A Method to Evaluate the Frequency Dependence of Retinal Vascular and Neural Metrics to Stimulus Modulation

Abstract

Purpose: Flickering light stimulation induces functional hyperemia, characterized by vasodilation, blood flow augmentation, and venous oxygen elevation. We present a new method to investigate the frequency dependence of metrics associated with functional hyperemia.

Methods: A novel optical imaging system was developed to quantify retinal blood vessel diameter (D), oxygen saturation (SO₂), and the inner retinal oxygen extraction fraction (OEF) before and after light flicker at different frequencies. Measurements were performed in 10 visually normal subjects (20–62 years) at flicker frequencies from 2 to 30 Hz. In addition, a measure of neural function was obtained by steady-state pattern electroretinography (ssPERG) across a similar range of frequencies.

Results: Flicker stimulation greater than 2 Hz increased D, increased SO₂ in veins, and decreased OEF. The maximum response for all metrics was obtained between 16 and 30 Hz, indicating that vascular and oxygenation metrics share a similar frequency response with light flicker. ssPERG amplitudes were positively correlated with flicker-induced increases in venous D and SO₂. ssPERG amplitude was negatively correlated with flicker-induced decreases in OEF.

Conclusions: We present a novel retinal imaging method to evaluate the frequency dependence of changes in D, SO₂, and OEF to light flicker stimulation. The relationship between these metrics and ssPERG amplitudes was evaluated.

Translational Relevance: The frequency-dependent response of retinal D, SO₂, and OEF established in healthy individuals herein has the potential to serve as a biomarker of vascular and tissue abnormality in future studies of retinal disease.

Introduction

The retina is one of the most metabolically active tissues in the human body and responds actively to visual stimuli via neurovascular coupling mechanisms.^1–3 During diffuse light flicker stimulation in humans, the retina undergoes a functional hyperemia response characterized by increases in blood vessel diameter (D), blood flow, and oxygen saturation of hemoglobin (SO₂) in veins.^4–13 Moreover, the inner retinal oxygen extraction fraction (OEF), the fraction of oxygen extracted from the retinal vasculature for metabolism, characteristically decreases during functional hyperemia.^14,15 This is due to a stimulus-induced increase in oxygen delivery that exceeds the increased oxygen metabolic demand. OEF abnormalities in response to flicker have been demonstrated in diabetic retinopathy¹⁶ and may be useful for studying how neurovascular coupling is affected in other retinal diseases. 

Much of the work to characterize functional hyperemia and neurovascular coupling has been conducted using light flicker stimulation at or near 10 Hz, as supported by earlier studies that characterized the frequency dependence of flicker-induced changes in D and relative blood flow at the optic nerve head (ONH).^17–20 Work has also been done to evaluate the frequency dependence of functional hyperemia and associated metrics in animal models.^21–23 However, to the best of our knowledge, the frequency dependence of inner retinal hemodynamic and oxygen metrics in response to light flicker has not been described in humans. Moreover, the association between flicker-induced changes in the retinal vasculature and neural function is not well understood. The electroretinogram (ERG) provides a useful approach to quantify retinal neural function. Previous work has used this approach to study neurovascular coupling, showing that flicker ERG responses are associated with flicker-induced vasodilatory responses.¹⁸ 

Here, we introduce a new optical imaging system for quantification of D, SO₂, and OEF before and after light flicker stimulation across a range of temporal frequencies. These responses are compared to neural responses of the inner retina measured with the steady-state pattern ERG (ssPERG) across a similar range of stimulus frequencies. This study is intended to provide new insight into the frequency dependence of neurovascular coupling and to provide an approach that may be useful for evaluating neurovascular coupling in health and disease. 

Methods

Subjects

This study was approved by an institutional review board at the University of Illinois Chicago and adhered to the tenets of the Declaration of Helsinki. Each subject provided written informed consent prior to participation. Data were obtained from 10 visually normal individuals with no history of ocular or systemic disease (four males and six females). Subjects ranged in age from 20 to 62 years (mean ± SD age, 44.5 ± 12.7 years). 

Retinal Imaging System

A novel optical imaging system was designed to assess D, SO₂, and OEF before and after light flicker stimulation from 2 to 30 Hz, a range of frequencies that has previously been shown to induce vasodilation in humans.^18,19 Figure 1 provides a schematic of the primary components of the optical system. A slit lamp biomicroscope (Haag-Streit, Köniz, Switzerland) was modified to accommodate a customized filter wheel (printed using polylactic acid on a Prusa 3D printer; Prusa Research, Prague, Czechoslovakia), a stepper motor to rotate the filter wheel (ROB-10846 Stepper Motor; SparkFun Electronics, Niwot, CO, USA), three hard-coated bandpass optical filters (Thorlabs, Newton, NJ, USA), a “white” light-emitting diode (LED; XLamp, CMB1306-18V; Cree LED, Durham, NC, USA), and a complementary metal oxide semiconductor (CMOS) camera (Kiralux CS126MU; Thorlabs, Newton, NJ, USA). Based on the stepper motor control, the white light was passed through optical filters at 606 ± 5, 570 ± 5, or 530 ± 5 nm to provide illumination of the retina for imaging by the CMOS camera. Light at 606 nm and 570 nm was used to acquire images for oximetry, corresponding to oxygen-sensitive and oxygen-insensitive wavelengths, respectively.²⁴ Light at 530 nm was used to elicit a maximal response during light flicker stimulation. The hardware modifications were controlled using a microcontroller (Uno; Arduino, Monza, Italy) programmed with customized software written in the LabVIEW programming language (National Instruments, Austin, TX, USA). 

Figure 1.

 

Schematic diagram of the optical imaging system for the evaluation of inner retinal vascular and tissue responses to light flicker stimulation. System components are indicated. Dashed lines indicate the optical path, and thin lines indicate electronic connections between system components.

Our previous optical imaging system for the assessment of inner retinal vascular and tissue responses provided light flicker stimulation at a fixed 10-Hz frequency by periodically obstructing the light source with a light block attached to a solenoid.¹⁴ For the novel instrument described here, square-wave light flicker stimulation was software controlled by modulating the state (i.e., on or off) of the white LED. This has several advantages, including frequency control above and below 10 Hz (limited only by software and computing time), illumination field control during light flicker, and the ability to precisely match the average stimulus power during image acquisition and light flicker stimulation. 

Retinal Imaging Acquisition and Analysis

Each subject was seated at the modified slit lamp biomicroscope with their head stabilized by a forehead and chin rest. Retinal reflectance image sequences were acquired immediately before (i.e., a baseline) and after each light flicker stimulation in a peripapillary region approximately 6 × 6 mm and centered on the ONH. Each sequence contained 14 images (seven at 570 nm and seven at 606 nm) and was acquired within 2 seconds. Light flicker stimulation was provided at 530 nm for 60 seconds at 2, 4, 8, 16, and 30 Hz, administered in the same increasing order for each subject. These frequencies were selected to closely match those produced by the ERG system and facilitate comparison of vascular and neural frequency responses. Thus, over the course of the imaging protocol, five repeated baseline image sequences were acquired, each immediately prior to one of the flicker stimuli. At least 60 seconds of rest with room lights turned on was provided to the subject between successive frequency conditions. This rest time was provided according to the previously reported time course of changes in vascular metrics and blood flow following the cessation of light flicker.^11,17,25 

The image analysis software has been previously described.¹⁴ Briefly, image sequences were manually inspected to remove those with poor illumination, blinks, or saccades. After registering images, mean images at 570 nm and 606 nm were generated. Vessels within the image were segmented using a Frangi vesselness filter.²⁶ The ONH was manually fitted with a circular overlay, and a circumpapillary region of interest extending from 0.5 to 1.5 disc diameters from the periphery of the ONH was generated. Centerlines for the vessels within the region of interest were determined by Euclidean distance transform, and perpendicular intensity profiles along the vessel contour length were generated. Vessel diameters were determined from these intensity profiles using the full width at half maximum. Manual inspection of the vessel boundaries was performed by a trained operator during image analysis, and grossly erroneous diameters were either eliminated or adjusted to ensure accuracy of the automated vessel definitions. Mean arterial D (D_A) and venous D (D_V) were calculated from each image sequence. Optical density ratios (ODRs) were calculated for both arteries and veins as the ratio of optical density at 606 nm to 570 nm. Optical densities were calculated at both wavelengths as the log-transformed ratio of intensity outside the vessel to intensity inside the vessel. Arterial and venous ODR values were independently corrected to remove artificial dependence on D, background pigmentation, and/or overall illumination.²⁷ Overall illumination was measured as the intensity of the vessel background at 570 nm. To the best of our knowledge, this is the first study to correct ODR for overall illumination, which was necessary to eliminate potential effects of changing lighting conditions on the reflectance images acquired over the 10-sequence acquisition protocol (before and after flicker at five frequencies). Vessel SO₂ was determined by linear calibration of the corrected ODR to healthy arterial and venous values from the literature (92.2% and 57.9%, respectively).²⁸ Mean arterial SO₂ (SO_2A) and venous SO₂ (SO_2V) were calculated from each image sequence. OEF was calculated as (SO_2A – SO_2V)/SO_2A, and values ranged from 0 to 1.¹⁴ 

The effect of light flicker stimulation on D_A, D_V, SO_2A, SO_2V, and OEF was assessed as the ratio of values after light flicker to before (termed D_AR, D_VR, SO_2AR, SO_2VR, and OEFR, respectively). A flicker-induced ratio of 1.05, for example, indicates an increase of that metric by 5%. This provides a normalized response for comparison among subjects. The ratio at each frequency was calculated using data from the baseline image sequence acquired immediately before that flicker stimulation; thus, each normalized ratio response precluded any cumulative effects of repeated flicker stimulation. The flicker-induced ratios were averaged among all subjects at each frequency to generate mean frequency response profiles. The mean frequency response profiles were fitted in MATLAB 2015 (MathWorks, Natick, MA, USA) to a second-order polynomial of the form y(x) = p₁x² + p₂x + p₃, where p₁, p₂, and p₃ are polynomial coefficients. This waveform was selected for fitting based on visual inspection of the vasodilatory and ONH blood flow frequency responses from literature.^17,18,20 

ERG Protocol and Data Analysis

The ssPERG was recorded as it primarily reflects the function of inner-retina neurons, including retinal ganglion cells, consistent with the assessment of D, SO₂, and OEF. Recordings were performed in general accordance with the International Society for Clinical Electrophysiology of Vision (ISCEV) PERG standards.²⁹ In brief, the stimulus consisted of a 31° × 31° checkerboard pattern, which corresponds to the ISCEV standard large-field PERG. Given the size and location of the stimulus used in the imaging protocol, we elected to capture the electrophysiological function beyond the macula (ISCEV standard 15° small-field PERGs were not obtained). The checkerboard was comprised of 1.0° × 1.0° white (261.7 cd/m²) and black (4.02 cd/m²) checks generated by a liquid-crystal display (LCD) monitor with luminous intelligence (Diagnosys, Lowell, MA, USA) that corrects for luminance artifacts inherent in LCD displays. The display was controlled with an Espion E3 Visual Electrophysiology System (Diagnosys). The checkerboard reversal rate ranged from 2 to 25 Hz in steps of approximately 0.3 log units (the reversal rate was set by the 74-Hz refresh rate of the display). These frequencies were selected to match the flicker frequencies used in the evaluation of D, SO₂, and OEF as closely as possible, given the constraints of the ssPERG hardware. 

Subjects were optimally refracted for the 28-cm test distance, and the non-tested (right) eye was occluded. ssPERGs were recorded with DTL corneal electrodes (Diagnosys) referenced to the ear and grounded at the forehead with gold cup electrodes. The stimulus duration was approximately 2 seconds, and it contained four to 50 stimulus cycles, depending on the stimulus temporal frequency. A minimum of five responses were obtained and averaged for each stimulus frequency, which were bandpass filtered (1–100 Hz) prior to analysis. Fourier analysis was used to derive the response amplitude and phase. As described elsewhere,^29,30 the ssPERG is defined by the second harmonic component, as a response is elicited by each check contrast reversal. Consistent with previous definitions,³¹ the response noise was defined as the amplitude at frequencies in the Fourier spectrum that neighbor the second harmonic (0.5 Hz above and below the second harmonic stimulus frequency). A criterion response signal-to-noise ratio of 2.82 was used to define responses that were statistically greater than noise, per convention.³¹ For each stimulus frequency, the noise was subtracted from the signal (second harmonic response) for analysis. The response of the fundamental component did not exceed noise at any temporal frequency, supporting the absence of meaningful flash artifacts. 

Statistical Analyses

Effects of vessel diameter, background pigmentation, and overall illumination on ODR values were determined by linear mixed-effects modeling. Equations for corrected ODR values were generated from model parameter estimates. Repeatability of D_A, D_V, SO_2A, and SO_2V was assessed by coefficient of variation between five repeated measurements before light flicker stimulation (i.e., baseline) in each subject. Data from one subject were removed at 30 Hz due to a physiologically implausible negative OEF. To assess reproducibility, D_A, D_V, SO_2A, and SO_2V were measured before light flicker stimulation by two independent observers, and their data were compared by intraclass correlation. The light flicker responses were averaged among subjects to generate mean frequency response profiles. Fitting polynomial coefficients generated for the mean frequency response profiles was used to calculate the frequency for maximal flicker-induced responses in D, SO₂, and OEF. The relationship between the ssPERG log amplitude and the flicker-induced changes in D, SO₂, and OEF was determined by computing Pearson correlation coefficients. All statistical analyses were performed using SPSS Statistics 25 (IBM, Chicago, IL, USA) or RStudio 2024.12.0+467 (R Foundation for Statistical Computing, Vienna, Austria), with statistical significance accepted at P ≤ 0.05. 

Results

ODR Correction, Repeatability, and Reproducibility

Arterial ODR was corrected to eliminate the significant effect of overall illumination (P = 0.014); however, there were no significant effects of D or background pigmentation (P ≥ 0.173). Venous ODR was corrected to eliminate the effects of D, pigmentation, and overall illumination (P ≤ 0.074). Average repeatability values calculated from repeated baseline measurements of D_A, D_V, SO_2A, and SO_2V before light flicker were 4.1%, 3.3%, 4.0%, and 5.2%, respectively (n = 10). The intraclass correlation coefficients (reproducibility) for D_A, D_V, SO_2A, and SO_2V were 0.958, 0.972, 0.578, and 0.822, respectively (n = 10). 

Frequency Responses of D, SO₂, and OEF

The mean (± SEM) frequency response profiles for the ratios are shown in Figure 2, with the values provided in the Table. D_AR and D_VR were greater than unity (i.e., an increase in D_A and D_V) for flicker stimulation above approximately 2 Hz, indicating vasodilation. Additionally, SO_2VR was greater than unity (i.e., an increase in SO_2V) above 2 Hz, whereas OEFR was below unity (i.e., a decrease in OEF) over the same frequency range. SO_2AR was near unity (i.e., no change with flicker), as expected because SO_2A should not change with light flicker stimulation. The frequency response profiles of D_AR, D_VR, SO_2VR, and OEFR were fitted with a second-order polynomial, which provided a good description of the data (adjusted R² values from 0.74 to 0.95). The frequencies that elicited the maximum flicker-induced responses were between 16 and 30 Hz, with the estimated maximum responses from best-fit polynomial functions between 20.8 and 23.4 Hz (Table). 

Figure 2.

 

The mean light-flicker frequency response profiles in 10 normal subjects. (Left) The frequency response profiles of D_AR (red) and D_VR (blue), with color-matched best-fit polynomial curves. (Center) The frequency response profiles of SO_2AR (red) and SO_2VR (blue), with the color-matched best-fit polynomial curve for SO_2VR. (Right) The frequency response profile of OEFR with the best-fit polynomial curve. Error bars indicate SEM, and the horizontal dashed lines indicate the ratio value of unity (i.e., no flicker-induced response).

Table.

 

Mean Flicker-Induced Ratio Values (± SEM) for D_AR, D_VR, SO_2AR, SO_2VR, and OEFR at All Tested Frequencies

 

Frequency Response of Retinal Neurons

Mean ssPERG waveforms for the 10 subjects are shown in Figure 3 (left) for the five stimulus temporal frequencies. Periodic signals are apparent at each frequency, with the exception of the 25-Hz stimulus. The amplitude (Fig. 3, center) and phase (Fig. 3, right) derived by fast Fourier transform of the ssPERG responses are also shown. Each data point represents the mean ± SEM noise-corrected amplitude and phase at each stimulus reversal frequency. The best-fit curve in the center panel represents a second-order polynomial fit, similar to the one that was applied to the flicker-induced D, SO₂, and OEF ratios (Fig. 2). The amplitude function had a bandpass shape, with a peak at 7.3 Hz. The amplitude at the 25-Hz reversal rate did not exceed noise for any subject and was therefore not plotted or included in further analyses. Specifically, the amplitude in the Fourier spectrum at 25 Hz (signal) and at 50 Hz (second harmonic) did not exceed the amplitude of the neighboring frequencies (noise) by at least 2.82×, which is the criterion used to define a signal that is statistically greater than noise.³¹ The phase of the ssPERG response decreased monotonically with increasing reversal frequency. 

Figure 3.

 

ssPERG waveforms (left), amplitudes (center), and phase (right) are shown. Each data point in the center and right panels represents the mean (± SEM) for the 10 subjects.

Relationships Among D, SO₂, OEF, and Neural Responses

Figure 4 shows the relationships between log ssPERG amplitude and each of the flicker-induced ratios. In these plots, each circle represents a single subject at a particular temporal frequency, excluding the highest temporal frequency that did not exceed noise for the ssPERG. There was no significant relationship between ssPERG amplitude and D_AR (Fig. 4A), whereas ssPERG amplitude was marginally correlated with D_VR (Fig. 4B). Likewise, there was no significant relationship between ssPERG amplitude and SO_2AR (Fig. 4C), but ssPERG amplitude was significantly correlated with SO_2VR (Fig. 4D). Figure 4E shows that greater decreases in OEFR were associated with larger ssPERG amplitudes. 

Figure 4.

 

(A–E) Log ssPERG amplitude is plotted as a function of the flicker-induced ratios: D_AR (A), D_VR (B), SO_2AR (C), SO_2VR (D), and OEFR (E). Solid lines represent linear regression fits to the data, and the Pearson correlation value (with associated P value) is provided in each panel.

Discussion

The primary purpose of this study was to develop and apply a new optical imaging system for evaluating the frequency dependence of inner retinal vascular and tissue metrics. Flicker-induced changes in D, SO₂, and OEF were quantified at frequencies up to 30 Hz. A second purpose of this study was to compare the effects of light flicker stimulation among D, SO₂, OEF, and neural function. The primary findings of the study were that (1) D and SO₂ were repeatable to within ∼5% and were reproducible between two independent observers (intraclass correlation coefficients ≥ 0.58); (2) the light-flicker frequency responses of D, SO₂, and OEF were measured simultaneously and described for the first time, to our knowledge; and (3) flicker-induced changes in SO₂ and OEF were correlated with inner retinal neural function. Notably, the repeatability of baseline data is comparable to our previous system, which acquired images before and during light flicker at a single frequency,¹⁴ indicating no substantial cumulative effects of repeated flicker stimulation on measurements. 

The frequency responses of D and blood flow at the ONH have been evaluated previously. Polak and colleagues¹⁷ reported 2% to 5% arterial vasodilation during light flicker stimulation from 2 to 64 Hz and 0.5% to 2% venous vasodilation from 4 to 40 Hz. Falsini et al.¹⁸ reported relative blood flow changes at the ONH for stimuli between 2.4 and 64 Hz, demonstrating peak augmentation at 12 to 20 Hz. Riva et al.¹⁹ demonstrated a similar bandpass function for blood flow at the optic nerve in response to light flicker, with an apparent peak around 20 Hz. From the best-fit polynomial functions, the maximum vasodilatory response was estimated to occur at approximately 20 Hz, in general agreement with these prior studies. Moreover, these results are consistent with findings in animal models where blood flow at the ONH demonstrated a broad bandpass shape from 5 to 40 Hz.^21,23,32 Of note, there was minimal vasodilatory response at lower frequencies, which may be attributed to measurements being performed after the cessation of flicker, rather than during flicker. In addition, the luminance and wavelength of the flicker stimulus used in the present study differed from those used in previous studies. Nevertheless, previous work combined with the present study indicates that retinal flicker stimulation dilates retinal arteries and veins, permitting increased blood flow. 

We have previously reported changes in SO₂ and OEF with light-flicker stimulation at 10 Hz.¹⁴ As vessels dilate and blood flow is augmented during light flicker, the increased oxygen delivery is expected to outpace the augmentation to the inner retinal oxygen metabolism, driving down OEF and leaving more oxygen present in the veins. To the best of our knowledge, this is the first time SO_2VR and OEFR have been examined at different flicker frequencies. Across most flicker frequencies, SO_2VR was greater than unity and OEFR was lower than unity, consistent with prior findings at 10 Hz.^11,14 There were minimal flicker-induced changes in SO_2VR and OEFR at low temporal frequencies, consistent with the lack of vasodilatory responses. These results are generally consistent with a previous study in rodents, which reported changes in vascular oxygen tension at flicker frequencies between 2 and 14 Hz,²² although the human flicker response extended into higher frequencies as well. Moreover, whereas the human response to light flicker stimulation seems to be characterized by an overcompensation of inner retinal oxygen delivery relative to inner retinal oxygen metabolism, in rats at 10-Hz stimulation, the OEF increases according to a decrease in venous oxygenation, thus signaling an apparent undercompensation.³³ These differences may be attributable to variations in neuron and glial cell counts, densities, and distributions between humans and rodents. For example, in rodents with nocturnal vision, rods photoreceptors are dominant and have been shown to predominantly respond at lower flicker frequencies compared to cones.³⁴ Moreover, rodent cortical astrocytes are smaller and have fewer processes than human counterparts,³⁵ which may impair feedforward glial-mediated mechanisms of neurovascular coupling compared to humans.³ Based on polynomial fitting of our human data, the SO_2VR and OEFR maximal responses were estimated to occur around 20 Hz, consistent with blood flow augmentation at the ONH observed in previous studies.¹⁸ 

In addition to D, SO₂, and OEF measurements, neural function was assessed by ssPERG across a range of similar temporal frequencies. The ssPERG amplitude frequency response was well described by a bandpass function, similar to the mean vascular and tissue profiles. The bandpass shape of the ssPERG function is consistent with a previous report.³⁰ However, the stimulus temporal frequency that elicited the largest ssPERG (approximately 7.3 Hz) was considerably lower than the stimulus temporal frequency that elicited the largest vascular responses (approximately 20 Hz). Of note, 7.3 Hz corresponds to a stimulus reversal rate of 14.6 reversals per second, which is close to the frequency that elicited the largest vascular responses. Thus, when considering the second harmonic frequency or expressing the amplitude in terms of reversals per second (2 × stimulus frequency), the ssPERG amplitude and vascular response have approximately similar temporal frequency profiles across the 2- to 30-Hz range. 

Previous work has examined neurovascular coupling by comparing flicker-induced increases in blood flow at the ONH to flicker-induced responses of the neural retina using ERG. For example, Falsini et al.¹⁸ reported that large flicker ERG amplitudes were associated with large flicker-induced increases of blood flow at the ONH. The results of the present study support this stimulus-driven neurovascular coupling, in that large stimulus-induced increases in D_V were associated with large ssPERG amplitudes. In contrast to the approach of Falsini et al.,¹⁸ the present study recorded the ssPERG rather than the flicker ERG. An advantage of the ssPERG is that the response is closely linked with inner-retinal neurons (retinal ganglion cells) that require blood supply from the retinal vasculature that is imaged with our approach. Thus, the current approach is likely to reflect neurovascular coupling of the bipolar and retinal ganglion cells. The cellular origins of the flicker ERG measured by Falsini et al.¹⁸ are complex and depend on the temporal frequency of the flicker stimulus. However, the non-linear (second harmonic) component of the flicker ERG that they reported likely has considerable inner-retina contributions, providing an alternative index of inner-retina neurovascular coupling. Here, we report the first-time associations of flicker-induced responses of SO_2V and OEF with inner retinal neural function. Both SO_2V and OEF reflect oxygen extraction by the retinal tissue for metabolism. An increase in SO_2V (or equivalently, with a constant SO_2A, a decrease in OEF) indicates a surplus of oxygen availability, which is associated with retinal function. Future studies are necessary to determine the effect of retinal disease on neurovascular temporal relationships. 

The present study was limited in that data were derived from relatively few subjects. However, the primary purpose of this study was to report a new methodology, rather than define a normal reference range of flicker-induced changes in D, SO₂, and OEF. Future studies with larger sample sizes will be necessary to define the normal ranges of these values and provide insight into the effect of light flicker stimulation. A second limitation is that the design of the system precluded image acquisition during light flicker stimulation, as opposed to our previous system, which could acquire image sequences during 10-Hz luminance flicker.¹⁴ The time courses of D, SO₂, and OEF returning to pre-flicker levels following the cessation of light flicker at 8 to 10 Hz have been reported to be between 10 and 26 seconds,^11,17 and a recent study indicated that blood flow may remain elevated as long as 60 seconds following flicker.²⁵ Although image sequences were acquired within 2 seconds following the cessation of flicker, our data may underestimate the true flicker-induced changes. Third, vessel diameters were manually inspected and could be eliminated or modified by a trained observer. Although infrequently needed, this intervention introduced a subjective component to the analysis. However, results between two independent observers were highly reproducible, suggesting negligible effects of this intervention between observers. Fourth, ODR measurements were corrected to remove artificial dependence on D, background pigmentation, and overall illumination. A recent study reported that SO₂ also weakly depends on metrics derived from blood velocity measurements during the cardiac cycle.³⁶ Given the current imaging system cannot quantify blood velocity, correction of ODR and SO₂ values for blood velocity was not possible. However, this lack of correction did not likely significantly affect our measurements, as D changes have a much larger effect on blood flow than blood velocity. Nevertheless, future studies incorporating instrumentation for blood velocity measurements may enhance the accuracy of SO₂ measurements and our derived calculations. Fifth, the frequencies of maximal responses to light flicker were estimated from best-fit polynomial functions without empirical data between 16 and 30 Hz. Future studies to evaluate responses at more frequencies are required to substantiate the estimated frequencies for maximal responses. Finally, the ssPERG was elicited by a contrast reversing checkerboard rather than uniform luminance flicker, as used in the D, SO₂, and OEF assessment. Although the ssPERG is more closely linked to retinal ganglion cell function than the diffuse flicker ERG, the stimuli used for the ssPERG and vascular responses differed. Nevertheless, D, SO₂, OEF, and ssPERG arise from inner-retinal sites, which is a strength. 

In conclusion, a novel optical imaging system is described that may be useful for elucidating the frequency response of inner retinal vascular and tissue responses to luminance flicker. This system provides a means to evaluate functional changes of the retina, particularly as they relate to pathologies that impact neurovascular coupling. 

Acknowledgments

The authors thank Andrew Cross for assistance with subject recruitment and Anubhav Pradeep for assistance with statistical analyses. 

Supported by grants from the National Institutes of Health (R01EY026004 to JM; R01EY030115 to MS; P30EY001792 to UIC Department of Ophthalmology and Visual Sciences); and by unrestricted departmental grants from Research to Prevent Blindness (UIC, USC). 

Disclosure: A.E. Felder, None; G. Balasubramanian, None; A. Arroyo, None; J.C. Park, None; M. Shahidi, None; J.J. McAnany, None