1.

Introduction

The potential of noninvasive retinal imaging techniques to investigate the physiological mechanisms regulating blood hemodynamics in the human retina have long been recognized. However, techniques are now available that enable the noninvasive quantification of volumetric retinal blood flow in absolute units by simultaneously measuring vessel diameter (in micrometers) and centerline blood velocity (in millimeters per second) to derive flow in real units^{1, 2, 3} (microliters per minute). Blood flow assessment using the Canon laser blood flowmeter (CLBF) is repeatable and reproducible with low variability.^{4, 5, 6}

Disturbance of retinal blood flow is a feature of many ocular diseases,^{7, 8} including diabetic retinopathy, ^{6, 7, 8, 9, 10} age-related maculopathy,¹¹ and a subset of patients with glaucoma.^{4, 12} In the future, the assessment of retinal blood flow may play a diagnostic role in the early detection of some of these ocular diseases.¹⁰ Retinal hemodynamic assessment techniques may also be used to monitor aspects of disease progression and to evaluate the retinal vascular response to surgical and drug treatments.⁴

Lens opacity, or cataract, frequently occurs concomitantly with other eye diseases, including diabetic retinopathy, age-related maculopathy, and glaucoma. The impact of lens opacity on the assessment of retinal blood flow is unknown, despite the fact that the technique has been used widely in elderly subjects. The clarity of fundus visualization using any fundus camera-based imaging device is limited in the presence of lens opacity due to intraocular light scatter that results in reduced image contrast. Recent studies from our group have quantified the influence of artificial light scatter on various retinal imaging instruments.^{13, 14, 15} This study was designed to determine the impact of artificial light scatter on quantitative measurements of retinal arteriolar hemodynamics.

2.

Methods

One eye from each of 10 healthy young subjects between the ages of 18 and 30 [mean age 23.6 years, standard deviation (SD) 3.4] was randomly selected. All volunteers were informed about the details of the study and gave their consent to participate. The research followed the tenets of the Declaration of Helsinki and was approved by the Research Ethics Board of the University Health Network and also by the Office of Research Ethics of the University of Waterloo. All subjects were drug free with no systemic disease, no ocular abnormalities, and no history of any ocular surgery. Subjects had no lens opacity, exhibited intraocular pressures less than $21\mathrm{mm}\mathrm{Hg}$ , a logMAR (logarithm of the minimum angle of resolution) visual acuity of 0.0, or better, and a refractive error $<\pm 6.00$ DS and $<\pm 2.50$ DC.

2.2.

Quantitative Assessment of Retinal Arteriolar Blood Flow

Quantitative assessments of retinal arteriolar blood flow were acquired using the CLBF, model 100. The CLBF utilizes bidirectional laser Doppler velocimetry (LDV) to quantify the centerline blood velocity in the large retinal vessels, densitometry to measure vessel diameter, and an image stabilization system to minimize the impact of eye movement.^{4, 16} A red diode laser ( $675\mathrm{nm}$ , $80\times 50\mathrm{\mu }\mathrm{m}$ oval) is used to acquire bidirectional LDV measurements. The frequency shift of the light scattered by the moving blood cells at the illuminated site is simultaneously focused by two distinct photodetectors separated from each other by a fixed, known angle.¹⁷ The maximum frequency shift detected by each photodetector is subtracted to enable the absolute quantification of centerline blood velocity, irrespective of the angle between the moving particle and reflected beam.^{18, 19} The resulting Doppler signal is analyzed using a previously described algorithm.¹⁸ The frequency shift is determined as the frequency at which there is an abrupt reduction in the amplitude of the fluctuations in the Doppler shift power spectrum. This determination does not depend on any presumed shape of the average power spectral density curve. Velocity measurements are acquired automatically every $0.02\mathrm{s}$ throughout the 2-s measurement window, resulting in a velocity-time trace that depicts the systolic-diastolic variation of blood velocity.

Retinal vessel diameter is determined by projecting a rectangular $(1500\times 150\mathrm{\mu }\mathrm{m})$ green $\left(543\mathrm{nm}\right)$ $\mathrm{HeNe}$ diode laser perpendicular to the vessel segment measurement site. Densitometry analysis of the cross-sectional vessel image on the CLBF array sensor is used to calculate vessel diameter (Fig. 1 ). The minimum point $\left({\mathrm{Min}}_1\right)$ in the central portion of the image is designated as the nominal vessel center with $D_1$ being its signal level from the background level. When mirror reflection occurs from the vessel wall, the second minimum point is observed in the neighborhood of ${\mathrm{Min}}_1$ with signal level $D_2$ . The maximum points on either side are detected as the vessel edges with $D_3$ and $D_4$ signal levels. The half-height points ( $X_1$ and $X_2$ ) are points on the image whose signal levels are $(D_1+D_3)∕2$ and $(D_2+D_4)∕2$ , respectively. The width between $X_1$ and $X_2$ is taken to represent the uncorrected vessel diameter and is converted into micron units after correction for axial and refractive magnification effects.^{16, 20} Diameter measurements are acquired every $4\mathrm{ms}$ during the first and last $60\mathrm{ms}$ of the 2-s velocity acquisition window.

Fig. 1

Densitometry signal of a cross-sectional image acquired from a retinal arteriole: ${\mathrm{Min}}_1$ , minimum point in the central portion of the image designated as the nominal vessel center; $D_1$ , signal level of ${\mathrm{Min}}_1$ from the background level; $D_2$ , signal level of the second minimum point in the neighborhood of ${\mathrm{Min}}_1$ ; $D_3$ and $D_4$ , signal levels of the maximum points on either side designated as the vessel edges; $X_1$ and $X_2$ , the half-height points whose signal levels are $(D_1+D_3)∕2$ and $(D_2+D_4)∕2$ , respectively. The width between $X_1$ and $X_2$ is taken to represent the uncorrected vessel diameter.

An integral “eye tracking” mechanism stabilizes the laser system on the selected measurement site. Light from the red laser is projected into the center of the green rectangle.¹⁷ The green laser is manually adjusted perpendicular to the vessel segment. The resulting cross-sectional image of the vessel is focused on the CLBF array sensor. The array sensor detects any lateral motion of the vessel and, via a negative feedback loop, controls the galvanometer steering system to stabilize the green tracking laser on the selected measurement site. This system also enables identification and postacquisition rejection of velocity measurements impacted by significant saccades.

Two sequential measurements (path 1 and path 2) both of blood velocity and of vessel diameter are taken to ensure consistency of each parameter and are then averaged to give one reading. The time average of the centerline blood velocity $V_{\mathrm{mean}}$ is used along with the diameter $D$ to calculate retinal blood flow $F$ assuming that the vessel has a circular cross section and that the flow characteristics obey Poiseuille’s law. Retinal blood flow is derived using the formula: $F=\frac{1}{2}\left[\pi \right(D^2∕4\left)\right](V_{\mathrm{mean}}\times 60)$ , where flow is in microliters per minute, diameter is in micrometers, and $V_{\mathrm{mean}}$ is in millimeters per second. Magnification effects associated with refractive and axial components of ametropia are corrected to provide absolute measurements of diameter (in micrometers), velocity (in millimeters per second), and flow (in microliters per minute).

3.

Results

Group mean ( $\pm$ SD) retinal arteriolar diameter, centerline blood velocity and blood flow as a function of artificial light scatter condition are summarized in Table 1 . The change of each of the retinal hemodynamic parameters relative to the baseline is shown in Fig. 2 .

Fig. 2

Change in retinal arteriolar diameter (upper), centerline blood velocity (middle), and blood flow (lower) relative to baseline for all cell conditions. The error bars represent $\pm 1$ standard error of the mean (* denotes significant difference from baseline, i.e., the no-cell condition).

Table 1

Group mean ( ± SD) retinal arteriolar diameter, centerline blood velocity, and flow as a function of artificial light scatter condition.

The group mean retinal arteriolar diameter and blood flow significantly increased (reANOVA, $p<0.0001$ ) with increase in microsphere concentration; however, centerline blood velocity was unaffected by artificial light scatter (reANOVA, $p=0.4740$ ). The LSD post hoc test revealed that the 0.006 and 0.008% microsphere concentrations produced significantly higher diameter and flow values than the baseline condition (i.e., the no-cell condition) (Fig. 2). Importantly, the water-only cell condition produced almost identical values as the baseline for each of the three hemodynamic parameters (Table 1). The laser tracking system of the CLBF had difficulty stabilizing the laser on the measurement site for the highest microsphere concentration; for this reason, it was impossible to acquire any measurements on three subjects using the 0.008% cell.

Table 2 and Fig. 3 depict the hemodynamic assessment of 4 of the original 10 subjects using both high and low laser intensities. Diameter values attained using the high-intensity laser were closer to the baseline; only the 0.008% concentration cell resulted in significantly greater diameter values than the baseline $(p=0.0008)$ . Diameter values were significantly less with the high-intensity laser setting than the equivalent measurement with the low-intensity laser $(p=0.0007)$ . Diameter values measured with the 0.006% concentration cell were also significantly smaller when the high-intensity laser was utilized $(p=0.0306)$ . Centerline blood velocity was unaffected by laser intensity $(p=0.2817)$ . Similarly, blood flow values attained using high-intensity laser were closer to the baseline with only the highest light scatter condition being significantly greater than the baseline $(p=0.0313)$ and were significantly less than the equivalent measurement with the low-intensity laser $(p=0.0221)$ .

Fig. 3

Retinal arteriolar diameter (upper), centerline blood velocity (middle), and blood flow (lower) attained at the low (left) and high (right) intensity laser settings for the no-cell, 0.006 and 0.008% microsphere concentrations. The error bars represent $\pm 1$ standard error of the mean (* denotes significant difference from the baseline). The dashed line denotes a significant difference between identical cell conditions for the low and high laser settings.

4.

Discussion

Our light scatter model resulted in an artifactual increase of retinal arteriolar diameter $(p<0.0001)$ , which in turn led to increased retinal blood flow values $(p<0.0001)$ . The increase in blood flow was secondary to the induced change in arteriolar diameter since flow is directly proportional to the second power of diameter, according to the formula used by the CLBF. Centerline blood velocity, however, was not affected by light scatter.

The densitometry technique that is used by the CLBF to measure vessel diameter is increasingly impacted as the magnitude of artificial light scatter increases (Fig. 2). We hypothesize that as the light scatter increases, the densitometry image of the vessel loses contrast and the slopes of the densitometry profile broaden (Fig. 4 ). This results in a greater separation of the half-height points causing an artifactual increase in the diameter of the vessel. Ultimately, light scatter will result in breakdown of the CLBF laser tracking system but artifactually increased vessel diameter values can be acquired prior to this point. On the other hand, it is clear that the bidirectional LDV technique used to measure centerline blood velocity is robust to the optical effects of artificial light scatter.

Fig. 4

Densitometry profiles of a vessel measured by the CLBF (left) and schematic representation (right): Upper, baseline (no-cell condition); lower, 0.006% microsphere concentration. DW is the half-height width.

Table 2

Group mean ( ± SD) retinal arteriolar diameter, centerline blood velocity, and flow as a function of artificial light scatter condition and laser intesity.

It is apparent that the impact of light scatter on the densitometric estimation of retinal vessel diameter can, in part, be negated by increasing the intensity of the $\mathrm{HeNe}$ laser (Fig. 3). Future instrumentation could incorporate a negative feedback loop that automatically increases $\mathrm{HeNe}$ laser intensity in situations where the slope of the densitometry profile becomes shallow. In addition, when assessing retinal blood flow in patients with concomitant cataract, we recommend using the high-intensity laser setting of the CLBF. CLBF parameters that do not rely on diameter estimation but still provide useful hemodynamic information, such as centerline blood velocity and maximum to minimum velocity ratio, should be given greater attention when assessing retinal blood flow in patients with concomitant cataract.

The data preclude the possibility that a real change in diameter occurred. To explain, if the apparent increase in arteriolar diameter was real then this would be reflected by an increase in blood velocity (since vascular resistance would decrease). The magnitude of apparent increase in arteriolar diameter, if real, would be predicted to result in an increase of blood velocity. An increase in retinal arteriolar diameter of $3.2\pm 1.4\%$ following hypercapnic provocation resulted²¹ in an increase of blood velocity of $26.4\pm 7.0\%$ . In effect, the finding of an increase in arteriolar diameter in the absence of an increase in blood velocity is contradictory and can be explained only as an artifact of the light scatter cells on the densitometric estimation of vessel diameter. Furthermore, the magnitude of apparent increase in diameter for the 0.008% light scatter cell is outside the physiological range of diameter change produced by extreme hypercapnic and hyperoxic provocations.^{21, 22}

Are the results of this study relevant to only the limited number of CLBF users worldwide and to those individuals with a specific interest in retinal hemodynamic assessment? A number of techniques designed to assess retinal vessel diameter, including the retinal vessel analyzer²³ and the assessment of retinal photographs used in the Atherosclerosis Risk in Communities Study,²⁴ have been introduced. In particular, recent results from the Atherosclerosis Risk in Communities Study found that retinal vascular abnormalities are predictive of cardiovascular disease.²⁵ However, both of these techniques^{23, 24} use algorithms to assess retinal vessel diameter that are very similar to that used in the CLBF and therefore may be susceptible to an artifact resulting from light scatter. Cataract is an inevitable consequence of aging and, therefore, care must be exercised in the interpretation of studies of retinal vessel diameter that utilize cohorts with age as a covariable.

The results derived from any model, however, must be interpreted with caution since the model may not accurately simulate the impact of true cataract. From a clinical perspective, we previously demonstrated the impact of each artificial light scatter condition on fundus visualization achieved with a digital fundus camera that utilizes a polychromatic light source.^{14, 15} Direct comparison of our artificial light scatter model with a subjective, slit-lamp-based cataract grading scale is difficult; however, the magnitude of light scattering medium used in this study is far less than that used to assess confocal optics based instrumentation.¹⁵ Unpublished data from our lab demonstrates that the artificial light scatter model impacts visual acuity and letter contrast sensitivity in a very similar manner to that of nuclear sclerotic cataract. Future work will investigate the impact of light scatter induced by cataract on retinal blood flow assessment.