Factors Associated With Ocular Perfusion Measurements as Obtained With Laser Speckle Contrast Imaging

Abstract

Purpose: To study the ocular and systemic factors affecting optic nerve head (ONH) perfusion data as obtained using a commercially available laser speckle flowgraphy (LSFG) device in a cohort of Caucasian subjects without ocular diseases. Also, to assess the intrasession repeatability and intersession reproducibility of ONH, macular, retinal, and choroidal perfusion.

Methods: Seventy-five healthy eyes of 75 Caucasian participants underwent LSFG and spectral-domain optical coherence tomography (SD-OCT) on the same visit. Perfusion of the ONH was assessed with LSFG, and SD-OCT was used to measure peripapillary retinal nerve fiber layer thickness (RNFLT) and macular ganglion cell plus inner plexiform layer thickness (GCIPLT). The intrasession repeatability and intersession reproducibility of ONH and macular perfusion and retinal and choroidal relative flow volume (RFV) were evaluated in 20 participants measured on three different days over a 6-month period.

Results: Intrasession and intersession intraclass correlation coefficients of LSFG parameters ranged from 0.787 to 0.967 and from 0.776 to 0.935, respectively. Intersession 95% prediction intervals for the ratio of two measurements were wider for RFV indices than for ONH and macular perfusion parameters. The multiple regression analysis indicated that higher ONH perfusion was associated with younger age, female sex, smaller optic disc area, and higher RNFLT. RNFLT was an independent predictor of ONH perfusion, whereas GCIPLT was not. Each 1-µm increase in RNFLT was associated with a 0.272 arbitrary unit increase in ONH perfusion.

Conclusions: LSFG measurements of optic disc perfusion are influenced by sex, age, and anatomical variations in optic disc area and RNFLT.

Translational Relevance: Better evaluation of ocular blood flow will result in better diagnosis and treatment of various ocular diseases.

Introduction

Several ocular diseases are caused by or related to circulatory disturbances, such as diabetic retinopathy, age-related macular degeneration, glaucoma, and ischemic optic neuropathy. As well as being relevant in many ophthalmic diseases, ocular circulation can potentially serve as an accessible and clinically useful window to the microcirculation of other organs, such as the kidney and the heart.^1–3 Nonetheless, measurement of blood flow in the human eye circulation is challenging and is mainly restricted to research devices.⁴ Optical coherence tomography angiography (OCTA) has recently emerged as a reference technique for imaging the retinal and choroidal vasculature in clinical practice. Currently, there are commercially available OCTAs that provide flux values in the superficial retinal plexus,⁵ but quantification of flow in other vascular compartments is not yet available.⁶ Interest in laser speckle contrast imaging (LSCI) has grown in several fields of medicine, as it is a non-invasive and fast method for evaluation of relative blood flow.^7,8 Based on the laser speckle phenomenon,⁹ it has been used to capture the movement of blood cells within the circulation of the optic nerve head (ONH), macula, and iris.^10,11 The LSCI method has been validated in animal studies, where the LSCI-derived signal has been shown to be strongly correlated with ONH volumetric blood flow rates.^12,13 In patients, the technique has been shown to be useful in evaluating the effect of medical interventions, in confirming the presence of ocular diseases characterized by blood flow abnormalities, and in monitoring their progression.^14–22 However, a number of obstacles prevent this technique from being stably introduced into the clinical setting, such as incomplete knowledge of factors affecting the variability of its measurements. 

The purpose of the present study was to identify the factors affecting the ocular blood flow measurements as obtained using a commercially available LSCI device in a cohort of Caucasian subjects without ocular diseases. In particular, the relationship between the thicknesses of the peripapillary retinal nerve fiber layer and the macular ganglion cell layer plus inner plexiform layer and ONH blood flow was observed. In addition, we tested for intra- and intersession LSCI variability. 

Methods

Subjects

This cross-sectional study included 75 eyes of 75 participants selected at two centers (Eye Clinic of the University Hospital of Parma, Italy, and Eye Clinic of the University Hospital of Basel, Switzerland), from two sources: (1) subjects attending the eye clinic for a routine ophthalmological examination, and (2) subjects who participated in previously published studies.^20,23 A subgroup of 40 eyes of 20 participants were included in the repeatability and reproducibility analysis. 

The inclusion criterion was the absence of previous or ongoing retinal or optic nerve disease. Participants in the age range from 10 to 82 years were included. The gender ratio was 33 (female) to 42 (male). Exclusion criteria were absolute spherical equivalent > 8 diopters (D), absolute cylinder > 2.50 D, best-corrected visual acuity worse than 20/25, intraocular pressure > 21 mm Hg, previous ocular surgery except for uncomplicated cataract surgery performed at least 6 months prior to enrollment, significant opacities of the optical media (e.g., corneal scars, Lens Opacities Classification System III [LOCS III] grading ≥ 3, posterior capsule opacification, vitreous opacities), uveitis, neurological diseases, diabetes mellitus, uncontrolled hypertension with systolic blood pressure > 170 mm Hg and/or diastolic blood pressure > 100 mm Hg. 

All research adhered to the tenets of the Declaration of Helsinki, and the study protocols were approved by the Ethics Committees of the two centers. Written informed consent was obtained from all subjects. 

Study Procedures

Subjects were instructed to abstain from alcohol, tea, and coffee in the 6 hours preceding each study visit, as these substances may affect LSCI measurements.^24,25 All subjects underwent a comprehensive examination that included medical history, refraction measurement using autorefractometry, best-corrected visual acuity testing using standard Early Treatment of Diabetic Retinopathy Study charts, slit-lamp biomicroscopy, fundus examination, measurement of intraocular pressure with a Pulsair non-contact tonometer (Keeler, Berkshire, UK) and measurement of systolic and diastolic blood pressures at the left brachial artery at the height of the heart in a sitting position with an automatic sphygmomanometer (Dräger Infinity Delta; Dräger USA, Houston, TX). If both eyes of a subject were eligible, the study eye was randomly chosen. LSCI and OCT were then carried out. 

Laser Speckle Contrast Imaging

A commercially available laser speckle flowgraphy (LSFG) device was used (LSFG-RetFlow; Nidek, Aichi, Japan). It consists of a fundus camera equipped with an 830-nm diode laser as the light source and a charge-coupled device sensor (750 × 360 pixels) as the detector. One LSFG recording takes about 4 seconds to acquire 118 frames. The image angle of the LSFG is 21°. The primary output parameter of LSFG is the mean blur rate (MBR), which represents a measure of relative blood flow velocity expressed in arbitrary units (AU). 

The LSFG acquisition protocol consisted of two consecutive scans centered at the ONH, followed by two consecutive scans centered at the fovea without moving the subject from the device. A single experienced operator took LSFG images at each center. In the few subjects for whom LSFG images could not be acquired in miosis, the pupil was dilated with 0.5% tropicamide eye drops. A previous study showed that pupil dilation with 0.5% tropicamide had no effect on repeatability or absolute MBR values.²⁶ Phenylephrine was avoided because of its effect on retinal vessels.²⁷ The LSFG Analyzer 3.8.0.4 (SoftCare, Fukuoka, Japan) was used for the analysis. A double automatic evaluation of image quality was performed. First, at the end of each acquisition, a binary quality check was performed by the software to exclude low-quality scans with gross artifacts such as blink artifacts; all scans passing this first level were checked frame by frame, and low-quality frames were automatically detected and eliminated. ONH blood flow was analyzed within a region of interest (ROI) that was drawn in a semiautomated fashion to fit the optic disc borders while referring to the color fundus photograph, as previously described.^28,29 In eyes with an optic disc contour that deviated from the ellipsoid, such as eyes with myopic discs, the region of interest was drawn using the “Spline” function. The optic disc area measured in pixels on the LSFG map was recorded. The MBR of different areas of the ONH was calculated after automatic binarization to differentiate between ONH larger vessels (MV, mean MBR of larger vessels) and ONH tissue capillaries (MT, mean MBR of ONH tissue capillaries) according to the chosen ROI. For the total optic nerve head area, the average MBR of all areas (MA) was calculated. The average of the values obtained from the two consecutive scans was then calculated. ROIs plotted at baseline were saved and used in subsequent scans. For macular blood flow, a rectangular ROI including the entire macular scan was drawn, as previously described,²⁹ and the average MBR (MA) was analyzed. The signal from this region is believed to be derived largely from choroidal circulation.^30,31 Finally, we analyzed the relative flow volume (RFV) parameter in retinal arteriolar, retinal venular, and choroidal vessel segments.^32,33 A line was drawn on the vessel center, and a rectangular ROI was placed around it accordingly. The vessel diameter inside this ROI represents about 1/3 of the total ROI width. The choroidal vessel segment was selected in areas without overlying large retinal vessels, as previously described.³³ A representative LSFG image with all of the regions of interest is shown in Figure 1. 

Figure 1.

 

LSFG analysis. The LSFG perfusion map is superimposed on the color fundus photograph, with the optic disc region marked by the white elliptical ROI. The marked vessel segment regions show the retinal venular segment (blue ROI), the retinal arteriolar segment (red ROI), and the choroidal vessel segment (yellow ROI). The large white rectangle represents the ROI used for measuring macular perfusion.

Optical Coherence Tomography

OCT measurements were taken with a CIRRUS spectral-domain OCT (SD-OCT) device (with software version 8.1.0.117; Carl Zeiss Meditec, Dublin, CA). Parameters used were peripapillary retinal nerve fiber layer thickness (RNFLT) and macular ganglion cell layer plus inner plexiform layer thickness (GCIPLT), which were automatically calculated using the optic disc cube 200 × 200 protocol and the macular cube 512 × 128 protocol, respectively. These layers were chosen specifically for their clinical relevance in glaucoma and other optic neuropathies. The optic disc area (in mm²) and the cup-to-disc area ratio were also obtained from the optic disc cube 200 × 200 protocol. All included scans had ≥6/10 signal strength and were checked for segmentation errors or blink artifacts by an experienced operator. When an OCT scan was judged to be of low quality because of signal strength or the presence of errors or artifacts, it was repeated until good quality was achieved. 

Intersession Reproducibility

To test for intersession reproducibility, longitudinal analysis was carried out in 40 eyes of 20 subjects who were measured by both LSFG and OCT on three different days at the same time of the day over a 6-month period. Testing the variability of the SD-OCT device was not necessary, as this was reported earlier.³⁴ 

Statistical Analysis

Categorical data are presented as absolute frequencies (%), and data for continuous variables are presented as mean and standard deviation or as median and interquartile range (IQR), as appropriate. Multiple regression analysis was used to detect the factors contributing to ONH MA, ONH MT, or ONH MV adjusting for the confounding effects of other factors. Sex, age, spherical equivalent, ocular perfusion pressure, optic disc area (in pixels), RNFLT, and GCIPLT were included as explanatory variables. Ocular perfusion pressure in the sitting position was calculated as follows: mean arterial pressure = diastolic blood pressure + 1/3(systolic blood pressure – diastolic blood pressure) and ocular perfusion pressure =2/3 mean arterial pressure – intraocular pressure.³⁵ Each parameter was normalized to estimate the standardized partial regression coefficient. The choice of predictors was based on the clinical judgment of the investigators and previous literature; no univariable prefiltering was performed.³⁶ Differences between age groups were analyzed by Pearson's χ² tests, analysis of variance (ANOVA), and Kruskal–Wallis tests, as appropriate. Intrasession repeatability and intersession reproducibility of seven LSFG parameters (ONH MA, ONH MV, ONH MT, macula MA, arteriolar RFV, venular RFV, and choroidal RFV) were evaluated using intraclass correlation coefficients (ICCs) in a random-effect framework.³⁷ Intrasession repeatability was evaluated using the two consecutive LSFG scans taken during each session, and intersession reproducibility was calculated from the three sessions over a 6-month period. Intrasession and intersession 95% prediction intervals of the ratio of two measurements were calculated according to a log–linear random effects model, with eye/session as the random effect for the intrasession analysis and eye and session as the nested random effects for the intersession analysis. Values of P < 0.05 were considered statistically significant without correction for multiple testing. 

Results

LSFG Intrasession Repeatability and Intersession Reproducibility

Forty eyes of 20 participants (10 males, 10 females; mean age, 34.5 ± 11.1 years) were included in the longitudinal assessment. In all of these subjects, all LSFG images were acquired without pupil dilation. ICCs with 95% confidence intervals (CIs) for intrasession repeatability and intersession reproducibility are shown in Table 1 and in Figure 2. Intrasession ICCs ranged from 0.787 to 0.967, and intersession ICCs ranged from 0.776 to 0.935. The ratios between the intersession and intrasession standard deviations of the LSFG measurements ranged from 1.000 to 1.779. Intrasession and intersession 95% prediction intervals of the ratio of two measurements are shown in Table 2. Intersession prediction intervals were wider for RFV indices than for ONH and macular measurements. 

Table 1.

 

Intraclass Correlation Coefficients With 95% Confidence Intervals

 

Figure 2.

 

LSFG intrasession repeatability and intersession reproducibility. Dots indicate ICCs, and the whiskers span the 95% CIs for intrasession (within visit) repeatability and intersession (within 6 months) reproducibility.

Table 2.

 

95% Prediction Intervals for the Ratio of Two Measurements

 

LSFG and SD-OCT Results

Seventy-five eyes (38 right eyes and 37 left eyes) of 75 participants were included in the cross-sectional analysis (42 males, 33 females; median age, 35 years; range, 10–82 years). Their ONH blood flow, RNFLT, and GCIPLT values are listed in Table 3 and Supplementary Tables S1 to S3 based on three age groups. Seven subjects were smokers, and 12 subjects were taking medication for arterial hypertension. Dilation with tropicamide was performed before LSFG acquisition in seven subjects (9%). The ONH region of interest on the LSFG map was drawn using the “Spline” function in eight subjects (11%). The multiple regression analysis results with standardized partial regression coefficients and 95% CIs are shown in Table 4 and in Figure 3. 

Table 3.

 

Demographics and Baseline Characteristics of Participants Included in the Cross-Sectional Analysis

 

Table 4.

 

Multiple Linear Regression Analysis

 

Figure 3.

 

Multiple linear regression analysis. Dots indicate standardized partial regression coefficients, and the whiskers span the 95% CI for each predictor.

Female sex was associated with higher global ONH perfusion (ONH MA, P = 0.005) and higher ONH arteriolar/venular perfusion (ONH MV, P = 0.020), whereas the partial correlation between sex and ONH MT was not significant (P = 0.457). All three ONH blood flow parameters showed statistically significant declines with increasing age (P < 0.001, P = 0.045, and P = 0.005 for MA, MT, and MV, respectively). Each additional year of age was associated with a decrease of 0.157, 0.058, or 0.211 AU in ONH MA, MT, or MV, respectively. A smaller optic disc area was associated with higher ONH MA and MT (P = 0.011 and P < 0.001, respectively), whereas the partial correlation between optic disc area and MV was not significant (P = 0.163). Each 1000-pixel decrease in optic disc area was associated with a 0.249-AU or a 0.244-AU increase in ONH MA and MT, respectively. RNFLT was a statistically significant predictor of all three ONH blood flow parameters (P = 0.003, P <0.001, and P = 0.020 for MA, MT, and MV, respectively), but GCIPLT was not (P = 0.765, P = 0.145, and P = 0.710 for MA, MT, and MV, respectively). Each 1-µm increase in RNFLT was associated with increases of 0.272 AU, 0.214 AU, and 0.372 AU in ONH MA, MT, and MV, respectively. Spherical equivalent and ocular perfusion pressure were not statistically significant predictors of any of the ONH blood flow indices after multiple regression analysis (Table 2). The inclusion of “diagnosis of arterial hypertension” and “cup-to-disc area ratio” as explanatory variables did not improve the model (Supplementary Table S4). 

Discussion

LSFG-derived ONH blood flow measurements of Caucasian subjects are influenced by sex, age, optic disc size, and RNFLT. This is the first time, to our knowledge, that RNFLT has been shown to be an independent predictor of ONH blood flow measurements as obtained with LSCI in healthy eyes. This finding is consistent with previous OCTA studies assessing retinal nerve fiber layer and retinal perfusion in the peripapillary region.^38,39 

The first LSFG device was introduced as a noninvasive method of imaging circulation in the choroid and ONH and was primarily used to monitor changes at a single site in the same eye. The latest version of LSFG, on the other hand, has been validated for interindividual and intergroup comparisons at the level of the ONH.¹³ The LSFG-RetFlow signal is indeed influenced by the absorption of the laser beam by the target tissue, and absorption at the level of the retinal pigment epithelium differs depending on fundus pigmentation. However, a study in albino and pigmented rabbits showed that, regardless of the fundus pigmentation, the LSFG measurement at the ONH, which is a pigment-free tissue, is highly correlated with capillary blood flow measured with the hydrogen gas clearance technique. This finding suggests that the LSFG measurement in the ONH tissue is usable for interindividual and intergroup comparisons.¹³ Previous studies have reported ONH measurements obtained with an earlier version of the LSFG (LSFG-NAVI; Softcare Ltd., Fukuoka, Japan) in normal individuals.^26,40 However, data are missing regarding the LSFG-RetFLow. Thus, the present study provides reference values that could serve as benchmarks for future studies with the LSFG-RetFlow, in particular for sample size estimations. 

Previous studies have found sex-specific differences in ONH MBR, with females having higher values.^40–42 Possible explanations included differences in sex hormone levels, hemoglobin concentration, rate of atherosclerosis, and stature. Notably, one study found that sex was not but hemoglobin concentration was an independent factor correlated with the ONH MA.⁴² In the present study, female sex was an independent factor correlated with higher ONH MA and MV, whereas sex did not affect ONH MT. Differences in the origin of LSFG parameters may be a possible explanation. The ONH MV is primarily an index of blood flow within the retinal circulation, whereas the ONH MT is an index of capillary blood flow within the ONH tissue, which has a complex vascular supply. The surface is supplied by the retinal circulation, whereas the prelaminar area and the lamina cribrosa are supplied mainly by the ciliary circulation. Therefore, it is possible that sex affects ciliary circulation differently than retinal circulation. 

Cross-sectional studies have shown that ONH blood flow decreases with increasing age, particularly the MA and MV parameters.^{26,28,40–42} On the other hand, two studies found that the MT parameter is not affected by age.^28,40 We found an age-dependent decline in all three ONH blood flow indices and showed a stronger effect of age on MA and MV than on MT. This is consistent with the age-related loss of radial peripapillary capillaries and optic nerve fibers.^43,44 

Recently, the effects of structural parameters of ONH, such as disc and cup area, on LSFG indices were studied.²⁸ A negative correlation between optic disc area and MT measurements was found, which agrees with our findings, and it was hypothesized that a higher MT in smaller discs could reflect higher blood flow rate per unit area in discs having a higher nerve fiber density. To date, no studies have examined the relationship between the measurement results of LSFG and the peripapillary RNFLT and macular GCIPLT in healthy eyes. In the present study, we showed that the RNFLT, but not the GCIPLT, is an independent factor associated with ONH blood flow in healthy eyes. In particular, the effect of RNFLT was stronger on tissue blood flow (i.e., ONH MT). Histological studies of the normal human retina supported a close correlation between RNFLT and radial peripapillary capillaries volume.⁴⁵ Furthermore, RNFLT was significantly correlated with retinal vessel diameters,⁴⁶ OCTA-derived vessel area density of the parapapillary region,^38,39 retinal blood flow measured with Doppler OCT,⁴⁷ and theoretical predictions of absolute retinal blood flow⁴⁸ in healthy eyes. The findings of the present study also support such a relationship. It is possible that higher RNFLT is associated with higher oxygen and nutrient demand and thus with higher perfusion. We found a stronger association with RNFLT than with GCIPLT, in line with the results of OCTA studies and other retinal vascular parameters.⁴⁹ This could be because GCIPLT measurement is only derived from a portion of the macula, whereas peripapillary RNFLT and ONH LSFG measurements represent retina-wide indexes. 

Next, we assessed the intrasession repeatability and intersession reproducibility of LSFG measurements. Intrasession repeatability of LSFG has already been reported. Coefficients of variations ranging from 1.9% to 6.81% for ONH blood flow and 4.1% for macular blood flow or ICCs ≥ 0.90 were found using the LSFG-NAVI.^26,50 Coefficients of variation for the RFV parameters were found to range between 5.9% for retinal arteriolar RFV and 7.7% for choroidal RFV (ICC = 0.86).^32,33 We found that intrasession ICCs were >0.90 for all LSFG parameters, except for ONH MT and MV. In general, we would always suggest acquiring two consecutive LSFG scans and using the average value to improve reliability, especially when considering MV and MT. Although the first LSFG device was introduced as early as 1990, there are no substantive data on intersession reproducibility of the method. Only limited research has been conducted using prototype devices and small sample sizes.^51,52 Intersession ICCs were slightly lower than intrasession ICCs, as also found in previous OCTA studies with a similar design.^53,54 Comparing the intra- and intersession variability results with studies performed with SD-OCT and OCTA, we found generally comparable ICCs.^34,53,54 Finally, we determined 95% intersession prediction intervals for the ratio of two measurements. These are important to consider for any evaluation of short-term changes in blood flow that could be due to an intervention or a pathological process. For example, given a 95% prediction interval of 0.83 to 1.21 for ONH MA, an 18% decrease or a 22% increase in this parameter over a 6-month period would represent a change beyond the expected variability. Of note, prediction intervals were wider for RFV indices than for ONH and macular measurements, indicating greater variability in the former. 

Our study has limitations. First, we did not measure axial length and did not correct the OCT and LSFG measurements for ocular magnification. As partial compensation for this, we included the spherical equivalent in the regression model. However, a stricter ametropia cut-off point or correction of the lateral scale before extraction of quantitative data from the devices would be necessary to avoid the effect of magnification. Second, some ocular parameters that were found to influence optic disc blood flow, such as the β-peripapillary atrophy area²⁸ or hemoglobin concentration, were not measured. Third, a few participants were smokers or on hypertension medication, and there is evidence of altered ocular hemodynamics in these conditions. We considered the data of participants with hypertension separately and included them in the multiple regression analysis. We found that the effect was only minor, which is demonstrated in the Supplementary Material. However, further studies should be performed on the effect of systemic hypertension and vascular risk factors on ophthalmic LSFG measurements. Determining the impact of these factors would be important to further the clinical use of LSFG. Fourth, a limitation of the current LSFG system is the manual work that must be performed in terms of both scan acquisition (e.g., there is no autofocus) and analysis (e.g., the ROI must be drawn semimanually). Greater automation of the technique would be important to make it more examiner independent, as it is for OCT and other imaging techniques commonly used in clinical practice. Finally, the intersession prediction intervals given here refer to healthy eyes, but they might be different under pathological conditions; therefore, these values should be extrapolated to clinical studies with caution. 

In conclusion, we investigated the factors affecting the variability of ocular blood flow measurements as obtained with a commercially available LSCI device in Caucasian subjects and found that these are independently affected by sex, age, optic disc area, and RNFLT. When considering patient follow-up, given the four factors that have been described as affecting blood flow, age and RNFLT would be of relevant influence. 

Acknowledgments

Disclosure: J. Fröhlich, None; M. Cattaneo, None; P. Valmaggia, None; P.M. Maloca, None; K. Gugleta, None; L. Schmetterer, None; H.P.N. Scholl, None; G. Calzetti, None