1.

Introduction

Abnormalities in ocular blood flow have been connected with a variety of ocular diseases including diabetic retinopathy, age-related macular degeneration, and glaucoma.^1–5 However, quantification of retinal blood flow is difficult. Color Doppler imaging measures blood velocities in retrobulbar vessels, but no information on vessel diameters is available.⁶ Laser Doppler flowmetry and speckle flowgraphy can be used to assess choroidal and optic nerve blood flow in relative units.^7,8 For several years, the only technique providing information about retinal blood flow was the combined measurement of retinal blood velocity via laser Doppler velocimetry (LDV) and of vessel diameters using fundus photographs.⁹ By measuring retinal vessel diameters using the dynamic vessel analyzer (DVA) and retinal blood velocities using bi-directional LDV, however, it was shown that the range of retinal blood flow values in healthy humans is wide.¹⁰ As such, it may be more attractive to study the response of the retinal vasculature to stimuli such as flicker light or 100% oxygen breathing. Indeed, it has been shown that the hyperemic response to flicker stimulation^11,12 and the vasoconstrictor response to 100% oxygen breathing are compromised in early diabetic retinopathy.¹³

In the recent years, optical coherence tomography (OCT), a noninvasive high-resolution imaging technique that offers the possibility of performing in vivo optical ‘‘biopsy’’ of tissue structures, has become standard in the imaging of the anterior and posterior segments of the eye.¹⁴ Doppler OCT, as a functional extension of OCT, allows the extraction of phase shifts in the backscattered light, thus the ability to gain information on movements in the probed tissue region.^11,12,15 Recently, several systems were realized aiming toward the quantification of retinal blood flow using Doppler OCT in humans.^16–20

We set out to develop a novel system that is capable of measuring changes in retinal blood flow in the rat. The system uses the combination of Fourier-domain Doppler OCT for the extraction of retinal blood velocities and a fundus camera-based optical part of the system for the measurement of retinal vessel diameters. The capability of this system to quantify changes in retinal blood flow during 100% oxygen breathing and stimulation with diffuse luminance flicker was demonstrated in vivo.

2.

Methodology

2.1.

Doppler Fourier-Domain-Optical Coherence Tomography

Blood flow velocity measurements were carried out via Doppler Fourier-domain OCT (FDOCT) operating in the near-infrared wavelength range. To measure blood flow velocities via FDOCT, the phase difference $\mathrm{\Delta }\mathrm{\Phi }(z)$ at the same point between adjacent A-line recordings after Fourier transform was calculated. Since one only has access to phase changes parallel to the incident probe beam direction, the tilting angle between the velocity vector under study and the probe beam, i.e., the Doppler angle, has to be taken into account. The flow velocity $V$ is then given by

Eq. (1)

$$V(z)=\mathrm{\Delta }\mathrm{\Phi }(z)\frac{{\lambda }_0}{4\pi n\tau \cos \alpha },$$

where ${\lambda }_0$ is the central wavelength of the light source and $\tau$ is the time period between subsequent A-scans. $n$ in Eq. (1) denotes the group refractive index of blood which was—as an average of values for 632.8 and 1080 nm²¹—estimated to be 1.37. The theoretic maximum accessible longitudinal velocity $V_{\max }$ is limited by the detection speed of the charge-coupled device (CCD) and can be obtained from Eq. (1); with a central wavelength of 841 nm and an A-scan rate of 20 kHz and assuming $\mathrm{\Delta }{\mathrm{\Phi }}_{\max }=\pi$ and $\alpha =0$, it amounts to $3\mathrm{mm}/\mathrm{s}$. However, since the measurements presented here were performed on the posterior pole of the eye on vessels around the optic nerve head (ONH)—with a distance of about one to two disk diameters from its rim—the incidence angle is much larger, in the range of 90 deg, and higher velocity values can be measured. $V_{\max }$ without any wrapping artifacts was—depending on the Doppler angle—in the range of 10 to $20\mathrm{mm}/\mathrm{s}$. Yet higher velocities can be measured by compensating for the wrapping artifacts as described in the next section. The minimum velocity is given by the phase noise $\mathrm{\Delta }{\mathrm{\Phi }}_{\mathrm{err}}$ present in the system, and can–for a single-beam Doppler OCT system–calculated as $V_{\min }=\lambda ·\mathrm{\Delta }\mathrm{\Phi }/(4\pi ·\tau )$.²² However, this equation only holds true for a Doppler angle of zero degrees, which does not hold when measurements are performed in the posterior pole of the eye. With the phase noise of $\mathrm{\Delta }{\mathrm{\Phi }}_{\mathrm{err}}=0.18\text{rad}$, measured when scanning was performed, we were able to assess velocities as low as $2.5\mathrm{mm}/\mathrm{s}$.

The optical setup of the measurement system is depicted in Fig. 1. As light source, a superluminescent diode (Superlum, Carrigtwohill, Cork, Ireland) with a central wavelength ${\lambda }_0$ of 841 nm and a full width at half maximum bandwidth $\mathrm{\Delta }\lambda$ of 50 nm was used. This resulted in an axial resolution in air of $6\mu \mathrm{m}$. A beamsplitter with a splitting ratio of 50:50 divided the light coming from the source into the sample and reference arms. The free-space pathway of the reference arm contained a variable neutral density (ND) filter and a pair of prisms for balancing dispersion due to the optic components in the sample arm. In the sample arm, light was collimated by means of a fiber collimator, passed two galvanometric mirrors (GVS002; Thorlabs GmbH, Dachau/Munich, Germany) for scanning in two dimensions, and illuminated the eye via the lenses L1 and L2, where L2 is the ophthalmic lens of the retinal vessel analyzer. The interference spectrum returning from the interferometer was directed onto a $50\times 50\mathrm{mm}$ transmission grating with $L=1200$ lines per mm (Wasatch Photonics, Logan, Utah) using a collimator with a focal length $f=100\mathrm{mm}$ (OZ Optics, Ottawa, Canada). The dispersed light emerging from the transmissive grating was imaged onto a CCD camera (Atmel AVIIVA M2 CL2014, Aviva, Essex, UK) with a pixel size of $14\times 14\mu \mathrm{m}$ by means of an achromatic lens ($f=100\mathrm{mm}$).

Fig. 1

Schematic representation of the Doppler optical coherence tomography (OCT) system and coupling to the rat fundus camera. SLD, superluminescent diode; PC, polarization controller; OI, optical isolator; FC, fiber collimator; BS, beam splitter cube; NDF, neutral density filter; DC, dispersion compensation; $M_i$, Mirror; DM, dichroic mirror; L1, L2, lens; HL, halogen lamp; DG, diffraction grating; CL, camera lens; HM, hole mirror.

The transversal resolution, given by the diameter of the collimated probe beam at the cornea and the focal length of the rat eye, is approximately $11\mu \mathrm{m}$. The oversampling factor (OF) of the phase tomograms is defined as $\mathrm{OF}=w·N/d$, where $w$ is the spot size, $N$ is the number of sampling points, and $d$ is the geometric width of the tomogram.²³ With the above given spot size, $N=1000$ sampling points for in vivo measurements, and a scan width of 1.1 mm, one obtains an OF of 10. The power of the probe beam incident on the rats’ corneas was measured to be $650\mu \mathrm{W}$. The time period $\tau$ between two subsequent CCD recordings (A-scan) was set to $50\mu \mathrm{s}$, which—with a lateral tomogram dimension of 1000 A-lines for in vivo measurements—gives a frame rate of $20{\mathrm{s}}^{-1}$.

As a first OCT postprocessing step, the tomograms were corrected for sample motion using a histogram-based method.^24,25 Vessel positions within the tomogram were detected via convolution of the phase image with an elliptical template.

Phase values calculated via Fourier transform lie in the range $[+\pi ,-\pi ]$. However, in larger arteries and veins, where flow velocities are quite high, phase values can exceed this unambiguous range, leading to phase wrapping artifacts. These artifacts result in a seeming reversal of the flow in the vessel center. By determining the “true” flow direction close to the vessel wall and adding or subtracting $2\pi$ to/from the corresponding phase value, these regions were unwrapped.

Further, several postprocessing steps were performed: missing Doppler data points that occur as a result of fringe washout, particularly in the center of the vessel, were reconstructed by applying a parabolic fit to the available data points.¹⁸ In addition, resampling of the vessel data using bi-cubic interpolation was employed for achieving equal lateral and axial resolution. For the calculation of the average phase within the vessel cross-sectional area, the circular approach introduced by Szkulmowska et al.²⁶ was used. Briefly, the method uses the fact that the phase differences $\mathrm{\Delta }\mathrm{\Phi }$ are randomly distributed around the actual values. When phase averaging is performed in the angular domain, this leads to an underestimation of the mean phase difference, especially at high velocities close to the $\pm \pi$ limit. Therefore, the phase values are transformed into a complex representation and the averaging is performed by calculating the argument of the complex sum. The average flow velocity $V_{\mathrm{avg}}$ was computed over a minimum of four to five pulse periods.

Retinal vessel diameters were assessed by means of a fundus camera-based DVA system (DVA rodent, Imedos, Jena, Germany) that does not require contact with the eye, and does not influence the retinal microcirculation. Recently, the application of the DVA technique in rodents has been described for such a system using a retinal camera adapted to the rat eye, an illumination unit and the DVA software platform also used for the human system.²⁷ The DVA system used for the current experiments allows for the online measurements of retinal arterial and venous diameters with excellent reproducibility.⁸ This is because the optical system is adapted specifically to the optical properties of the rat eye with its specific illumination and imaging requirements. The measurement resolution of the system is, due to the higher magnification as compared to the human system and the applied algorithms for vessel boundary detection, up to $0.1\mu \mathrm{m}$. This allows for assessment of vessel diameters in retinal vessels down $30\mu \mathrm{m}$ in both pigmented and nonpigmented animals.

The Doppler OCT and the DVA rodent were coupled by means of a dichroic mirror DM, placed between the ophthalmic lens L2 and the hole mirror HM of the DVA system. This dichroic mirror lets the DVA’s light in the visible wavelength range pass unimpeded while redirecting the OCT probe beam (near-infrared wavelength range) into the fundus cameras’ optical path.

2.3.

Measurement Protocol

All results presented in this paper were gained from blood flow velocity and vessel diameter measurements in large retinal vessels. The measurement protocol for the hyperoxia experiments is shown in Fig. 2(a): the rodents were narcotized and ventilated with ambient air for 10 min. Thereafter, 100% oxygen was administered for another 10 min. In each animal, one major retinal vein was selected and measurements were carried out at a distance of about one disk diameter from the ONH. Starting at 8 min of the air breathing phase, the vessel diameters were continuously recorded during the whole session. Blood flow velocity measurements were performed at 8 and 9 min of the air breathing phase (baseline) and at 2, 4, 6, 8, 9, and 10 min of the oxygen breathing phase. The OCT recordings lasted 14 s to allow averaging of the phase data and calculating the mean blood flow velocities.

Fig. 2

Measurement protocols: timelines for (a) hyperoxia and (b) flicker experiments.

For all light stimulus experiments, the ambient light was dimmed to avoid corruption of the measured effects. Diffuse flicker illumination was applied at a frequency of 12 Hz via the fundus illumination path of the DVA. Each measurement session started with fundus illumination on: both the vessel diameter (via DVA) and blood flow velocity (via Doppler OCT) were measured at time point zero. After 60 s, the fundus illumination was switched off and Doppler OCT measurements were performed at time points 80 and 200 s. The dark phase lasted 3 min. Thereafter, the diffuse flicker phase started and lasted 60 s. Again, both vessel diameter and blood flow velocity of a single retinal vessel were measured at time point 260 s. At time point 300 s, the flicker illumination was switched off and the rat’s fundus was illuminated via the DVA’s halogen lamp for 60 s. At time point 320 seconds, further Doppler OCT and DVA measurements were performed before switching off the fundus illumination and repeatedly measuring flow velocities at time points 380 and 500 s. The timeline for the conducted measurements is depicted in Fig. 2(b).

3.

Results

Two phase tomograms as obtained from Doppler measurements on a rat’s retina are shown in Fig. 3. In Fig. 3(a), the phase tomogram including a vein with a diameter of approximately $50\mu \mathrm{m}$ or 100 pixels under basal conditions is shown. The measured vessel clearly shows a wrapping artifact in the center. The phase profile within the vessel after unwrapping is depicted in Fig. 3(c) and is close to parabolic. In Fig. 3(b), the phase tomogram of the same vein is shown during 100% oxygen breathing. In the phase tomogram, the vasoconstriction, indicated by a smaller vessel lumen, and the reduction in blood velocity, cognizable by the disappearance of the wrapping artifact, are clearly visible. This is also evident from the phase profile presented in Fig. 3(d), which, despite the largely reduced blood velocity, remains almost parabolic and shows a decrease of the vessel diameter by about 15 pixels.

Fig. 3

Phase tomograms including a retinal vein and extracted phase profile (a), (c) under baseline conditions and (b), (d) during breathing 100% oxygen. The extracted phase is directly proportional to the blood velocity.

In Fig. 4, the time courses of vessel diameter (black line), relative blood flow velocity (red squares), and relative blood flow (blue squares) in a single retinal vessel during breathing of 100% oxygen are depicted. As seen, the vessel diameter starts to decrease soon after the beginning of the 100% oxygen breathing. At approximately 360 s, full vasoconstriction is obtained and, thereafter, the vessel diameter remains constant. The pronounced reduction in retinal blood velocity and retinal blood flow is already visible 120 s after the start of oxygen breathing (i.e., 240 s after the beginning of the experiment). In agreement with the diameter data, almost stable conditions are achieved at 360 s after the start of the experiment. Both velocity and flow decrease almost in parallel because the reduction in blood velocity is more pronounced than the reduction in vessel diameter or vessel cross-sectional area.

Fig. 4

Time courses of retinal venous diameter D, phase shift ($\mathrm{\Delta }\mathrm{\Phi }$) and blood flow (Q) during 100% oxygen breathing. Red symbols indicate the phase shift, which is proportional to the mean blood flow velocity. VF, ventilator frequency. Dashed lines indicate the change of breathing gases.

The average data as obtained in the six rats are presented in Fig. 5. At 600 s after the start of the oxygen breathing, we observed a reduction of $42.6\pm 5.2\%$, $13.2\pm 5.2\%$, and $56.7\pm 7.4\%$ in blood flow velocity, vessel diameter and blood flow, respectively ($p<0.001$ versus baseline each).

Fig. 5

Time courses of the relative change in (a) blood flow velocity, (b) venous diameter and (c) blood flow during 100% oxygen breathing ($n=6$). Data are presented as $\text{means}\pm \mathrm{SD}$. The vertical line indicates the onset of 100% oxygen as breathing gas.

The time courses of the vessel diameter, blood flow velocity, and blood flow in one rat during stimulation with diffuse flicker light are presented in Fig. 6. During the dark periods, no measurements of vessel diameters via DVA were carried out. Typically, a short-lasting increase in vessel diameters was immediately seen after changing from dark to flicker stimulation, but the vessel diameter returned almost to its baseline value thereafter. In contrast, Doppler OCT data could be also obtained during the dark periods. As seen, the retinal blood velocity was typically slightly lower in the dark as compared to the light conditions. Since no information on the vessel diameter was available, no flow values were calculated during darkness. Stimulation with flicker light increased both blood velocity and blood flow. During lightening conditions, the retinal blood velocity stayed elevated as compared to darkness.

Fig. 6

Time courses of retinal venous diameter D, phase shift ($\mathrm{\Delta }\mathrm{\Phi }$) and blood flow (Q) during flicker stimulation. Red symbols indicate the time course of the phase shift, which is proportional to the blood flow velocity. During the periods in darkness no flow values are available, because vessel diameter was not measured.

In Fig. 7, the average increase in retinal blood velocity, vessel diameter, and blood flow in all four rats during the experiments is shown. The transition from light to dark did not change the blood velocity [Fig. 7(a)]. During flicker stimulation, on the other hand, a pronounced increase in blood velocity of 28.1% was observed ($p<0.001$ versus baseline). When the diffuse luminance flicker was turned off, the blood velocity decreased slightly but was still higher than at baseline ($p<0.001$). When the light was turned off, however, the retinal blood velocity returned to baseline again.

Fig. 7

Time courses of the relative change in (a) blood flow velocity, (b) vessel diameter and (c) blood flow during flicker stimulation ($n=4$). Data are presented as $\text{means}\pm \mathrm{SD}$. Dashed lines indicate the onset of the flicker stimulus.

Figures 7(b) and 7(c) show the effect of flicker stimulation on retinal vessel diameter and retinal blood flow. Retinal vessel diameters slightly increased by 3.6% during flicker stimulation ($p<0.05$ versus baseline) and returned to the baseline value thereafter. The retinal blood flow showed a strong increase (36.2%, $p<0.001$ versus baseline) during flicker stimulation and stayed elevated after cessation.

4.

Discussion

So far, no gold standard technique for the measurement of ocular blood flow has been realized.^6,19,28–30 Doppler OCT is a promising approach for measuring retinal blood flow. If absolute blood flow is to be measured, the angle ambiguity of the Doppler techniques has to be resolved. In humans, several techniques were realized to quantify retinal perfusion.^{7,16,18,20,25,31–36} In rats, a technique was proposed to measure the axial blood velocity in an en-face plane using raster scanning in a very small area around the central retinal artery. The retinal blood flow was calculated by integrating over the vessel area.²⁰ For measurements of pulsatile total flow, very high acquisition speeds with axial scan rates in the range of several 100 kHz are necessary. In addition, transverse sampling density has to be sacrificed. Alternatively, gated acquisition can be applied to observe pulsatile retinal flow.³⁶ Other investigators used a volumetric scanning protocol that asynchronously samples a single vessel with respect to the heartbeat, thus determining the averaged total blood flow.³⁷

However, to which degree it is useful to study total retinal blood flow in the rat is unclear. In humans, total retinal blood flow shows a wide range in healthy subjects¹⁰ and it is doubtful whether absolute blood flow measurements can be used for risk stratification in ocular disease. In rats, this is even more doubtful given the potential influence of anesthesia and the selected blood pressure value. As such, we focused on relative changes in retinal blood flow. One characteristic of our system is that it is coupled to a fundus camera allowing for the high-precision extraction of retinal vessel diameters from the fundus image, which is considered the gold standard technique for measuring retinal vessel width.³⁸ This is a critical issue, because the diameter of the vessels enters into the calculation of blood flow with the second order. Most other approaches determine the diameter from the OCT amplitude or phase images. The former method is difficult because of light scattering and absorption in retinal vessels, which often makes it complicated to identify the rear vessel wall. The latter is hampered by the phase noise and the lower limit of detectable velocity, which may lead to a velocity-dependent underestimation of the vessel width. However, both techniques are limited by the longitudinal resolution of OCT systems, which is usually in the range of 5 to $7\mu \mathrm{m}$.

In the present study, we used hyperoxia as a stimulus, which induces pronounced vasoconstriction. A variety of techniques were previously used to characterize the retinal blood flow response to 100% oxygen breathing including magnetic resonance imaging³⁹ and scanning laser Doppler flowmetry.⁴⁰ These earlier studies reported a reduction in retinal blood flow of 25% and 24.5%, respectively. Our data show a reduction of 50 to 60% in retinal blood flow, which is closer to the values obtained in humans.^41–46 Studying the response of retinal blood flow to hyperoxia is interesting, because it has been shown to be disturbed early in diabetes.⁴⁷ However, the mechanisms which lead to this reduced vasoconstrictor response to hyperoxia are poorly understood.^48,49

In contrast to hyperoxia, flicker light increases retinal blood flow due to a mechanism called neurovascular coupling. Increased neural activity as induced by light stimulation leads to an increased metabolic demand and to hyperemia of the retina and the ONH.^50–52 Several techniques were used to assess the retinal blood flow response to flicker stimulation including laser speckle flowgraphy,^53–55 fluorescent microspheres,⁸ the measurement of retinal vessel diameters,⁵⁶ and Doppler OCT.⁵⁷ Our results indicate that most of the increase in retinal blood flow is due to the increase in blood velocity, whereas retinal vessel diameters showed only minor changes. This is in good agreement with human data indicating that most of the vasodilator response occurs in the microvasculature.⁵⁸ A reduction in the retinal response to flicker stimulation can be seen in diabetic patients before any change in pattern electro-retinography becomes evident.⁵⁹ In diabetic rats, the abnormal retinal flicker response can be normalized by inhibition of inducible nitric oxide synthase.⁵⁶

In conclusion, we presented a method for measuring relative changes in retinal blood flow by combining the Doppler OCT technique with fundus camera-based measurement of retinal vessel diameters. The response to systemic hyperoxia and visual stimulation with flicker light was studied with this system. The technique has considerable potential in characterizing vascular abnormalities in animal models of retinal disease including diabetes.