Optical Assessment of Photoreceptor Function Over the Macula

Abstract

Purpose: The purpose of this study was to develop next-generation functional photoreceptor imaging using ultrahigh-speed swept-source optical coherence tomography (UHS-SS-OCT) and split-spectrum amplitude-decorrelation optoretinography (SSADOR) algorithm. The advancement enables rapid surveying of large retinal areas, promising non-contact, objective, and quantifiable measurements of macular visual function.

Methods: We designed and built a UHS-SS-OCT prototype instrument using a wavelength tunable laser with 1 MHz A-scan rate. The functional scanning protocol records 5 repeated volumes in 3 seconds. A flash pattern selectively exposes the imaged retina area. SSADOR quantifies photoreceptor light response by extracting optical coherence tomography (OCT) signal changes within the photoreceptor outer segment before and after the flash.

Results: The study prospectively enrolled 16 eyes from 8 subjects, demonstrating the ability to measure photoreceptor light response over a record field of view (3 × 3 mm²) with high topographical resolution (approximately 100 µm). The measured SSADOR signal corresponds to the flashed pattern, whose amplitude also correlates with flash strength, showing consistency and reproducibility across subjects.

Conclusions: The integration of high-performance UHS-SS-OCT and SSADOR enables characterizing photoreceptor function over a clinically meaningful field of view, while maintaining a workflow that can be integrated into routine clinical tests and trials. The new approach allows detecting changes in photoreceptor light response with high sensitivity and can detect small focal impairments.

Translational Relevance: This innovative advance can enable us to detect early photoreceptor abnormalities, as well as help to stage and monitor degenerative retinal diseases, potentially providing a surrogate visual function marker for retinal diseases and accelerating therapeutic development through a safe and efficient outcome endpoint.

Introduction

Preserving vision is the primary goal in ophthalmic care. Visual perception begins when light is converted to neuronal signal, a process known as phototransduction that occurs in specialized photoreceptor cells (rods and cones). Photoreceptor damage and death lead to vision impairment and blindness.¹ Therefore, the preservation of photoreceptors’ function is a meaningful objective in ophthalmic care and therapeutic targets. The development of therapeutic strategies require accurately characterizing photoreceptor health for evaluating safety and efficacy. 

Various visual function tests have been developed as surrogate biomarkers for photoreceptor health. Traditionally, best corrected visual acuity (BCVA) has long been used in clinics and trials.² Although BCVA is easy to measure and provides valuable information, it has limitations. Specifically, it primarily assesses a small retinal region, missing impairments beyond the fovea (i.e. only approximately 1.5 degree/0.5 mm of the central area [approximately 6 mm in diameter] responsible for fine central vision).³ Alternative methods, including psychophysiological examinations (e.g. visual field, perimetry/fundus tracked microperimetry,⁴ and contrast/luminance tests^5,6) and electrophysiological assessments (e.g. electroretinogram⁷ and multifocal electroretinogram^8,9), have been developed to address this unmet need. However, challenges remain, including the need for electrode contact, or relying on subjective feedback, which collectively lead to low spatial resolution and high measurement variability. There is a need for the development of noninvasive techniques capable of rapidly, accurately, and objectively assessing visual function with high resolution. 

Optoretinography (ORG) has shown promise to meet this need. ORG relies on intrinsic optical signals originating from changes in photoreceptor morphology and/or optical properties.¹⁰ Measurements are imaging based and there is no electrode contact and minimal discomfort. These photoreceptor intrinsic signals differ from neurovascular coupling, which reflects hemodynamic changes following the activation of bipolar and/or ganglion cells.¹¹ The first demonstration of ORG in a living human eye took place in 2000, using a single-spot densitometer to capture retinal reflectance changes independent of photoreceptor pigment bleaching.¹² Subsequent studies utilized near-infrared fundus cameras and spectral domain optical coherence tomography (SD-OCT) to correlate retina light back-reflection/backscatter alterations with areas of light stimulation.^13,14 More sophisticated techniques utilized optical phases to detect and measure nanometer-scale length changes in the photoreceptor outer segment (OS). Using coherent imaging light, adaptive optics (AO) fundus cameras have observed a chirped sinusoidal oscillation in the reflection intensity of the cone mosaic following a brief light flash.¹⁵ By using AO to track individual photoreceptors and achieve high imaging speeds to resolve the 2π phase ambiguity, state-of-the-art phase-sensitive AO-OCT can directly quantify changes in the optical path length.^16–20 This innovative approach has allowed for the investigation of diminished cone function along the transition zone in retinitis pigmentosa eyes.²⁰ 

Challenges remain that hinder the broader clinical adoption of ORG. The AO instruments are highly complex and not readily accessible in clinical settings.^21,22 Additionally, AO-based imaging often has a small field of view (i.e. limited to the isoplanatic patch around 0.8 degrees),²³ necessitating extensive image montaging to locate and capture most pathologies. Researchers are actively working to address these limitations, exploring approaches like prototype line-scan OCT²⁴ or OCT phase velocity mapping without AO aberration correction.²⁵ However, these approaches either require unconventional OCT setups that are not yet ready for commercialization or have a limited sampling field, such as single B-scans. 

In light of these challenges, we developed next-generation functional photoreceptor imaging capable of a single measurement of field of view of over 3 × 3 mm², the largest reported to date. Using a prototype ultrahigh-speed swept-source OCT (UHS-SS-OCT) and split-spectrum amplitude-decorrelation optoretinography (SSADOR) algorithm, this innovative approach (OCT SSADOR) makes the first instance of observing photoreceptor reactions throughout the entire macula lutea (fovea and parafovea within the Early Treatment Diabetic Retinopathy Study [ETDRS] grid). The split-spectrum approach has been shown to enhance the signal-to-noise ratio.^26,27 This advancement can lead to a paradigm shift in ophthalmic visual examinations, leading to an objective and quantifiable biomarker for macular vision function. Furthermore, it has the potential to measure and monitor the evaluation of retinal responses across diverse retinal areas and diseases, enriching our comprehension of ocular conditions and their manifestations. 

Materials and Methods

UHS-SS-OCT Instrument Design

The schematic diagram of the custom-designed UHS-SS-OCT and integrated retinal flash stimulator is illustrated in Figure 1. The swept-source OCT uses a microelectromechanical system (MEMS) tunable vertical-cavity surface-emitting laser (VCSEL, SVM10G-2201; Thorlabs Inc., USA). Operating at a sweep rate of 1 MHz and with a 50% duty cycle (unidirectional wavelength sweep), this laser emitted light at a center wavelength of 1060 nm with a 6 dB bandwidth of approximately 70 nm, corresponding to an axial resolution of 7.06 µm in air. The average laser output power is 21.3 mW. The OCT interferometer consists of four fiber couplers, which were adapted from our previous studies,^28,29 and optimizes light delivery and imaging efficiency. The incident power on the subject's cornea is 2.5 mW, well below safety standards.³⁰ The sampling arm includes an electrically tunable lens (EL-16-40-TC-NIR; Optotune, Switzerland) to compensate for the subject's refractive error. The system's relay optics were built using off-the-shelf optics, of which the parameters were optimized using ray tracing software (OpticStudio; ANSYS Inc., USA). The beam size on the pupil plane was around 2.5 mm. This results in a theoretical spot size of 9 µm (1/e² beam diameter) on the retinal plane, specifically when imaging an infant with an axial eye length of 17 mm. The other optics’ specifications and mechanical layout of the UHS-SS-OCT prototype can be found in Figs. S1 and S2 in the Supplementary Materials. A fast-steering mirror (OIM5002; Optics In Motion LLC, USA), located at the back focal plane of the ocular lens, allows precise positioning of the incident location on the pupil plane.³¹ The OCT interference fringe was detected by a balanced detector (PDB481C-AC; Thorlabs Inc., USA), filtered (SLP-1200+ and VAT-10+; Mini-Circuits, USA), and digitized using a 12-bit waveform digitizer board (ATS9373; Alazar Technologies, Inc., Canada). 

Figure 1.

 

Experimental setup for image acquisition and stimulation details. (A) Schematic representation of the UHS-SS-OCT system. The red lines represent fiber cables, whereas the blue lines depict electrical cables. VCSEL, vertical-cavity surface-emitting laser; FC1-FC4, fiber coupler with split ratios (FC1 [10:90], FC2 [50:50], FC3 [50:50], and FC4 [25:75]); NC, not connected; BD, balanced detector; C1-C2, collimator; ETL, electrically tunable lens; L1-L7, lenses; M, mirror; BS, beam splitter; F, filter; DM, dichroic mirror; Galvo-X/Galvo-Y, fast/slow axis of galvanometer scanner; FSM, fast steering mirror; NI DAQ, multifunctional data acquisition and control card; PC, polarization controller; GPU, graphics processing unit. (B) Schematic representation of the imaging acquisition and light stimulus sequence. (C) Three-dimensional volume rendering of OCT data. The red box highlights the imaging area capturing the response to the flash pattern. (D) Raster scanning pattern. The solid blue arrows depict fast scans (B-scans), whereas the cyan dashed arrows represent the flyback delay (0.4 ms for each B-scan).

Visual Stimulation and Imaging Protocol

The OCT SSADOR imaging dataset comprised 5 continuously acquired OCT volumes, with a total duration of 0.6 × 5 = 3 seconds. The flash stimulus was delivered at the beginning of the third volume and persisted for 0.2 seconds (see Fig. 1B). Each OCT volume has a raster sampling of 600 × 600 A-scans. We used a unidirectional raster scan with a fast flyback (0.4 ms, or equivalent to 400 A-scans) to avoid scanning hysteresis and reduce decorrelation noise. 

SSADOR Signal Processing

OCT SSADOR measures subtle alterations in photoreceptor backscattering/back reflectance following flash stimulus to assess their functional response. Therefore, it is sensitive to involuntary eye motion and physiological motions, including heartbeat and breathing, leading to elevated background decorrelation noise. To mitigate these challenges, a series of corrective measures were implemented. As a first step, intra-volume horizontal shifts were corrected using adjacent B-scans using cross-correlation. Inter-volume movements were corrected using OCT en face structural image registration, as demonstrated in Figures 2A to 2C. Finally, axial displacements were corrected between registered B-scan pairs. Figure 2D shows representative registration performance. The calculation of split-spectrum amplitude-decorrelation was performed following spatial OCT registration. We empirically selected a four-fold split using a Gaussian window function and short-time Fourier transform to generate four pairs of wavelength-multiplexed OCT volumes. This 4× spectral split corresponds to an increased coherence length of approximately 20 µm, resulting in an axial resolution of approximately 10 µm. Photoreceptor response is retrieved using split-spectrum amplitude decorrelation, integrated over the photoreceptor OS (Figs. 2E to 2G). For a more detailed description of the three-dimensional spatial registration algorithm, readers can refer to our earlier publication.²⁷ 

Figure 2.

 

ORG signal processing. (A) En face OCT images of the second volume (color-coded in red) and fourth volume (color-coded in cyan) were acquired during the imaging session. (B) An exemplary transverse displacement field was utilized for registering the target OCT volume to the reference volume; in this case, the second volume was chosen as the reference. (C) A comparison of OCT projection images before and after spatial registration. (D) Cross-sectional views illustrating the impact of A-scan registration before and after processing. (E) An illustration of segmented ORG signal responses in cross-sectional views. (F) Three-dimensional rendering of flattened B-scans with highlighted regions indicating the occurrence of ORG signal responses. This step visually depicts the locations where ORG signals are observed. (G) Three-dimensional rendering of selected layers covering the outer segment to observe the ORG signal response in detail.

Study Subjects

Eight healthy adult participants were recruited from the Casey Eye Institute at the Oregon Health & Science University (OHSU) for this pilot study. The protocol was approved by the OHSU Institutional Review Board (IRB)/Ethics Committee and conforms to the Declaration of Helsinki. Signed informed consent was obtained before study procedures. Five participants were myopic and had spherical errors ranging between –6.0 and –3.0 diopters. Although undilated imaging is possible by controlling the beam wandering on the pupil plane, the flashing stimulus light caused pupil constriction, resulting in vignetting artifacts. To ensure unobstructed beam passage, four subjects received pupil dilation and cycloplegia to achieve optimal OCT imaging and stimulus delivery. 

Results

Sixteen eyes from eight subjects were included in the study (5 male subjects and 3 female subjects, mean age = 33.5 ± 5.9 years, range = 25 to 43 years). OCT SSADOR imaging began after 5 minutes of dark adaption. Unless otherwise specified, a 0.2 seconds flash, with a retina irradiance of 2.31 mW/cm², bleached 23.2% of cone pigments for OCT SSADOR measurements. In all subjects, we successfully detected photoreceptor responses, which are consistent and reproducible. 

Temporal Analysis of SSADOR Cone Response

The OCT SSADOR protocol continuously acquires 5 OCT volumes, where the flash coincides with the beginning of volume 3 (Figs. 3A to 3E). Direct en face visualization of structural OCT using the same slab for SSADOR (specifically, the layer approximately 15–35 µm above the retinal pigment epithelium [RPE]/Brunch's membrane interface) did not reveal photoreceptor changes following stimuli. The images before (see Figs. 3A, 3B) and after (see Figs. 3C to 3E) the flash appeared identical and did not show the flash pattern. The SSADOR algorithm, however, isolates and amplifies OCT amplitude changes originating from microscopic cone response, revealing an SSADOR decorrelation map corresponding to the patterned stimulus. Figures 3G to 3I depict the spatial-temporal response across all post-flash volumes using the second volume as a reference for SSADOR computing. The raster scan encodes the time axis in the scanning location, therefore providing insights into the temporal response of photoreceptors, assuming they share similar dynamics. The area within the yellow dashed box in Figure 3G, corresponding to about 0.3 seconds following the onset of the flash (partially overlaps with the 0.2-second flash), has a low decorrelation signal. Progressively higher decorrelation signal emerged approximately in the middle of the third volume (approximately 0.3 seconds), reaching its highest level before the fourth volume. The fourth and fifth volumes clearly showed the flash pattern (see Figs. 3H, 3I), with no noticeable distinction between them. The observed temporal response agrees with previous ORG studies on photoreceptor dynamics.³² 

Figure 3.

 

Temporal assessment of ORG response. (A) to (E) En face OCT structural images with the same slab for SSADOR captured during a single imaging session. Within the same imaging session, amplitude decorrelation with the (F) second (before stimulation), (G) third, (H) fourth, and (I) fifth volumes is computed using the second volume as a reference. The red arrows denote variations in backscattered light intensity due to blood flow within the vessels. Brighter regions, indicated by yellow arrowheads, highlight instances of excessive and uncorrectable eye movement between volumes. Scales bars are 300 µm.

Dependency of ORG Response on Stimulus Intensity

The SSADOR cones’ response strongly correlates with stimulus intensity (Fig. 4). We modified the intensity of the patterned flash stimuli, which induced cone opsin bleaching ranging from 2.3% to 25.6% within a single measurement. Using a patterned stimulus, instead of repeated scans, reduced variations from instrument realignment. These results align with prior research; Pandiyan et al. observed similar responses to varying light exposure intensities.²⁰ Averaging the SSADOR decorrelation within each subregion yields a quantitative metric for the mean cone response magnitude. At the lowest bleaching intensity, the response may be close to the noise floor, making it challenging to distinguish “a1” from the background. Mean SSADOR decorrelation increases with stronger flash intensity, within each tested subject (red markers) and in population mean (black circle; Fig. 4C). Mean SSADOR decorrelation and cone pigment bleach level show an exponential relationship, where curve fitting yielded the equation y = −0.0752  × e^−0.0838x + 0.1644 and an R² of 0.96. This relationship is depicted by the blue dashed line in Figure 4C. 

Figure 4.

 

Dependency of ORG response on stimulus intensity. (A) En face OCT structural image indicating the imaging area. The flash pattern in the lower left inset consists of eight regions with varying grayscales. (B) Corresponding ORG response from an adult volunteer with –0.5 diopters. Each region is labeled from “a1” to “a8” based on its intensity. (C) Different red markers represent eight individual measurements from the same subject for all data. The y-axis represents the mean SSADOR decorrelation value. The black markers with error bars indicate the mean of these 8 measurements, with error bars denoting ± 1 standard deviation (SD) from the mean. The blue dashed line represents the exponential curve fitting. (D) Rearrangement of the eight regions based on stimulus intensity, ordered from low to high. (E) Cone pigment bleaching levels for each region. Scales bars are 300 µm.

Exploring the High Topographical Resolution of the ORG

The integration of UHS-SS-OCT with the SSADOR algorithm represents a powerful and versatile platform for precisely identifying the cone response at high topographical resolution. To demonstrate this capability on healthy subjects, we designed flash patterns with intricate features. Figures 5A and 5B demonstrate the ORG response in a flash pattern with a dark background. In Figure 5B, it is evident that the ORG response can resolve fine features (approximately 130 µm), indicated by the black arrows. Meanwhile, Figures 5C to 5F showcase the ORG response to patterns with a bright background. These visuals emphasize the adaptability of our technique in capturing and characterizing the ORG response with exceptional topographical resolution, which holds significant potential for detecting small photoreceptor defects in future clinical applications. The accompanying structural images from the same scan indicate the imaging area, but again do not reveal the cone response. Moreover, displaying the SSADOR decorrelation map using a pseudo color scale helps distinguish subtle variations in photoreceptor activity across the various flash patterns. 

Figure 5.

 

Visualization of ORG responses to complex flash patterns. En face OCT structural images delineating the imaging area are shown in (A), (C), and (E). The corresponding flash patterns are depicted in the lower left inset of (A), (C), and (E). Color-coded ORG responses to distinct flash patterns are presented in (B), (D), and (F). The ORG response effectively resolves fine features (approximately 130 µm), as indicated by the black arrows in (B). These responses were obtained from three adult volunteers, each with different refractive errors: –5.0 diopters (A) and (B), –0.5 diopters (C) and (D), and –4.0 diopters (E) and (F). Scales bars are 300 µm.

Discussion

Functional photoreceptor imaging offers a noninvasive and objective surrogate marker to assess vision.³³ The technology represents a transformative advancement over existing ophthalmic tests, including BCVA, electroretinogram, microperimetry, etc., allowing to survey a large retinal area with high topographical resolution. SSADOR decorrelation amplitude corresponds to stimuli intensity. Ensemble temporal response reveals a rising phase of approximately 0.3 seconds, consistent with prior reports on cone ORG dynamics.^25,27 Our results strongly suggest that the SSADOR signal correlates with the magnitude of cone response. 

The “opto-genesis” of ORG contrast (coined after “electrogenesis” in electroretinogram) deserves further exploration. High-speed AO ORG revealed a brief initial shortening (approximately 5 ms) of the optical path length (OPL) at OS, followed by a prolonged elongation (approximately 1 second).^16,18,34 These studies reported OS length change on the order of hundreds of nanometers, which depends on the stimulus strength. Other studies and ours found that the magnitude (i.e. change in OPL or decorrelation) of the later phase is proportional to stimulus intensity,³⁴ and exhibits consistent reciprocity to equal-energy stimuli.³⁵ The origins of the ORG signal were ruled out as being attributed to thermal expansion and light-induced disk shedding due to their slower dynamics and restricted temperature-induced expansion.^36–38 The potential role of refractive index changes cannot be entirely dismissed. However, extrapolating ex vivo experiments on vertebrate photoreceptors and ion concentration/flow calculations under near-total bleaching conditions suggest that refractive index changes alone cannot fully explain the observed response, which typically extends to hundreds of nanometers.^39–41 Investigation of human cone OS elongation kinetics using AO OCT revealed 2 distinctive exponential components, with time constants of 80 to 90 ms (component I) and 1 to 1.3 seconds (component II).³⁴ The authors tested and rejected several hypotheses based on mathematical models of molecular and biochemical dynamics. The following were deemed most likely and not mutually exclusive.³⁴ Component I is driven by increased osmotic pressure resulting from the hydrolysis of transducing-bound guanosine triphosphate (GTP) to guanosine diphosphate (GDP) and free phosphate. This hydrolysis is catalyzed by G-protein signaling 9 (RGS9). The component II involves chromophore isomerization, hydrolysis of all-trans chromophore, and the subsequent reduction to membrane-localized all-trans retinol, which collectively leads to membrane swelling. 

The clinical management of age-related macular degeneration (AMD) and inherited retinal diseases,⁴² two major causes of retinal degeneration and vision loss, can benefit from monitoring the functional integrity of photoreceptors. For instance, in AMD, the assessment of photoreceptor function helps identify the stage and severity of the disease, allowing for more tailored treatment plans.^42–44 In cone-rod dystrophy and best vitelliform macular dystrophy, detecting and monitoring functional loss is essential for early diagnosis and tracking of disease progression.^45,46 It also aids in identifying potential therapeutic targets. In addition, the ability to monitor photoreceptor function facilitates drug development by providing an endpoint marker that is closely linked to visual perception. It allows researchers to evaluate the safety and efficacy of experimental treatments and pharmaceutical interventions, ultimately expediting the development of novel therapies. 

OCT SSADOR imaging offers the ability to obtain co-registered retinal structural and photoreceptor functions in one scan. Integrating both functional and structural imaging within a unified system holds the potential to offer comprehensive insights into retinal diseases. Such an integrated structural-functional approach has the potential to enhance diagnostic accuracy because structural imaging can offer an anatomic reference for studying functional alterations, a perspective not directly accessible through other ophthalmic tests. Furthermore, this structural-functional correlation also provides an image registration basis in longitudinal follow-up studies. The capability to correlate longitudinal structural and function findings represents a unique advantage over histology and other ophthalmology tests, which may enable a deeper understanding of disease mechanisms and pathophysiology in vivo. 

The primary advantage of SS-OCT in the study over SD-OCT is its significantly higher imaging speeds (≥ 1 MHz A-scan rates versus 200 – 300 kHz). The fast acquisition speed helps to reduce eye motion artifacts and enhance image quality in challenging subjects. Our imaging system's setup facilitates a straightforward transition for other researchers and future higher-performance commercial instruments to adopt the demonstrated technology. This method offers the advantage of current imaging systems with cost-effective components compared to more sophisticated techniques like line-field and full-field OCT.⁴⁷ Unlike phase-sensitive OCT or AO-OCT,^25,47 which limit the measuring field of view due to the necessity of resolving or tracking single photoreceptors, our methodology does not require such resolution or tracking. Our OCT SSADOR approach does not necessitate explicitly measuring OS length changes; instead, it calculates OCT amplitude decorrelation, which captures subtle changes in OS length and reflectivity. Indeed, our results demonstrated that SSADOR decorrelation is correlated with stimulus intensity and, consequently, is expected to correlate with the magnitude of cone response amplitude/OS length change. This observation provides a foundation for future studies on impaired cone function in retinal degeneration. By prioritizing broader coverage, our approach facilitates a comprehensive functional assessment of photoreceptors over the macula. 

Limitations include susceptibility to noise and artifacts, specifically motion artifacts (e.g. from the scanning mechanism and involuntary eye movements) and retinal blood flow projection artifacts. Additionally, the available light source limited the axial resolution to 7.1 µm, making it challenging to distinguish retinal structures between thin layers. Although our results are promising, further research is necessary to refine optical performance, expand image coverage to encompass the entire macular region, conduct additional testing on subjects with and without retinal disease, and consider variations across different age groups in future studies. 

Combining ultrahigh-speed OCT imaging with the SSADOR algorithm leads to an innovative way to investigate photoreceptor function and health, boasting an unprecedented combination of extensive coverage, exceptional topographical resolution, and short testing time. The extraordinary capability represents a significant advancement in ophthalmic vision tests, promising widespread advancements in the care of retinal diseases that include photoreceptor degeneration. 

Acknowledgments

Supported in part by grants R01HD107494, R01EY023285, and P30EY010572 from the National Institutes of Health, an unrestricted departmental funding grant and a Career Advancement Award from Research to Prevent Blindness, and the Bright Focus Foundation. 

Disclosure: S. Ni, None; S. Khan, None; A. Jiménez-Villar, None; M.E. Pennesi, None; D. Huang, Visionix (F, P, R), Genentech (P, R), Intalight (F), Canon (F), Cylite (F). These potential conflicts of interest have been reviewed and managed by OHSU; Y. Jian, None; S. Chen, None