Relative Flow Volume, a Novel Blood Flow Index in the Human Retina Derived From Laser Speckle Flowgraphy

Abstract

Purpose.: We investigated the accuracy and reproducibility of relative flow volume (RFV), a novel index of blood flow in the human retina derived from laser speckle flowgraphy (LSFG).

Methods.: Pre- and postbranch retinal RFV measurements were compared in 34 retinal venous bifurcations in 34 healthy volunteers (mean age, 49.0 ± 14.8 years) to determine the accuracy of RFV. Next, the coefficient of variation (COV) of RFV was determined for 30 temporal retinal arteries in a second group of 18 healthy volunteers (mean age, 30.3 ± 7.7 years). Finally, laser Doppler velocimetry (LDV) data were obtained from the same study population and compared to RFV data from the retinal vessels.

Results.: A comparison of RFV measurements in a trunk vessel of the retina and the sum of its two daughter vessels revealed a strong correlation (r = 0.98, P < 0.001). Reproducibility analysis showed that the COV for RFV was 5.9% ± 3.6%. Linear regression analysis revealed that RFV was correlated significantly with LDV measurements of mean retinal blood velocity (v _mean) and retinal blood flow (F_LDV, v _mean, r = 0.61, P < 0.001; F_LDV, r = 0.51, P = 0.004, respectively), but not significantly correlated with ocular perfusion pressure (r = −0.04, P = 0.76).

Conclusions.: These results suggest that RFV values obtained with LSFG can be considered an accurate and reliable index of relative blood flow in the human retina. Thus, RFV, a novel LSFG-derived variable, has potential for assessing retinal blood flow alterations in ocular disease.

Introduction

Altered retinal blood flow has been reported to have an important role in the pathogenesis of many ocular diseases, including diabetic retinopathy and glaucoma. ^1–4 Thus, an understanding of hemodynamic abnormalities in the retina is of critical importance in determining the pathophysiologic features of these diseases. Although a variety of techniques for the measurement of retinal blood flow have been developed, including fluorescein angiography, the blue field simulation technique, and laser Doppler velocimetry (LDV), all have various limitations. Fluorescein angiography is used widely in clinical practice, but it is invasive, can cause severe complications, including anaphylactic shock, and its results are limited by its methodology. ^5–7 The blue field simulation technique has only a small capture area that only measures the macular region of the retina. ^8–10 In combination with fundus photography, the LDV device is capable of measuring absolute blood velocity in individual retinal vessels, but requires a relatively long measurement time and is dependent on the subjects' compliance. ^11–13 Recently, Doppler Fourier domain-optical coherence tomography (FD-OCT) has been developed to measure retinal blood flow. ^14–17 Although it may have the potential to assess absolute retinal blood flow, ^18,19 it has not yet been made commercially available. 

Laser speckle flowgraphy (LSFG), a noninvasive technique based on the laser speckle phenomenon, allows the assessment of hemodynamics in the retinal vessels, choroid, and optic nerve head simultaneously. ²⁰ Although the pattern of speckle contrast produced by LSFG is known to be closely associated with blood velocity, ²¹ previous studies have suggested that the speckle signal obtained from the retinal vessels is affected not only by the velocity of the erythrocytes they contain, but also by blood flow in the underlying choroid. ^22,23 Therefore, a new approach is needed to overcome the problem of background choroidal blood flow in the human retina. 

A recent version of LSFG's accompanying software (LSFG Analyzer, v. 3.1.6; Softcare Co., Ltd., Fukutsu, Japan) provides a novel measurement parameter in the retinal vessel region, relative flow volume (RFV). It is produced by subtracting the background choroidal blood flow from the overall blood flow value of a region of interest centered on a retinal vessel, reflecting retinal flow velocity and vascular diameter. This new parameter is potentially a more accurate assessment of retinal blood flow in the superficial layer of the retina. 

Thus, the purpose of this study was to test the accuracy and reproducibility of LSFG-derived RFV measurements as an index of blood flow in the human retina, by comparing RFV measurements of a trunk vessel in the retina and the sum of the measurements of its two daughter vessels, by assessing the intrasession reproducibility of RFV, and by comparing its results to those from a bidirectional LDV system. 

Materials and Methods

Determination of RFV in the Retinal Vessel With LSFG

The principles of LSFG have been described in detail previously. ²⁰ This study used the LSFG-NAVI device (Softcare Co., Ltd.), which has been approved by the Pharmaceuticals and Medical Devices Agency in Japan. Briefly, this instrument consists of a fundus camera equipped with a diode laser (830 nm wavelength) and an ordinary charge-coupled device sensor (750 × 360 pixels). This camera is used to produce an image of the pattern of speckle contrast produced by the interference of a laser scattered by blood cells moving in the ocular fundus. The main measurement parameter of LSFG is mean blur rate (MBR), a measurement of the relative velocity of blood flow that is expressed in arbitrary units (AU). ^24–26 Images are acquired continuously at the rate of 30 frames per second over a 4-second period and then averaged to produce a composite map of ocular blood flow. The MBR values for retinal vessels always include the background intensity of choroidal blood flow. However, the influence of choroidal blood flow can be removed by manually selecting a region of interest centered on a retinal vessel and subtracting the background choroidal blood flow from the overall MBR value. This results in our novel LSFG measurement parameter, RFV (Fig. 1). First, the threshold between MBR (MBR_threshold) values in the retinal vessels and the background choroidal blood flow is determined with the following calculation:  

 

 

Figure 1

 

Determination of RFV. The MBR_threshold is the threshold between MBR values in the retinal vessels and the background choroid; f(x) is the distribution function of MBR in a cross-sectional area of the blood vessel; the width of the function at MBR_threshold is represented by m and n. The RFV in the retina was calculated by subtracting choroidal MBR from overall MBR.

In this formula, f(x) indicates the distribution function of the MBR in a cross sectional area of the blood vessel, the offset is the measured MBR outside the area of the vessel, and e indicates a mathematical constant that is based on the natural logarithm. 

Next, RFV is calculated as follows, as depicted in Figure 1. The width of the function at MBR_threshold is represented by m and n:  

 

 

Bidirectional LDV

A blood flow velocimeter (Laser Blood Flowmeter, CLBF 100; Canon, Tokyo, Japan) equipped with a tracking system was used to measure diameter (D_LDV), mean blood velocity (v _mean), and blood flow (F_LDV) in the major retinal vessels. Velocity measurements were based on the bidirectional Doppler velocimetry method. The LDV device also is capable of measuring vessel diameter. ²⁷ Details on the device have been reported elsewhere. ²⁷  

Human Experiments and Testing Protocol

The procedures in all experiments followed the tenets of the Declaration of Helsinki and were approved by the Institutional Review Board of the Tohoku Graduate School of Medicine. The experiments were consecutive and included 52 healthy subjects (mean age, 42.5 ± 15.6 years; male-to-female ratio = 22:30), recruited from volunteers at Tohoku University Hospital, Miyagi, Japan. The inclusion criteria were baseline IOP less than 22 mm Hg in eyes, as measured by Goldmann applanation tonometry and normal findings on a slit-lamp or funduscopic examination. Exclusion criteria were history of ophthalmic or general disorders, ocular laser or incisional surgery in either eye, systemic or topical medication, and refractive error greater than −6.0 diopters. All subjects abstained from alcohol and caffeine for at least six hours before the measurements were performed. On the day of the test, following a slit-lamp examination, 0.4% tropicamide (Mydrin M; Santen Pharmaceutical Co., Ltd., Osaka, Japan) was used to dilate the pupil. The patients rested in a sitting position for 10 minutes in a dark room before the examination. Blood pressure (BP) then was measured with an automated BP monitor (HEM-759E; Omron Corporation, Kyoto, Japan). Mean arterial blood pressure (MAP) and ocular perfusion pressure (OPP) were calculated as follows: MAP = diastolic BP + 1/3 (systolic BP − diastolic BP), and OPP = 2/3 MAP − IOP. Finally, the four independent experiments were performed according to their individual protocols. 

Experiment 1: In Vivo Measurement of Venous Bifurcations

We included in this experiment 34 retinal venous junctions of an independent group of 34 healthy subjects (mean age, 49.0 ± 14.8 years; male-to-female ratio = 13:21). As shown in Figure 2, the accuracy of retinal RFV measurements was tested by comparing them in the two vessels before the junction (trunk vessel 1 [Q1]) and in the vessel after the junction (daughter vessels 2 [Q2] and 3 [Q3]). All measurements were made in the temporal retina. There were 16 bifurcations located inferiorly and 18 bifurcations superiorly. 

Figure 2

 

Measurement of venous junctions. Representative retinal image of ocular blood flow, taken with laser speckle flowgraphy. Regions in the trunk vessel (white square 1) and daughter vessels (white squares 2 and 3) were manually selected in the images.

Experiment 2: Reproducibility Assessment of RFV and LDV Variables

This experiment included 30 temporal retinal arteries of 18 healthy subjects (mean age, 30.3 ± 7.7 years; male-to-female ratio = 9:9). To assess the intrasession reproducibility of RFV and LDV variables, major retinal vessel blood flow was calculated with LDV. The LDV variables included D_LDV, v _mean, and F_LDV, and were measured five times consecutively. Retinal blood flow then was assessed with LSFG at the same location in three consecutive examinations on the same day. Averaged variables were used for the statistical analysis, as in previous studies. ^27,28 The coefficient of variation (COV) was calculated for the LSFG measurements, as well as for the LDV variables. 

Experiment 3: Comparison of Retinal Blood Flow Data Obtained With LSFG and LDV

Total flow was measured in 30 temporal retinal arteries in the 18 volunteers from Experiment 2. The retinal blood flow data obtained with LSFG and LDV were compared to the data obtained in Experiment 2. 

Experiment 4: Comparison of Retinal Blood Flow Data Obtained With LSFG and Physiological Variables

A comparison of LSFG measurements of RFV and physiological variables, including systemic BP, MAP, IOP, and OPP was performed with the data obtained in Experiments 1 and 2. There were 64 retinal RFV measurements of 52 volunteers included in this experiment. 

Statistics

All data are presented as mean ± SD. Spearman's rank correlation test was used to evaluate single correlations between variables. All statistical analyses were performed with JMP software (Pro version 10.0.2; SAS Institute Japan, Inc., Tokyo, Japan). The significance level was set at P < 0.05. 

Results

In Vivo Measurement of Venous Junctions

Table 1 summarizes the clinical characteristics of the participants and their retinal vessel RFV measurements (see Supplementary Table S1 for details). As shown in Figure 3, linear regression analysis revealed a highly significant correlation between retinal blood flow in the trunk vessel (Q1) and the sum of blood flow measured in the daughter vessels (Q2 + Q3, r = 0.98, P < 0.001). 

Figure 3

 

Linear correlation analysis of the relationship between retinal blood flow in a trunk vessel (Q1: y) and the sum of blood flow in its daughter vessels (Q2 + Q3: x).

Table 1

 

Clinical Characteristics of Subjects and RFV Measurements of Retinal Vessels (Experiment 1)

Subjects, n	34
Age, y	49.0 ± 14.8
Sex, male:female	13:21
Vessel location, upper:lower	18:16
RFV in trunk vessel, AU	373.6 ± 209.3
RFV in daughter vessel 1, AU	256.9 ± 164.7
RFV in daughter vessel 2, AU	136.5 ± 63.4
RFV in sum of 2 daughter vessels, AU	393.4 ± 221.6
Systolic blood pressure, mm Hg	123.3 ± 19.0
Diastolic blood pressure, mm Hg	75.8 ± 12.1
Mean arterial blood pressure, mm Hg	91.6 ± 13.3
IOP, mm Hg	13.4 ± 2.9
Ocular perfusion pressure, mm Hg	47.7 ± 8.4

 Data are expressed as the mean ± SD. The first vessel before the bifurcation is the trunk vessel, while daughter vessels 1 and 2 are the vessels after the junction. The RFV measurements made with laser speckle flowgraphy in each area are shown in the Table.

Reproducibility of RFV and LDV

The RFV in the retinal vessels measured with LSFG was 264.9 ± 68.4 AU (range, 166.8–440.9 AU). The LDV measurements in the same area were as follows: D_LDV was 112.0 ± 12.7 μm (range, 90.2–151.1 μm), v _mean was 36.1 ± 7.9 mm/s (range, 20.6–50.1 mm/s), and F_LDV was 11.2 ± 3.8 μL/min (range, 5.1–19.5 μL/min). The COV was 5.9 ± 3.6% for RFV, while it was 3.3 ± 2.7% for D_LDV, 16.6 ± 4.9% for v _mean, and 17.6 ± 6.1% for F_LDV. 

Comparison of Retinal Blood Flow Data Obtained With LSFG and LDV

Table 2 is an overview of the clinical characteristics of the subjects and their LSFG and LDV retinal measurements (see Supplementary Table S2 for details). As shown in Figures 4A and 4B, linear regression analysis revealed that RFV in the retinal vessels was significantly correlated with v _mean and F_LDV measurements made with bidirectional LDV (v _mean, r = 0.61, P < 0.001; F_LDV, r = 0.51, P = 0.004, respectively). However, there was no correlation between RFV and D_LDV (r = 0.05, P = 0.81). 

Figure 4

 

(A, B) Linear correlation analysis of the relationship between RFV in a retinal vessel (x), and retinal blood velocity and blood flow measured at the same site with laser Doppler velocimetry (y).

Table 2

 

Clinical Characteristics of Subjects and Measurements Made With Laser Speckle Flowgraphy and Laser Doppler Velocimetry in Retinal Vessels (Experiments 2 and 3)

Subjects, n	18
Age, y	30.3 ± 7.7
Sex, male:female	9:9
Vessel measurements, number	30
Vessel location, upper:lower:temporal	18:16:1
RFV, AU	264.9 ± 68.4
Diameter, μm	112.0 ± 12.7
Velocity, mm/s	36.1 ± 7.9
Flow, μL/min	11.2 ± 3.8
Systolic blood pressure, mm Hg	115.8 ± 12.3
Diastolic blood pressure, mm Hg	69.7 ± 11.7
Mean arterial blood pressure, mm Hg	85.1 ± 11.3
IOP, mm Hg	15.3 ± 1.6
Ocular perfusion pressure, mm Hg	41.4 ± 7.4

 Data are expressed as the mean ± SD. The RFV was measured with laser speckle flowgraphy. Diameter, mean velocity, and blood flow were measured with laser Doppler velocimetry.

Comparison of Retinal Blood Flow Data Obtained With LSFG and Physiological Variables

There was no correlation between RFV and systolic BP (r = 0.05, P = 0.69), diastolic BP (r = −0.15, P = 0.24), MAP (r = −0.06, P = 0.61), IOP (r = −0.13, P = 0.32), or OPP (r = −0.04, P = 0.76). 

Discussion

This study found that RFV, a unique parameter of LSFG, had a similar value before and after the bifurcation of a retinal vessel, indicating that it is a reliable index of retinal blood flow. In addition, the COV for RFV was better than indices of velocity and blood flow derived from LDV, indicating that the reproducibility of RFV is high, and that it can be used reliably to obtain accurate evaluations of retinal blood flow. Finally, RFV measurements and the F_LDV of the same vessel were correlated significantly, indicating that RFV is a possible substitute index for the absolute flow value (F_LDV) measured by LDV. According to the results of this study, therefore, RFV can be considered a useful index for evaluating retinal hemodynamics. 

The correspondence of RFV measurements of a trunk vessel and the sum of measurements of its two daughter vessels revealed by this study was essential to demonstrate RFV's reliability as a calculated index of retinal blood flow. Notwithstanding the similar correspondence that was found in a previous study that examined the reliability of Dual-Beam Bidirectional Doppler FD-OCT, ¹⁸ LSFG should be a useful part of future investigations, as it has the advantage of being able to acquire an image of ocular blood flow in just a few seconds. This should help elucidate blood flow changes over time in specific areas of the human retina, an area of research that previously has been possible only in rabbit retinas. ²⁹  

The high reproducibility of other indexes derived from LSFG measurements, including MBR and waveform variables, has been demonstrated in our previous reports. ^26,28,30 Additionally, we found that the reproducibility of RFV compared favorably with that of LDV. The COV of RFV was lower (5.9% ± 3.6%) than the COVs of LDV (COV of 16.6% ± 4.9% for v _mean, and 17.6% ± 6.1% for F_LDV). These results for LDV agree with a previous study, ²⁷ and indicate that RFV measurements are an accurate representation of retinal blood flow. 

Currently, measurements of retinal blood flow in clinical situations are mainly made with LDV. This study, therefore, included a comparison of retinal blood flow measurements of the same vessel made with LSFG (RFV) and LDV (v _mean and F_LDV). Although RFV is a relative value, LSFG provides an easy and quick method of acquiring measurements. On the other hand, LDV provides absolute values, but requires some skill to operate. Despite these differences, RFV in the retinal arteries was correlated significantly with v _mean and F_LDV. This result supports previous findings that LSFG measurements were correlated significantly with the microsphere technique and LDV in rabbit and human retinas. ^23,29 Tamaki et al. ²⁹ reported a significant correlation between measurements of retinal blood flow made with a previous version of LSFG and absolute measurements made with the microsphere technique in rabbits (r = 0.59, P < 0.001). Although we used a different wavelength laser, our results showed a similar correlation between RFV and the LDV variable F_LDV. Additionally, Nagahara et al. ²³ found that applying a formula derived from an in vitro model of retinal blood flow measurements made with an early version of LSFG returned values that were significantly correlated with absolute LDV measurements of D_LDV, v _mean, and F_LDV in the human retina (D_LDV, r = 0.56, P = 0.032; v _mean, r = 0.59, P = 0.023; F_LDV, r = 0.83, P = 0.005). This correlation implies that LSFG theoretically is capable of producing precise retinal blood flow measurements when factors related to the underlying choroidal vessels are accounted for. Moreover, our finding of a significant correlation between RFV and LDV measurements of v _mean and F_LDV suggested that RFV could accurately isolate retinal blood flow from choroidal blood flow. However, the correlation between RFV and F_LDV in this study was lower than the correlation between the values in the study of Nagahara et al. ²³ Additionally, those values also were correlated with D_LDV, whereas RFV was not. It remains unclear why this was so, but we speculate that the current technical limitations of LSFG may be responsible, as the most current version of the technology cannot accurately assess vessel diameter. 

The present study had several limitations. Although the results revealed the potential of RFV as an index of retinal blood flow and velocity, with a linear correlation between RFV measurements of a trunk retinal vessel and the sum of its two daughter vessels, the possibility remains that our RFV measurements were influenced by absorption and/or scattering in the underlying choroidal circulation, due to the long wavelength of the diode laser. However, our finding that RFV was not significantly correlated with systemic BP, IOP, MAP, or OPP may indicate that RFV values reflect the impact of retinal blood flow autoregulation and are independent of choroidal blood flow. To confirm this, a comparison of the dynamic effects of alterations in perfusion pressure on RFV measurements of the retinal vessels and MBR measurements of the macula may be a promising avenue of research, since, as we have previously reported, LSFG is able to assess significant alterations in the hemodynamics of the optic nerve head and choroid that occur in response to posture change. ²⁶  

Although this study only assessed a small number of normal subjects, we found that maps of RFV made with LSFG were highly reproducible and reliable. Future studies using RFV to investigate diabetic retinopathy and retinal vein occlusion, common vascular disorders that can cause vision-threatening complications, such as macular edema, may lead to valuable findings. Since vascular endothelial growth factor has a key role in the progress of macular edema, intravitreal antivascular endothelial growth factor agents have emerged as a new treatment strategy. ³¹ The RFV may lead to new insights into the pathophysiology of many ocular diseases, including diabetic retinopathy and retinal vein occlusion, and help assess the efficacy of new treatments. In the future, a large multicenter study should be performed, including subjects with various ocular diseases. 

In conclusion, although the RFV index of blood flow derived from LSFG measurements is a relative value, this study found that it was an accurate and reliable method of evaluating alterations in retinal blood flow, because of its significant correlation with changes in v _mean and the F_LDV index, absolute values obtained with LDV. Thus, RFV, the novel LSFG-derived variable described in this study, has a potential use in future assessments of retinal blood flow alteration in ocular disease. 

Supplementary Materials

pdf S1 

pdf S2 

Acknowledgments

The authors thank Miyuki Nagahara and Makoto Araie for the use of the LDV device and providing technical support, and Tim Hilts for reviewing the manuscript. 

Disclosure: Y. Shiga, None; T. Asano, None; H. Kunikata, None; F. Nitta, None; H. Sato, None; T. Nakazawa, None; M. Shimura, None 

Footnotes

 YS and TA are joint first authors.

Footnotes

 YS and TA contributed equally to the work presented here and should therefore be regarded as equivalent authors.