Variability and repeatability of retinal blood flow measurements using the Canon laser blood flowmeter☆

Available online at www.sciencedirect.com R

Microvascular Research 65 (2003) 145–151

www.elsevier.com/locate/ymvre

Variability and repeatability of retinal blood flow measurements using the Canon Laser Blood Flowmeter夞 Kit Guan,a Chris Hudson,a,b,* and John G. Flanagana,b a

Department of Ophthalmology, University of Toronto, Toronto, Ontario M5T 2S8, Canada b School of Optometry, University of Waterloo, Waterloo, Ontario N2L 3G1, Canada Received 1 November 2002

Abstract The purpose of this work was to determine the within-session variability and between-session repeatability of the Canon Laser Blood Flowmeter (CLBF), Model 100, an instrument that permits the noninvasive measurement of retinal blood flow. The CLBF calculates flow in ␮l/min based on the Poiseuille principle. One eye of 20 normal subjects (mean age 36.5; SD 9.7 years) was randomly chosen. A minimum of five measurements was acquired of a temporal arteriole approximately 1 disc diameter from the optic nerve. Measurements were repeated within a 1-month period. Blood pressure and intraocular pressure were measured. The coefficient of variation (COV) and the coefficient of repeatability (COR) were calculated for each individual. The individual COVs for diameter, velocity, and flow ranged from 0.5 to 6.5% (median 2.0%), 4.8 to 39.7% (median 19.9%), and 4.8 to 37.3% (median 19.3%), respectively. The group mean CORs for diameter, velocity, and flow were 5.2 ␮m (relative to a mean effect of 104.6 ␮m), 8.8 mm/s (relative to a mean effect of 33.9 mm/s), and 2.6 ␮l/min (relative to a mean effect of 8.8 ␮l/min), respectively. The CLBF gave consistent and repeatable measurements of blood flow within retinal arterioles in normal subjects. Given the range of individual variation in the velocity measurement, and thus flow, confidence limits for retinal hemodynamics need to be determined on an individual basis. © 2003 Elsevier Science (USA). All rights reserved.

Introduction The retinal vasculature is distributed primarily within the inner retina and is thought to nourish most retinal structures apart from the pigment epithelium and photoreceptors, which are supplied by the choroidal vasculature. It is unique in that the vessels lack an intrinsic sympathetic nerve supply and lumen integrity is maintained despite the presence of a high external tissue pressure. Blood flow into the capillary bed of the retina is determined by the contraction of the upstream arterioles whose smooth muscle cells are highly sensitive to intrinsically (i.e., emanating primarily from the endothelium)

夞 Aspects of this work were presented at the 2002 annual meetings of the Association for Research in Vision and Ophthalmology and of the American Diabetes Association. The authors have no proprietary interest in the Canon Laser Blood Flowmeter. * Corresponding author. Fax: ⫹1-416-603-5126. E-mail address: [email protected] (C. Hudson).

generated vasodilators and vasoconstrictors (Haefliger et al., 2001). In addition, the downstream retinal capillaries are thought to have contractile and vasoactive properties and, consequently, are able to further regulate blood flow in response to local tissue demands, such as the level of oxygenation (Alm, 1992). These mechanisms in combination maintain constant blood flow over a wide range of perfusion pressures (i.e., autoregulation). Disturbance of retinal capillary blood flow is a feature of many ocular diseases, including diabetic retinopathy, agerelated maculopathy and a subset of patients with glaucoma. Blood flow perturbation in the retina has been suggested to be a precursor of clinically visible retinopathy in diabetic patients and is accepted as a surrogate marker of early diabetic complications (Archer, 1999). In diabetes, vasodilation is proposed to occur prior to the onset of clinically evident retinopathy. An increase in retinal blood flow is thought to eventually lead to the development of diabetic retinopathic complications (Schmetterer and Wolzt, 1999). In addition, disturbances of optic nerve blood flow have

0026-2862/03/$ – see front matter © 2003 Elsevier Science (USA). All rights reserved. doi:10.1016/S0026-2862(03)00007-4

146

K. Guan et al. / Microvascular Research 65 (2003) 145–151

been implicated in the development and progression of low-tension glaucoma (Hayreh, 2001). Numerous noninvasive techniques have been developed to assess ocular blood flow in humans. One such technique, bi-directional laser Doppler velocimetry, provides a quantitative, pointwise measurement of retinal blood velocity in the major arterioles and venules of the retina. This technique has been used to assess blood velocity in a variety of studies in clinically normal subjects and in patients with ocular disease (Konno et al., 1996; Yoshida et al., 1998; Fujio et al., 1994; Ogasawara et al., 1992; Nagaoka et al., 2002). A new commercial instrument, the Canon Laser Blood Flowmeter (CLBF) Model 100 has recently been released. The CLBF determines centerline blood velocity (mm/s) and vessel diameter (␮m) of the retinal arterioles and venules. It subsequently calculates flow in ␮l/min based on the Poiseuille principle. The assessment of retinal blood flow with the CLBF may provide a clinical “tool” to predict those patients at risk of developing sight-threatening eye disease (Feke et al., 1985). At this moment in time, however, no consensus has been reached concerning the precise clinical significance of change in retinal capillary blood flow. Indeed, the variability and repeatability of retinal capillary blood flow measurement needs to be established in order to define significant change and to use this technique in a clinical setting. The aim of the study was to determine the within-session variability and between-session repeatability of the CLBF.

Materials and methods Sample The sample consisted of 20 eyes from 20 healthy volunteers (8 men, 12 women; mean age 36.5, SD 9.7, range 21–57 years). All subjects had a visual acuity of 20/40 or better. Subjects were excluded for any eye disease, refractive error greater than ⫾ 6.00 DS or ⫾ 2.50 DC, glaucoma or diabetes in a first-degree relative, history of central nervous system disorders or psychiatric illness, or medications with known effects on blood flow (e.g., anti-convulsants, muscle relaxants, or anti-inflammatory medications). The study was approved by the Research Ethics Board of the University Health Network, University of Toronto, Canada. The research followed the tenets set out in the Declaration of Helsinki. Informed consent was obtained from all subjects after the nature and possible consequences of the study were explained. Canon Laser Blood Flowmeter The principle underlying the CLBF is based on the Doppler effect. Briefly, the frequency of reflected laser light is shifted by a moving particle. This frequency shift (⌬f ) is

proportional to the velocity of the moving particle. A vessel that exhibits Poiseuille flow will have a range of velocities and thus a range of frequency shifts up to a maximum frequency shift ⌬fmax that corresponds to the maximum velocity of blood moving at the centre of the vessel. Light scattered from stationary tissue is unshifted and acts as the reference beam from which a relative change in retinal blood flow speeds can be measured (Feke and Riva, 1978). By utilising two photomultipliers separated by a known angle, the frequency shifts are subtracted to remove the uncertainty of scattering angles. This allows the absolute noninvasive measurement of centreline blood velocity in the human retinal vasculature (Riva et al., 1979; Feke et al., 1987). The CLBF automatically determines the maximum Doppler shifted frequency from the two photomultipliers and converts this to maximum velocity. A red diode measurement laser (675 nm, 80 ␮m ⫻ 50 ␮m oval) is used to measure up to a maximum velocity of 120 mm/s. The measurement window of 2 s gives continuous velocity readings (every 0.02 s) and plots a velocity time curve. Calibration of the CLBF to measure blood velocity and vessel diameter is done during assembly. Once installed, periodic adjustments of laser power and detector sensitivity and daily optical system checks are undertaken to maintain the calibration of the instrument. The CLBF also uses a green diode vessel tracking laser system (543 nm, 1500 ␮m ⫻ 150 ␮m rectangle) that is used to stabilize, and measure the diameter of, the vessel of interest (Fig. 1) (Milbocker et al., 1991; Delori et al., 1988). The vessel tracking system allows a graph of eye position to be superimposed on the velocity time curve to aid in artifact rejection. Multiple diameter readings are acquired during the first and final 60 ms of the 2-s velocity measurement window every 0.04 ms. Two sequential bi-directional readings (i.e., path 1 and path 2) are taken to ensure consistency and averaged to give one reading. In combination with the average velocity over a pulse cycle and the diameter, flow through the vessel can be calculated as F ⫽ Vmax S/2, where S is the cross-sectional area of the vessel at the measurement site (for technical summary see Kida et al., 2002; Canon, 1997). Procedures One eye of each subject was randomly assigned to the study. Refraction, logMAR visual acuity, and resting blood pressure were assessed prior to dilation of the study eye with 1% tropicamide. All subjects were seated for a minimum period of 15 min prior to blood flow measurements in order to facilitate a physiological cardiovascular and respiratory resting state. Five retinal blood flow readings were taken using the CLBF Model 100 (Canon, Tokyo, Japan) of a temporal arteriole approximately 1 disc diameter from the edge of the optic nerve head. Intraocular pressure was measured by Goldmann applanation tonometry after readings were taken with the CLBF. Axial length was measured by I3

K. Guan et al. / Microvascular Research 65 (2003) 145–151

147

Fig. 1. (a) The CLBF showing the acquisition camera, power module, and personal computer. (b) A CLBF fundus image during data acquisition showing the red measurement laser (centered on the vessel) superimposed on the green tracking, and diameter measurement, laser.

System ABD A-scan ultrasound (I3 Innovative Imaging Inc, Sacramento, CA) to correct the CLBF measurements for magnification effects. The CLBF readings were repeated within 1 month of the initial visit at the same time of day and under the same conditions as the first session. A single operator (K.G.) acquired all the data.

Analysis A postacquisition analysis of the velocity waveforms was performed using a standardized protocol to remove aberrant waveforms affected by eye movement, tear film breakup, or improper tracking of the measurement laser.

148

K. Guan et al. / Microvascular Research 65 (2003) 145–151

Fig. 2. Graph of CLBF-derived velocity time curve over a 2-s measurement window. (Left) Aberrant waveform. (Right) Superimposed eye tracking errors that caused the aberrant waveform. This data was excluded from the analysis.

Fig. 2 illustrates aspects of the artifact rejection protocol. The maximum number of acceptable pulse cycles was used in the data analysis for each measurement (Fig. 3). The coefficient of variation (COV) and coefficient of repeatability (COR) were calculated to assess the within-session variability and between-session repeatability, respectively, of the CLBF (Bland and Altman, 1986).

Results Subject characteristics (including age, gender, and right or left eye), axial length, intraocular pressure, and blood pressure are detailed in Table 1. A box plot of the COVs for arteriolar diameter, centreline blood velocity, and blood flow are displayed in Fig. 4. The individual COVs for diameter, velocity, and flow ranged from 0.5 to 6.5% (median 2.0%), 4.8 –39.7% (median 19.9%), and 4.8 to 37.3% (median 19.3%), respectively. The individual CORs for diameter, velocity, and flow ranged from 2.2 to 18.4 ␮m (median 5.4 ␮m), 8.7–28.4 mm/s (median 18.3 mm/s), and 2.0 –9.2 ␮l/min (median 4.4 ␮l/min), respectively. The group mean COR for diameter, velocity, and flow was 5.2 ␮m (mean effect 104.6 ␮m), 8.8 mm/s, (mean effect 33.9 mm/s), and 2.6 ␮l/min (mean effect 8.8 ␮l/min), respectively. A difference vs mean plot of diameter, velocity, and flow (Fig. 5) reveals clear outliers in each of the parameters (subjects 16 and 18) (Bland and

Altman, 1986). Removal of these outliers reduced the group mean COR for diameter, velocity, and flow to 3.3 ␮m (mean effect 104.1 ␮m), 6.8 mm/s (mean effect 33.1 mm/s), and 2.0 ␮l/min (mean effect 8.6 ␮l/min), respectively (Table 2).

Discussion This study shows that the CLBF can reliably and consistently measure blood flow in the major arterioles of the retina of normal human subjects. Compared to other methods of assessing retinal blood flow, the CLBF offers a quantifiable and repeatable method of assessing retinal hemodynamics. Blue field entoptic and fluorescein angiographic techniques provide mostly subjective or semiquantitative assessments of retinal blood flow. Laser Doppler flowmetry provides flow values only in arbitrary units and readings between subjects are open to interpretation. Pulsatile ocular blood flow assessments rely on controversial assumptions of ocular physiology and measure total blood flow, a majority of which arises from the choroidal circulation. In this study, the individual variability for arteriolar diameter was small, i.e., a median COV of 2.0% or 2.1 ␮m relative to a mean effect of 104.5 ␮m. There was a range of variability in individual velocities with a median COV of 19.9% or 6.7 mm/s relative to a mean effect of 33.9 mm/s. The individual variability in velocity may reflect the impact of eye motion, tear film quality, noncentreline blood flow

Fig. 3. Two different velocity time curves with different pulse rates. The maximum number of acceptable complete CLBF-derived pulse cycles was included in the analysis. The velocity was averaged between the two horizontal bars. Left; One complete pulse cycle. (Right) Two complete pulse cycles with averaging of velocity over a larger measurement window.

K. Guan et al. / Microvascular Research 65 (2003) 145–151

149

Table 1 Individual subject characteristics Name

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Age

34 28 28 32 35 43 23 39 57 25 34 45 38 44 42 21 41 47 25 49

M/F

M F F F F M M M F M F F M F F F M F M F

OD/OS

OD OD OS OS OD OS OS OS OS OS OS OD OD OS OD OS OS OD OS OD

AL

24.00 23.96 22.96 24.99 24.75 25.54 24.36 24.61 23.32 26.67 24.90 22.60 25.03 26.73 21.56 25.32 23.92 26.87 24.16 26.44

IOP

8 16 12 16 14 17 12 18 18 17 18 20 10 17 18 12 17 13 13 15

Visit 1

Visit 2

Sys

Dias

HR

Sys

Dias

HR

112 114 93 102 119 110 104 121 132 126 98 97 87 144 119 97 123 118 109 113

61 75 58 52 70 72 65 77 80 84 63 45 54 84 83 60 75 82 75 62

75 84 74 72 76 89 66 57 90 53 67 78 66 70 71 70 91 88 58 75

110 121 107 97 116 130 113 125 128 128 108 94 92 133 120 94 114 125 118 114

60 79 64 56 60 70 69 80 78 74 78 54 56 81 84 54 70 84 75 65

76 80 67 73 81 91 84 58 93 61 82 68 74 68 63 68 84 75 63 68

Note. AL, axial length (mm); IOP, intraocular pressure (mmHg); Sys, systolic blood pressure (mmHg); Dias, diastolic blood pressure (mmHg); HR, heart rate per minute.

measurements, or normal biological variation on blood flow speed. Flow is calculated based on the Poiseuille principle and varies directly with velocity and to the square of diameter. As a result, the variability in flow is compounded by the variability in these two measured parameters, the most variable of which is undoubtedly velocity. It could be predicted that the largest source of variation would arise from

Fig. 4. Box plot of individual COVs for diameter, velocity, and flow. The error bars show the non-outlier range (⫾1.5 times the height of the box).

measuring noncentreline blood velocities. Previous studies using a prototype of the CLBF reported COVs for velocity and flow of the order of 18% for both parameters in normal human subjects (Feke et al., 1989; Yoshida et al., 1996). More recently, the commercial CLBF has been used in the evaluation of retinal arteriolar blood flow changes following acute increases in blood pressure and glucose tolerance testing (Nagaoka et al., 2002; Kida et al., 2002). These papers reported group mean COVs for diameter, velocity, and flow ranging from 2.1 to 4.1%, 12 to 13.9%, and 10.8 to 16.9%, respectively, generally in agreement with our findings. A previous study on the variability of the CLBF reported overall COVs for diameter, velocity, and flow of 6.5% (range 1.0 to 27.3%), 16.6% (range 1.6 to 45.0%), and 18.7% (range 0.0 to 66.4%), respectively (Jonescu-Cuypers et al., 2001). Our study also showed a range of individual variation similar to that of Jonescu-Cuypers and co-workers (2001). Given the range of individual variation, confidence limits for significant change should be set on an individual basis. Multiple readings of a given site at a given time are needed to determine the COV. With controlled conditions and an experienced operator, the COV can be kept to a minimum, thus giving the ability to measure smaller effect sizes. Patients with high COVs could be excluded from a given study; the acceptable level of variability needs to be determined for each study based upon the magnitude of the expected effect size. The CLBF has a repeatability for diameter, velocity, and flow of 3.3 ␮m (3%), 6.8 mm/s (21%), and 2.0 ␮l/min (23%), respectively. One other study has examined the repeatability of blood velocity measurements of the CLBF

150

K. Guan et al. / Microvascular Research 65 (2003) 145–151

Fig. 5. Difference in diameter (top), velocity (middle), and flow (bottom) as a function of the mean value between sessions. The outer bars represent the limits of agreement (COR) and the center bar represents the mean of the differences between sessions.

and reported it to be of the order of 18%, i.e., very similar to our findings (Feke et al., 1989). Preliminary data collected in our laboratory on the variation in adjacent areas of a nonbifurcating arteriole reveal no significant changes in velocity readings. In certain subjects with tapering or curved arterioles, small but significant differences in vessel diameter (on the order of 10 ␮m) can be detected. In combination, this

may result in a difference in flow between these two sites. This illustrates the importance of measuring from the same location if one is looking at hemodynamic changes prospectively. The results of this study demonstrate that the primary source of the variability in blood flow is derived from the velocity component. Various factors can influence the measurement of velocity, with the major factors being eye motion and centreline displacement. The eye tracking mechanism of the CLBF compensates for a majority of eye motion artifacts. Fig. 2 highlights the importance of examining the acquired velocity waveforms to ensure accuracy of tracking. The accuracy of the eye movement detector is determined by the contrast of the retinal vessel under consideration and noise sources within the detector (Milbocker et al., 1991). Accuracy is determined by the compensatory step size of the instrument and the magnitude of the eye movement. It has been previously reported in a prototype instrument that the tracking system had an overall recentering error of 7.2 ⫾ 2 ␮m (Milbocker et al., 1991). In Poiseuille flow, centreline velocity is the most rapid and decreases in a parabolic manner towards the vessel wall. Small displacements of the measurement laser will bias the velocity measurement to lower velocities and increase the measurement variability. Other factors such as tear film break up can distort the measurement laser or may cause it to drift away from the centre of the vessel due to optical blurring of the image. Empiric observations have demonstrated that inadequate dilation, upper lid obstruction, and media opacities can impair the usefulness of the instrument in certain situations. All these factors can contribute to the individual variation of blood flow parameters measured by the CLBF. In a clinical study employing fluorescein angiographic techniques, Bursell et al. (1996) found a 32% difference in retinal blood flow between patients with diabetes and normal subjects. They used glucose clamping methodologies to raise glucose levels from euglycemia to 200 and 300 mg/dl and found a corresponding increase in retinal blood flow of 32 and 20%, respectively. Grunwald et al. (1984) used laser Doppler velocimetry to assess the hemodynamic response to 100% oxygen breathing in patients with diabetes that had varying levels of retinopathy. They found a 24 –54% drop in retinal blood flow with 100% oxygen that was related to the level of retinopathy. Fujio et al. (1994) employed a developmental model of the CLBF in evaluating blood flow in retinal vessel occlusions and found a 40 –50% drop in blood flow in branch retinal artery occlusions and a 80 –90% drop in branch retinal vein occlusions. Given the magnitudes of blood flow responses in these various studies, the CLBF is an ideal instrument to measure hemodynamic disturbances in a variety of clinical situations. Since altered blood flow has been accepted as an early feature of various ocular diseases, particularly diabetic retinopathy, the ability to reliably establish change of blood flow within the retinal vasculature would be of great value. It is clear that any treatment to preserve the vision of patients with diabetes must be instituted before the distur-

K. Guan et al. / Microvascular Research 65 (2003) 145–151

151

Table 2 Individual and group mean COR with and without outliers (subjects 16 and 18) Individual COR

Diameter Velocity Flow

Median

Range

5.4 18.3 4.4

2.2–18.4 8.7–28.2 2.0–7.4

Overall COR

Mean effect

Outliers removed

Mean effect

5.2 8.8 2.6

104.6 33.9 8.8

3.3 6.8 2.0

104.1 33.1 8.6

bance of retinal capillary blood flow and the associated loss of retinal vascular integrity (Archer, 1999). The present study has demonstrated that the CLBF can give consistent and repeatable measurements of blood flow within retinal arterioles in normal subjects. The effects, however, of eye movements, tear film break up, and improper tracking of the measurement laser need to be controlled. In summary, the magnitude of measurement variability and repeatability can vary substantially between individuals; confidence limits for differences in retinal blood flow from normal, and for change in retinal blood flow over time, need to be determined on an individual basis. The CLBF offers the opportunity to quantifiably track retinal blood flow disturbances in various ocular diseases such as diabetic retinopathy. The instrument also has application for the investigation of the physiological response of the retinal vasculature to provocation, such as isometric exercise, intraocular pressure manipulation, and hyperoxia.

Acknowledgments This work was funded by the Canadian Institutes of Health Research (Operating Grant to C.H. and J.G.F. and New Investigator Award to C.H.), by a Premier’s Research Excellence Award (recipient C.H.), and by a Vision Science Research Program Fellowship, University of Toronto (recipient K.G.).

References Alm, A., 1992. Ocular circulation, in: Hart, W.M. (Ed.), Adler’s Physiology of the Eye, ninth edition, Mosby-Year Book Inc., St. Louis, MO, pp. 198 –227. Archer, D.B., 1999. Bowman Lecture 1998. Diabetic retinopathy: some cellular, molecular and therapeutic considerations. Eye 13, 497–523. Bland, J.M., Altman, D.G., 1986. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1, 307–310. Bursell, S.E., Clermont, A.C., Kinsley, B.T., Simonson, D.C., Aiello, L.M., Wolpert, H.A., 1996. Retinal blood flow changes in patients with insulin-dependent diabetes mellitus and no diabetic retinopathy. Invest. Ophthalmol. Visual Sci. 37, 886 – 897. Canon, 1997. Laser Blood Flowmeter CLBF Model 100 Operation Manual. Canon Inc., Tochigiken. Delori, F.C., Fitch, K.A., Feke, G.T., Deupree, D.M., Weiter, J.J., 1988. Evaluation of micrometric and microdensitometric methods for measuring the width of retinal vessel images on fundus photographs. Graefes Arch. Clin. Exp. Ophthalmol. 226, 393–399.

Feke, G.T., Goger, D.G., Tagawa, H., Delori, F.C., 1987. Laser Doppler technique for absolute measurement of blood speed in retinal vessels. IEEE Trans. Biomed. Eng. 34, 673– 680. Feke, G.T., Riva, C.E., 1978. Laser Doppler measurements of blood velocity in human retinal vessels. J. Opt. Soc. Am. 68, 526 –531. Feke, G.T., Tagawa, H., Deupree, D.M., Goger, D.G., Sebag, J., Weiter, J.J., 1989. Blood flow in the normal human retina. Invest. Ophthalmol. Visual Sci. 30, 58 – 65. Feke, G.T., Tagawa, H., Yoshida, A., Goger, D.G., Weiter, J.J., Buzney, S.M., McMeel, J.W., 1985. Retinal circulatory changes related to retinopathy progression in insulin-dependent diabetes mellitus. Ophthalmology 92, 1517–1522. Fujio, N., Feke, G.T., Ogasawara, H., Goger, D.G., Yoshida, A., McMeel, J.W., 1994. Quantitative circulatory measurements in branch retinal vessel occlusion. Eye 8, 324 –328. Grunwald, J.E., Riva, C.E., Brucker, A.J., Sinclair, S.H., Petrig, B.L., 1984. Altered retinal vascular response to 100% oxygen breathing in diabetes mellitus. Ophthalmology 9, 1447–1452. Haefliger, I.O., Flammer, J., Beny, J.L., Luscher, T.F., 2001. Endotheliumdependent vasoactive modulation in the ophthalmic circulation. Prog. Retin. Eye. Res. 20, 209 –225. Hayreh, S.S., 2001. The blood supply of the optic nerve head and the evaluation of it—myth and reality. Prog. Retin. Eye Res. 20, 563–593. Jonescu-Cuypers, C., Harris, A., Kagemann, L., Rotenstreich, Y., Siesky, B., Trivedi, A., McCranor, L., Sinchai, O., Martin, B., 2001. The new canon CLBF-100. variability of CLBF retinal artery measurements. Invest. Ophthalmol. Visual Sci. 42, s212. Kida, T., Harino, S., Sugiyama, T., Kitanishi, K., Iwahashi, Y., Ikeda, T., 2002. Change in retinal arterial blood flow in the contralateral eye of retinal vein occlusion during glucose tolerance test. Graefes Arch. Clin. Exp. Ophthalmol. 240, 342–347. Konno, S., Feke, G.T., Yoshida, A., Fujio, N., Goger, D.G., Buzney, S.M., 1996. Retinal blood flow changes in type I diabetes. A long-term follow-up study. Invest. Ophthalmol. Visual Sci. 37, 1140 –1148. Milbocker, M.T., Feke, G.T., Goger, D.G., 1991. Laser Doppler velocimetry stabilized in one dimension. IEEE Trans. Biomed. Eng. 38, 928–930. Nagaoka, T., Mori, F., Yoshida, A., 2002. Retinal artery response to acute systemic blood pressure increase during cold pressor test in humans. Invest. Ophthalmol. Visual Sci. 43, 1941–1945. Ogasawara, H., Feke, G.T., Yoshida, A., Milbocker, M.T., Weiter, J.J., McMeel, J.W., 1992. Retinal blood flow alterations associated with scleral buckling and encircling procedures. Br. J. Ophthalmol. 76, 275–279. Riva, C.E., Feke, G.T., Eberli, B., Benary, V., 1979. Bidirectional Ldv system for absolute measurement of blood speed in retinal-vessels. Appl. Opt. 18, 2301–2306. Schmetterer, L., Wolzt, M., 1999. Ocular blood flow and associated functional deviations in diabetic retinopathy. Diabetologia 42, 387– 405. Yoshida, A., Feke, G.T., Ogasawara, H., Goger, D.G., McMeel, J.W., 1996. Retinal hemodynamics in middle-aged normal subjects. Ophthalmic Res. 28, 343–350. Yoshida, A., Ogasawara, H., Fujio, N., Konno, S., Ishiko, S., Kitaya, N., Kagokawa, H., Nagaoka, T., Hirokawa, H., 1998. Comparison of short and long-term effects of betaxolol and timolol on human retinal circulation. Eye 12, 848 – 853.