, Juan E. Grunwald, Stephen H. Sinclair, and Benno L. Perrig The distributions of blood velocity and volumetric flow rate in individual vessels of the normal human retina were determined as a function of vessel diameter. The mean velocity of blood, V mean, was calculated from the centerline velocity measured by bidirectional laser Doppler velocimetry (LDV). Volumetric flow rate was determined from V me an and the vessel diameter, D, measured from monochromatic fundus photographs. Diameter of the arteries and veins at the site of the LDV measurements ranged from 39 to 134 /tin and 64 to 177 nm, respectively. Flow velocity correlated with D (P < for both arteries and veins). Volumetric flow rate varied with D at a power of 2.76 ±0.16 for arteries and 2.84 ±0.12 for veins, in close agreement with Murray's law. Calculated from 12 eyes, the average total arterial and venous volumetric flow rates were 33 ±9.6 and 34 ±6.3 /xl/min, respectively. The good agreement between both flow rates suggests that the technique and the assumptions for calculating flow yield results that satisfy mass conservation. Total arterial and venous volumetric flow rates correlated with total arterial and venous vessel cross-section. Volumetric flow rate in the temporal retina was significantly greater than in the nasal retina, but the difference is likely to be due to the larger area of the temporal retina. No difference in flow rate was observed between the superior and inferior retinal hemispheres. Finally, blood velocity in the major retinal vessels measured under normal experimental conditions appears remarkably constant over short (hours) and long (months) periods of time. Invest Ophthalmol Vis Sci 26: , 1985 The retina is one of the few human tissues where the blood circulation can be observed directly and noninvasively. This has, on the one hand, facilitated a detailed description of retinal vascular morphologic changes for a variety of ocular and systemic diseases. On the other hand, in spite of the accessibility of the retinal vasculature and the importance of blood flow measurements, quantitative data on retinal hemodynamics in both health and disease is still lacking. The available information consists primarily of subjective and qualitative assessments of "normal" or "low" flow conditions obtained by fluorescein angiography. Attempts to obtain quantitative information using the fluorescein dye dilution technique have had limited value. Bidirectional laser Doppler velocimetry (LDV) allows direct quantitative and noninvasive measure- From the Scheie Eye Institute, Department of Ophthalmology, School of Medicine, University of Pennsylvania, Philadelphia, PA Supported by NIH Grant EY from the National Eye Institute, the Pennsylvania Lions Sight Conservation and Eye Research Foundation, Inc., and by the F. Roland and Marianne E. Sargent Teaching and Research Fund. Submitted for publication: August 13, Reprint requests: Dr. Charles E. Riva, Scheie Eye Institute, 51 North 39th Street, Philadelphia, PA ments of the centerline velocity of red blood cells (RBCs) in individual retinal vessels. 1 LDV combined with vessel diameter measurements enables the determination of volumetric flow rate in the major retinal vessels. In this article, we present quantitative measurements of blood velocity and volumetric flow rate in individual arteries and veins of the normal human retina and establish their relation to vessel diameter. Total arterial volumetric flow rate is compared with total venous volumetric flow rate to verify the conservation of mass, and the flow rates are correlated with total vessel cross-section. Finally, we demonstrate the stability over time of retinal blood flow velocity when measured under normal physiologic conditions. Materials and Methods Doppler shift power spectra of laser light scattered from RBCs in retinal vessels were recorded simultaneously for two directions of the scattered light using a fundus camera-based LDV system. The centerline or maximum velocity, V max, was determined according to the relation: = K-XeAf/(ncos 3) (0 where Af = f 2 fj. fi and f 2 are the cutoff frequencies of the power spectra recorded for two directions of 1124

2 No. 8 HUMAN RETINAL BLO FLOW / Rivo er ol the scattered light. 2 X is the wavelength of the incident helium-neon laser beam ( nm); e, the axial length of the eye as measured by A-scan ultrasonography; h, the index of refraction of the flowing medium (assumed to be equal to that of the vitreous, 1.336); and /3, the angle between the direction of V max and its projection on the plane defined by the two scattering directions. Since publication of the LDV system, 2 an instrumental modification has been made which enables the continuous rotation of this scattering plane so that (3 can be set equal to zero (cos /9 = 1) for each measured vessel. K is an instrumental constant related to the geometry of the optical pathways used in the detection system. For our instrument, K is equal to Refraction of the incident and scattered beams at the vessel wall was disregarded. For each vessel measured, the two photocurrent signals were recorded on a Honeywell 5600C magnetic tape system for 5 sec at 5-sec intervals for a total period of 3-5 min. Between each 5-sec recording, the fundus camera was recentered; the position of the incident laser beam and light-collecting optical fibers remained unchanged. The level of laser light illumination was approximately 0.08 Watt/cm 2, a value below the maximum permissible level of retinal irradiance for an illumination of 5-sec duration. 3 Throughout the recording session, the fundus was diffusely illuminated with monochromatic light (wavelength of 0.57 /im) at a constant intensity. This wavelength was chosen because it enhances the contrast between vessels and background 4 while it can be filtered out by the laser light transmitting optical filter placed in front of each photomultiplier. 2 The LDV recordings were made at sites which provided the most distinct cutoff frequencies when the laser beam was moved along the vessel. Sites close to arterial branchings, venous junctions or arteriovenous crossings were avoided. All locations of measurement were within one to two disc diameters from the center of the disc and were marked on Polaroid color fundus photographs for referencing diameter measurements. Processing the Doppler Signal The taped photocurrent signals were analyzed by a two-channel spectrum analyzer (Unigon, model 4520). One of the signals was also supplied to a loudspeaker. Only those portions of tape were analyzed from which a clearly pulsatile pitch (for arteries) or a monotonous, high frequency pitch (for veins) could be heard. The power spectra were displayed in pairs on the screen of a two-channel oscilloscope. Each spectrum was inspected for the presence of a distinct drop of both the spectral power and the magnitude of the spectral power fluctuations. If such a drop could be identified on both spectra, the frequencies at which it occurred were documented as cutoff frequencies fi and fj. The examiner was masked with regard to the location and diameter of the vessels. In arteries, V max has been found to be pulsatile, in synchrony with the heartbeat. Previously, we have recorded the power spectra for 0.1 sec at various phases of the cardiac cycle 5 and calculated a mean value of V max, (V max ) by integrating over the cardiac cycle. This procedure was abandoned because it was time consuming. Instead, we have calculated (V max ) using an empirical formula /v \ = v J + - rv v.. 1 IT\ \ T max/ * max.dia ' o I max.sys * max.diaj \^J which is derived from the relationship for the mean blood pressure 6 MBP = BP dia + - (BP sys - BP dia ). This formula was found to provide values of (V max ) in good agreement with those obtained by integrating V max over the cardiac cycle. Equation (2) can be rewritten as max.dia (3) where p, the pulsatility of the RBC velocity is defined as: _ * max.sys _ * 1,sys V ~ f v max.dia M,dia Vmax.dia was determined using equation (1) by averaging values of Af obtained from nine pairs of power spectra (Fig. 1), each taken during diastole at different cardiac cycles, p was calculated by averaging five values of the ratio fi, sys /fi,dia- Recording time of each pair of spectra was 0.1 sec. Pulsatile velocity was rarely observed in the veins measured. Therefore, for these vessels, the power spectra were recorded without regard to the phase of the heart cycle. (V max ) was calculated by averaging nine values of Af obtained from pairs of power spectra recorded during different cardiac cycles in 0.32 or 0.64 sec. Figure 2 displays a pair of power spectra obtained from a retinal vein. The vessel diameter, D, at the site of the LDV recordings was estimated from photographic negatives taken with an interference filter centered at 0.57 /im. The photographic negatives were projected onto a screen and the diameter of the vessels were measured with a caliper. Total magnification of the fundus'

3 1126 INVESTIGATIVE OPHTHALMOLOGY 6 VISUAL SCIENCE / August 1985 Vol. 26 oc UJ o Q. Equation (4) assumes a circular cross-section of the vessels and equation (5) uses a constant of proportionality of 1.6 between centerline red cell velocity and mean blood velocity. This relation has been demonstrated for blood flowing in glass tubes with a diameter in the range of those measured in our study (Baker and Wayland, 7 Lipowsky and Zweifach, 8 Damon and Duling 9 ). (V max ), Q, and D were obtained from retinal arteries and veins in 12 eyes of seven healthy volunteers, all males, ranging in age from 20 to 45 yr (mean, 34 yr). These subjects had normal intraocular pressure, no fundus pathology, and steady target fixation. Axial length ranged from 22.6 to 24.4 mm (mean, 23.5 mm). Informed consent was obtained from all subjects prior to undertaking the study. OC UJ o Q. oc UJ O FREQUENCY I KHZJ 20 Fig. 1. A pair of Doppler shift power spectra of laser light scattered in two directions from red blood cells in a retinal artery (78 ixm in diameter). Both spectra were recorded simultaneously during diastole. Time window: 0.1 sec. Frequency range: 0-20 khz. The cutoff frequencies f, and f 2 were determined by visual inspection of the spectra (arrows). projected image was XI67. This includes the magnification of the eye-fundus camera system, for which we assumed a value of X2.5 for all subjects. All vessel diameters were measured by the same individual and were obtained from an average of three measurements. The average coefficient of variation of D was 3%. Volumetric blood flow rate, Q, was estimated as Vc A * mean (4) with the mean blood velocity, V mean, calculated as V r (5) oc UJ o Q. FREQUENCY CKHZ3 Fig. 2. A pair of Doppler shift power spectra obtained from a retinal vein (92 /xm in diameter). Time window: 0.32 sec. Frequency range: 0-20 khz. Arrows indicate fi and f 2, the cutoff frequencies determined by visual inspection of the spectra. 20

4 No. HUMAN RETINAL BLO FLOW / Rivo er ol Arterial 3.5 Venous Fig. 3. (V max ) vs vessel diameter based on measurements in 64 retinal arteries and 71 retinal veins. Correlation coefficient, r, for the linear fit between (V max ) and D is 0.52 (P< 0.001) for the arteries and 0.66 (P < 0.001) for the veins Vessel Diameter (pm) Results Velocity azjd Volumetric Flow Rate Figures 3 ahd 4 demonstrate the dependence of (V^x) and Q upon the diameter of the arteries and veins which, at the site of the LDV measurements, ranged from 39 to 134 nm and from 64 to 177 /jm, respectively. For any given vessel diameter, (V max ) and Q were generally greater for arteries than for veins. A significant correlation was found between (V max ) and D for arteries (linear correlation coefficient r = 0.52, P < 0.001) and for veins (r = 0.66, P< 0.001), demonstrating that flow velocities increased with increasing diameter. Such a correlation was evident for seven arteries measured in one eye (r = 0.91, P < 0.01). The correlation coefficient between Q and D based on a power curve fit was 0.89 (P < 0.001) for the arteries and 0.92 (P < 0.001) for the veins. Volumetric flow rate was found to vary with the diameter at a power of 2.76 ±0.16 for the arteries and at a power of 2.84 ±0.12 for the veins. A significant correlation <Q > (pl/min) Arterial "* Venous Fig. 4. Volumetric flow rate vs vessel diameter, D, for 64 arteries and 66 veins. Continuous lines are the power curve fit to the data points from arteries and veins. Q A varies as " 5 D 276 and Q v as " 6 D 284 where D is in pm. Correlation coefficient, r, is 0.89 (P < ) for arteries and 0.92 (P < ) for veins. \. V; A- -15 ' / io. /:.. -5 / V Vessel Diameter (pm)

5 1128 INVESTIGATIVE OPHTHALMOLOGY 6 VISUAL SCIENCE / Augusr 1985 Vol ~ 1.0 > E Time (minutes) was calculated for three ranges of diameters, and was 26%, 17% and 16% for veins with diameters between 62 and 99 /im, 100 and 139 ixm, and 140 and 176 nm, respectively. For arteries, this coefficient was constant (34%) for the three ranges of diameters (40-69 tim, /im, and /mi)- The average coefficient of Q was only a few percent greater than that of <(V max ) since the coefficient of variation of D was only 3%. 2.0 Flow Measurements Along a Vessel I Time (days) Fig. 5. A, top, Retinal blood velocity in a 119-Mm vein of a normal subject measured during approximately 50 min. Error bars represent ±1 SD. B, bottom, Retinal blood flow velocity in the same vein measured over a period of 2 mo. between Q and D was also apparent for the arteries and the veins measured from the same eye. In all arteries, pulsatile flow velocity was observed to be in phase with the cardiac cycle. The pulsatility, p, ranged from 2.26 to For one eye in which the diameter of seven arteries ranged from 39 to 105 ixm, p was found to correlate significantly with D (linear correlation coefficient r = 0.77, P < 0.05). p decreased by 0.1 as D decreased by 10 fim. Venous pulsatility was seldom perceptible in the Doppler signal; it could be measured in only two of the 71 veins and was 1.68 and Reproducibility of LDV Measurements Intraobserver: LDV recordings from 19 vessels of six eyes were analyzed twice by the same observer in a masked manner. Nine pairs of spectra were recorded to determine Af, identifying f, and f 2 by visual inspection of the power spectra. The average absolute difference between both determinations, Afi - Af 2, was 10.7% of the average Af calculated as (Af, + Af 2 )/2. Interobserver: Recordings from 52 veins were analyzed independently by two observers. Average value of Af obs! Af obs2 was 13.8% of the average Af. Variability of (V max ) and Q The average coefficient of variation of (V max ), i0(m<v ma x)) <v max > 80 In two subjects, volumetric flow rate was measured in four veins at various sites between two junctions. Separation between the first and last sites was one and one-quarter disc diameter. The coefficient of variation of the mean flow rate ranged between 2 and 12%. The difference between flow rates obtained from any two sites along the same vessel ranged from 1 to 22% (mean 8.2%) of the average flow rate. Stability of Retinal Blood Flow Velocity The stability of V max was determined from repetitive BLDV measurements in veins of two subjects. The power spectra were analyzed by the same observer. Figure 5 shows V max values obtained from a 119-^m vein in one subject over a period of 50 min and on different days over a period of 2 mo. Each V max was calculated from 1 to 1.5 min of recording (approximately 20 pairs of power spectra). The coefficient of variation of V max was less than 10% in both cases. Similar results were obtained from a 127-fim vein of the other subject. Blood flow velocity, when determined with a time constant of 1 to 1.5 minutes, appears, therefore, to be remarkably constant in normal subjects since the variations in V max are within the range of errors of the technique. Furthermore, measurements performed twice, on different days and under the same experimental conditions, showed average coefficient of variation for V max of 10.6 ± 8.5% in 11 veins with a diameter between 64 and 172 ^m. Total Retinal Blood Flow In order to estimate total retinal arterial and venous volumetric flow rate, the flows for the individual arteries or veins were summed. In many eyes, however, not all vessels were measurable by LDV as shown by the value of %S A and %S V in Table 1. %S A or %S V represents the sum of the cross-sections of the arteries or veins which could be measured by LDV, expressed as a percent of the sum of the cross-sections of all

6 No. 8 HUMAN RETINAL DLO FLOW / Rivo er ol Table 1. Total arterial and venous retinal blood flow Subject Eye %s A %S V QA (nl/min) QACOIT (lil/min) Qv (til/min) Qvcorr (nl/min) Qvcorr/QAcorr ± 2.8* 35.1 ± ± 1.7* 34.1 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Mean 30 ± ± ± ± ± 0.16 * ±1 SD (calculated from the standard deviations of V max, D and p). %S A (%S V ): Sum of cross section of all arteries (veins) from which LDV measurements could be obtained, expressed as percentage of the cross section of all visible arteries (veins), S A (S v ). QA (QV): Sum of the arterial (venous) volumetric flow rates measured. QAcorr (Qvcorr): Corrected value of Q A (Qv) based on an estimation of volumetric flow rates in the vessels not measured by LDV. arteries or veins measurable on fundus photography. For example, a %S A of 98 means that only a very small artery could not be measured by LDV. Therefore, to include all the vessels in the calculation of total flow, we have estimated the volumetric flow rates of the vessels which were not measured based on the regression lines in Figures 4 and 5 and have added these values to the sum of the measured flows, Q A and Q v, to obtain the total corrected volumetric flow rates QACOIT and Qvcorr (Table 1). A highly significant correlation was found between Q AC orr and Qvcorr (P < 0.001). The average ratio Qvcorr/QAcorr was equal to 1.06 ±.16 (SD). Figure 6 is a plot of Q Acor r and Qvcorr vs S A and S v respectively, demonstrating a significant correlation (P < 0.02) between total volumetric flow rate and total vessel cross-section Fig. 6. Total arterial ( ) and venous (O) retinal volumetric flow rate vs the corresponding total cross-section. Correlation coefficients for the linear regression are 0.86 (P < 0.01) for arteries and 0.76 (P < 0.01) for veins. o ^ 30 H 0) J 20 2> Total Vessel Cross-section (cm 2. io 5 )

7 1130 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / August 1985 Vol. 26 Table 2. Vessel cross-section and blood flow in temporal and nasal retina Subject 1 Eye Si S»A SI SV Ql Q N A Ql n N \lv Mean ± SD 22 ± ± ± ± ± ± ± ±4.3 Sl(S ), Sv(Sv): Total arterial (A) or venous (V) cross-section (cm 2 10~ 5 ) of the temporal (T) or nasal (N) vessels. QA(QA), QV(QV): Corrected total arterial or venous flow rates (/xl/min) in the temporal or nasal retina. Correlation coefficients for linear regression of Qj vs Sj: r = 0.67, P < 0.02; Qj vs SJ r = 0.74, P < 0.01; Q VS Si? r = 0.82 P < 0.01; Q v vs S v r = 0.85 P < Regional Blood Velocity and Volumetric Flow Rate No difference was observed in the average velocities within arteries and veins of equal diameters when nasal and temporal regions of the fundus were compared. V maxj for nasal and temporal arteries between 80 and 100 /tm in diameter, for example, was 2.49 ± 0.22 (SE) cm/sec and 2.47 ± 0.25 (SE) respectively. For those between 100 and 140 /im in diameter, the nasal and temporal (V max ) were 2.51 ±0.13 (SE) and 2.47 ±0.16 (SE). The average velocity in superior and inferior vessels of similar diameter were also not statistically different. Table 2 shows the total crosssections of the temporal and nasal vessels with the corresponding corrected flow rates. The linear correlations between the arterial (venous) flows and arterial (venous) cross-sections are all significant. A significant difference was found (Table 2) between the average volumetric flow rates of the nasal and temporal retina (P < 0.02 for arteries and P < for veins) as well as between the nasal and temporal cross-sections (P < 0.05 for arteries and P < for veins). Average temporal flow rate and crosssection (arterial and venous) were 52% and 45% greater than average nasal flow rate and cross-section, respectively. There was however no significant difference in volumetric flow rates and cross-sections between the superior and inferior retinal hemispheres. Discussion This investigation attempts, for the first time, to determine retinal blood velocity and volumetric flow rate as a function of vessel diameter. It also provides the first measurement of total retinal volumetric flow rate in normal subjects and demonstrates that this parameter correlates with the total cross-section of the retinal vessels. The measured flow rates represent, obviously, an estimation of the true flow rates. Inaccuracy in the measurements can be expected from errors in the determination of the diameter and the velocity. An uncertainty is associated with the cross-sectional area estimates which assume a unique optical magnification of the eye-fundus camera system and which do not take into account the marginal plasma zone (2-4 /zm for the vessels measured in this study 10 ) nor the actual shape of the vessels. The use of a model eye and neglect of the refraction of the laser light at the vessel wall may also be a source of error when defining the actual scattering geometry. Furthermore, the assumption of a unique and invariant relationship between V max and V mean is based on in vitro measurements and may not apply in vivo." Our velocity and flow values from main superior retinal arteries are, however, in close agreement with the values of 2-4 cm/sec and 7 /il/min obtained from such arteries by high speed cinematography after carotid artery injection of fluorescein. 12 As observed in other vascular beds, 1314 the velocity in retinal vessels increases with the diameter. This increase is steeper for the arteries than for the veins, most likely due to the steeper blood pressure gradient in arteries. The velocity also varies greatly between vessels of approximately the same diameter. This

8 No. 8 HUMAN RETINAL DLO FLOW / Riva er ol variability results from instrumental uncertainties (variance <T 2 DV) anc * ^rom statistical fluctuations of the velocity (variance o- 2 tat) in different vessels of similar size. Assuming these two types of fluctuations to be uncorrelated (as we believe to be the case with our measurements), it is possible to determine the relative contribution of OLDV ar t0»d oftat ff m, the variance of the measured velocities, using the formula: ffm = OIIDV +ffstat- 15 Calculations for the three ranges of arterial and venous diameters specified earlier (Results, variability of (V max ) and Q) show that o- 2 tat represents 60-70% of oi, in arteries and 57-72% of a 2 m in veins, whereas (TLDV represents approximately 30-40% of the measured variability of (V max ). Presently, LDV measurements from veins are more reliable and less time consuming than measurements from arteries. Variability of arterial velocities is increased by two factors. First, the random motion of the laser beam across arteries produces larger variations of Af than for veins because the arteries are generally smaller than the veins. Second, the variability of the pulsatility measurements must be added to that of V maxdia. The average total volumetric flow rate of approximately 34 pil/min is comparable to previously reported values obtained from eyes of the macaque monkey by means of the microsphere technique: 25 ± 9 (SD) ^l/min 16 and 34 ± 8 (SD) /d/min. 17 Values published for the eye of the rhesus monkey are somewhat larger (50 ± 39 (SD) /il/min) 18 but the standard deviation is to large to allow meaningful comparison. The intersubject variability of the total arterial and venous volumetric flow rates, which range from approximately /Ltl/min and jil/min, appears to be large. It is, however, similar to the variability observed in eyes of monkeys. It may be explained by the large variability of the total arterial and venous cross-section with which it correlates significantly (P < 0.01) (Fig. 6). The difference between temporal and nasal flow rates (5-9 /til/min on the average) (Table 2) correlates significantly with the difference between temporal and nasal cross-sections. Some of these differences are explainable by the 20-25% larger size of the temporal retina. The rest could be due to a higher metabolic activity in the temporal retina. The exponent of 2.76 ±0.16 and 2.84 ±0.13 for the variation of the volumetric flow rate with the arterial and venous diameter respectively conforms closely to Murray's law. 19 This law predicts an exponent of 3 for a vascular system which either seeks an optimum compromise between blood volume and vascular resistance or which minimizes its resistance for a given volume. This observation may lead to interesting speculations on the properties of the retinal vascular system which, however, are beyond the scope of the present work. 20 Bidirectional laser Doppler velocimetry combined with monochromatic fundus photography provides a reliable and reproducible technique for retinal blood flow measurement. The measurements of V max and D and the assumptions of a unique and invariant relation between V mean and V max and circular vessel cross-section yield arterial and venous volumetric flow rates which satisfy mass conservation. Increased accuracy in the determination of the volumetric flow rates could be gained from more accurate measurements of V max, by recording the LDV signals only when the laser beam is perfectly centered on the vessel, and possibly of D, as described recently. 21 Clinical application of the technique will be facilitated by shortening the time of recording of the power spectra (presently acquisition of the data to obtain total retinal blood flow requires about 1 hr) and implementation of online computer determination of the cutoff frequencies. 22 Key words: retinal blood flow, red blood cell velocity, laser Doppler velocimetry, retinal hemodynamics, Murray's law Acknowledgments The authors thank Joan Baine for her skillful technical assistance and Virginia Mesibov for typing the manuscript. References 1. Riva CE and Feke GT: Laser Doppler Velocimetry in the measurement of retinal blood flow. In The Biomedical Laser: Technology and Clinical Applications, Goldman L, editor. New York, Springer-Verlag, 1981, pp Riva CE, Grunwald JE, Sinclair SH, and O'Keefe K: Fundus camera based retinal LDV. Appl Optics 20:117, Sliney DH and Freasier BD: Evaluation of optical radiation hazards. Appl Optics 12:1, Delori FC and Gragoudas ES: Examination of the ocular fundus with monochromatic light. Ann Ophthalmol 8:703, Feke GT and Riva CE: Laser Doppler measurements of blood velocity in human retinal vessels. JA 68:526, Folkow B and Neil E: In Circulation. New York, Oxford University Press, 1971, p Baker M and Wayland H: On-line volume flow rate and velocity profile measurement for blood in microvessels. Microvase Res 7:131, Lipowsky HH and Zweifach BW: Application to the "two-slit" photometric technique to the measurement of microvascular volumetric flow rates. Microvasc Res 15:93, Damon DN and Duling BR: A comparison between mean blood velocities and center-line red cell velocities as measured with a mechanical image streaking velocimeter. Microvasc Res 17:330, 1979.

9 1132 INVESTIGATIVE OPHTHALMOLOGY 6 VISUAL SCIENCE / August 1985 Vol Charm SE and Kurland GS: In Blood Row and Microcirculation. New York, John Wiley and Sons, 1974, p Proctor KG, Damon DN, and Duling BR: Inaccuracies in blood flow estimates in microvessels during arteriolar vasoconstriction. Microvasc Res 28:23, Dollery CT: Dynamic aspects of the retinal microcirculation. Arch Ophthalmol 79:536, Zweifach BW and Lipowsky HH: Quantitative studies of microcirculatory structure and function. III. Microvascular hemodynamics of cat mesentery and rabbit omentum. Circ Res 41:380, Le-Cong P and Zweifach BW: In vivo and in vitro velocity measurements in microvasculature with a laser. Microvasc Res 17:131, Bevington PA: Data Reduction and Error Analysis for the Physical Sciences. New York, McGraw-Hill, 1969, p Aim A and Bill A: Ocular and optic nerve blood flow at normal and increased intraocular pressure in monkeys (Macaca irus): a study with radioactively labelled micropheres including flow determinations in brain and some other tissues. Exp Eye Res 15:15, Aim A, Bill A, and Young FA: The effect of pilocarpine and neostigmine on the blood flow through the anterior uvea in monkeys: A study with radioactively labelled microspheres. Exp Eye Res 15:31, O'Day DM, Fish MB, Aronson SB, Pollycove M, and Coon A: Ocular blood flow measured by nuclide labeled microspheres. Arch Ophthalmol 86:205, Murray CD: The physiological principle of minimum work. I. The vascular system and the cost of blood volume. Proc Natl Acad Sci USA 12:207, Sherman TF: On connecting large vessels to small: the meaning of Murray's law. J Gen Physiol 78:431, Kennedy SJ, Schwartz B, Takamoto T, and Eu KT: Interference fringe scale for absolute ocular fundus measurement. Invest Ophthalmol Vis Sci 24:169, Petrig BL, Riva CE, and Grunwald JE: Computer analysis of laser Doppler measurements in retinal blood vessels. ARVO Abstracts. Invest Ophthalmol Vis Sci 25(Suppl):7, 1984.