In vivo lipofuscin quantification by visible-light optical coherence tomography-based fundus autofluorescence imaging

Summary

We developed visible-light optical coherence tomography-based fundus autofluorescence (VIS-OCT-FAF) for simultaneous FAF and OCT imaging. We previously showed that the ratio of quantified FAF to quantified OCT of the retinal pigment epithelium (RPE), qAF/qOCT, correlates with the A2E amount measured ex vivo. In this study, we further demonstrated that qAF/qOCT linearly correlates with the total RPE lipofuscin fluorophore concentration, confirming it as a true representation of total RPE fluorophores. Fluorescence intensities of lipids, extracted from the RPE-choroid of albino and pigmented rats after VIS-OCT-FAF imaging, were measured and compared with a fluorescence calibration curve of synthetic A2E to establish an intensity equivalence (INEQ) between total RPE fluorophores and A2E concentration. Pigmented rats exhibited higher qAF/qOCT and thus higher lipofuscin content than albinos, which can be revealed in vivo only when the true FAF intensity is quantified. These findings demonstrate that VIS-OCT-FAF enables quantitative in vivo assessment of RPE lipofuscin, expanding FAF clinical applications.

Graphical abstract

Highlights

• •VIS-OCT-FAF enables quantitative assessment of RPE lipofuscin in vivo
• •The ratio qAF/qOCT correlates linearly with total RPE fluorophore concentration
• •INEQ metric converts qAF/qOCT into an equivalent concentration of A2E
• •Pigmented rats show higher qAF/qOCT and INEQ than albinos

Medical imaging; Applied sciences

Introduction

Fundus autofluorescence (FAF) imaging maps the lipofuscin distribution non-invasively in the retinal pigment epithelium (RPE),¹ a single layer of pigmented cells located between the photoreceptors and the Bruch’s membrane playing an essential role for the survival and function of the overlying photoreceptors.^2,3 RPE lipofuscin is a complex aggregate of fluorescent substance accumulating with age and is believed to have pathological significance,^4,5,6 which can be detected with FAF imaging. For example, FAF images were used to classify retinal atrophies using deep leaning neuronal network.⁷

Translating FAF intensities into the quantities of RPE lipofuscin would make a paradigm shift for FAF clinical applications, as it allows quantitative comparison of images, acquired over time, from different patients under different imaging conditions, and with different imaging devices. However, two major technical issues present a challenge for translating FAF intensity to the amounts of lipofuscin. One is to obtain FAF intensities free of imaging device fluctuations, such as fluctuations of intensity of the illuminating light source and detector sensitivity, and signal attenuations caused by the pre-RPE media (the ocular tissues anterior to the RPE, including the cornea, the aqueous humor, the lens, the vitreous, and the neuronal retina) and the RPE melanin. The other major issue is to establish a quantitative relationship between FAF intensity and the quantities of RPE fluorophores so that a given FAF intensity reflects a given amount of RPE lipofuscin.

We have addressed the first issue by developing a novel multimodal FAF imaging technology based on visible light optical coherence tomography (VIS-OCT-FAF) that acquires VIS-OCT and FAF images simultaneously using a single broadband light source.^8,9,10 In VIS-OCT-FAF, the OCT signals from the RPE serve as a reference to compensate light attenuations by pre-RPE media and the RPE melanin. In addition, the OCT signals from a reflection reference target and the fluorescence signals from a fluorescence reference target provide references to compensate for fluctuations of the light-source intensity and the detector sensitivity. The reflection and the fluorescence reference targets are placed in the intermediate retinal imaging plane. Thus, the compensated FAF intensities calculated from the images acquired by the VIS-OCT-FAF represent a measure of the true fluorescence intensities from the RPE lipofuscin. In fact, we have demonstrated that the compensated FAF intensities acquired by VIS-OCT-FAF correlate linearly with the quantities of a major RPE fluorophore A2E measured with liquid chromatography/mass spectrometry (LC/MS) in RPE samples dissected from eyes immediately after VIS-OCT-FAF imaging.¹⁰ Moreover, we found that even though the raw FAF intensities in albino rats were higher than those in pigmented rats, the compensated FAF intensities were significantly higher in pigmented rats than that in albino rats, which were confirmed by direct measurement of A2E with LC/MS.¹⁰

Among the different fluorophores in RPE lipofuscin, A2E is a major fluorophore contributing to FAF and is the first identified and the most studied fluorophore in RPE lipofuscin.^11,12,13 Other fluorophores in RPE lipofuscin are also involved in FAF emission.^13,14,15 Without chemical analysis of RPE samples, it is impossible to identify the different fluorophores and quantify their contributions to FAF. The question is how to assess the amounts of all FAF-contributing fluorophores based on the FAF intensities measured in vivo.

In the present work, we further proved that the compensated FAF intensities, which were calculated as the ratio of the quantified FAF (qAF) over the quantified OCT signals of the RPE (qOCT), qAF/qOCT represents the quantity of total fluorophores in RPE lipofuscin. We also developed a method to convert qAF/qOCT into an equivalent concentration of A2E, named intensity equivalent of A2E, or INEQ, meaning the true fluorescence intensity emitted by all the RPE fluorophores is equivalent to that emitted by a given quantity of A2E. Using this method, a given compensated FAF intensity measured by VIS-OCT-FAF could be represented as the corresponding INEQ.

We imaged the eyes of both pigmented and albino rats in vivo with the VIS-OCT-FAF, and qAF/qOCT was calculated using the simultaneously acquired FAF and VIS-OCT images. RPE lipids were extracted immediately after imaging. The fluorescence intensities of the lipid extracts, dissolved in ethanol, were measured and converted to INEQ unit using a fluorescence calibration curve established with standard solutions of synthesized A2E. The compensated FAF intensities measured in vivo correlated well with the INEQ of the lipid extracts of the corresponding eyes measured ex vivo.

The combination of VIS-OCT-FAF and the INEQ unit makes it possible to quantitatively measure RPE lipofuscin in vivo. This work is an important step to help expand the clinical implementation of FAF for diagnosis, monitoring progression, and evaluating treatment outcomes of degenerative retinal diseases.

Results

The VIS-OCT-FAF system

An upgraded VIS-OCT-FAF system was used in the experiments. It has three subimaging modules as shown in Figure 1, a spectral-domain VIS-OCT with a center wavelength of 480 nm and a bandwidth of 30 nm, a spectral-domain near-infrared (NIR)-OCT with a center wavelength of 830 nm and a bandwidth of 50 nm, and a FAF detection channel in a confocal configuration. The confocal arrangement ensures that fluorescence from other components of the eye, such as the lens, is minimized in the detected FAF signals. The VIS-OCT and the FAF share the same illuminating light source. The VIS-OCT-FAF can simultaneously acquire VIS-OCT, NIR-OCT, and FAF images at a maximum line rate of 70K lines/s, which is determined by the line rate of the line-scan CMOS cameras of the corresponding OCT spectrometers. The VIS-OCT and FAF images are intrinsically spatially registered; one pixel in the FAF image spatially corresponds to one depth scan (A-line) in the three-dimensional VIS-OCT volume scans. All the other parameters of this VIS-OCT-FAF system are the same as that used in our previous studies,¹⁰ including depth resolution of the VIS-OCT (∼6 μm), determined by the center wavelength and bandwidth of the light source, and sensitivity (∼85 dB at a path-length difference of 0.5 mm). In the VIS-OCT-FAF configuration, the NIR-OCT is used mainly for alignment purposes. It helps reduce visible light exposure and discomfort for the subject of imaging, which is important especially for human imaging.

Figure 1.

A schematic of the VIS-OCT-FAF imaging system

The system has three imaging modules, a VIS-OCT system (blue), an NIR-OCT system (red), and a confocal FAF detection module. The VIS and NIR light beams are combined by two dichroic mirrors (DM1 and DM2) in the sample arms of the two OCT systems. The combined light is scanned together by the 2D galvanometer scanner (GM) and delivered to the retina by a pair of lenses (L1 and L2). The VIS- and NIR-OCT signals are detected by two corresponding spectrometers (SPEC-VIS and SPEC-NIR). The fluorescence signal passes through a set of filters (LPF and SPF), is focused into a pinhole (PH), and detected by a photomultiplier tube (PMT). SLD, superluminescent diode; SC, supercontinuum; VBPF, variable band-pass filter; M1–3, mirrors; S1 and 2, slit; FC1 and 2, 2 × 2 fiber coupler; FP1–4, fiber collimator; PC1 and 2, polarization controller; L1–3, lens; LPF, long-pass filter; SPF, short-pass filter.

Reference targets for fluorescence and reflection

Two reference targets were placed in the intermediate retinal imaging plane—a fluorescence reference target and a reflection reference target. The fluorescence reference target was a flat piece of optical-quality epoxy stained with synthetic A2E at a concentration of 0.1 mg/mL. The reflection reference target was a silver-coated ground glass diffuser (DG10-220-P01, Thorlabs). The ground glass diffuser was used to ensure that the reflectivity is not sensitive to the incident angle of the scanning illuminating light. To exclude light from the reference arm of the VIS-OCT when imaging the reference target, the shutter in the reference arm is closed so only the light reflected from the reference target is detected by the spectrometer.

A2E fluorescence calibration curve

We established an A2E calibration curve by measuring the fluorescence intensities of synthetic A2E in solutions with different concentrations. Chemically synthesized and purified A2E was dissolved in ethanol with concentrations of 0 (ethanol only), 0.0156, 0.0313, 0.0625, 0.0125, 0.025, and 0.05 mg/mL. The fluorescence intensities of the solutions were measured with a home-built multiwavelength confocal fluorescence imaging system using 488-nm laser light for excitation. The fluorescence intensity increases linearly as the A2E concentration increases in the range measured (Figure 2). The linear relationship between fluorescence intensity and A2E concentration can be fitted with a straight line, expressed as:

$$f\left(x\right)=63040x+107\text{,}$$ (Equation 1)

where, x is the concentration of A2E in mg/mL and f(x) is the fluorescence intensity in the unit of readings in 16-bit digitization. The small constant offset is caused by the background.

Figure 2.

A2E fluorescence calibration curve

Quantification of the extracted RPE fluorophore in vitro

We imaged 17 eyes of 11 rats, 4 albino Sprague Dawley (SD) and 7 pigmented Long Evans (LE) (as shown in Table 1). After imaging, the rats were euthanized and the eyes were excised. RPE-choroid preparations were collected from each eye, and total lipids were extracted. The lipid extract of each sample was dissolved in 150 μL ethanol, and the fluorescence intensity was measured with the multiwavelength confocal system used for measuring the A2E standard solutions. The fluorescence signals excited by the 488-nm laser were used to calculate the INEQ of A2E using Equation 1. The fluorescence intensities of the lipid samples and the calculated quantities of INEQ are summarized in Table 1.

Table 1.

Animals, fluorescence intensities from RPE lipids, and INEQ concentration

Rat No.	Strains	Age (months)	Eyes imaged	Fluorescence intensity (×103 a.u.)	INEQ (×10−3 mg/mL)
1	SD	5	L	1.173	16.90
			R	1.030	14.64
2	SD	5	R	1.089	15.57
3	SD	8.4	L	0.502	6.27
			R	0.896	12.52
4	SD	19.5	L	1.523	22.46
			R	1.111	15.93
5	LE	2.1	L	1.745	25.98
			R	1.645	24.40
6	LE	2.1	L	1.944	29.14
			R	1.899	28.43
7	LE	5	R	2.471	37.50
8	LE	5	R	2.498	37.93
9	LE	5	R	1.730	25.75
10	LE	5	L	2.798	42.69
11	LE	13.5	L	2.976	45.51
			R	4.248	65.69

Fluorescence intensity is in the unit of readings in 16-bit digitization. The INEQ concentration is calculated using Equation 1.

L, left eye; R, right eye.

In vivo VIS-OCT-FAF imaging

Eyes of albino and pigmented rats were imaged. Among the 22 eyes of 11 rats, 17 were imaged successfully and analyzed and 5 were excluded due to eye lesions that compromised image quality (Table 1). Figure 3 shows representative images from an SD (Figures 3A–3C) and an LE (Figures 3D–3F) rat. Figures 3A and 3D are the qAF images. Figures 3B and 3E are the en face view of the qOCT images of the corresponding rats. The fluorescence reference target is shown at the bottom of both Figures 3A and 3D. The reflection reference target is shown at the bottom of both Figures 3B and 3E.

Figure 3.

Representative qAF and qOCT images of albino and pigmented rats

(A and D) The qAF images were constructed by normalizing the pixel intensities of the raw FAF images to the mean intensity of the fluorescence reference target (A and D, bottom right).

(B and E) The qOCT images are the en face view of the segmented 3D OCT data projected onto the x-y plane and normalized to the mean intensity of the reflection target (B and E, bottom left).

(C and F) The locations of the B-scan OCT images were marked with the green dashed lines in (B and E), respectively. The red line in each OCT B-scan was drawn between the IS/OS of the photoreceptor layer and the RPE as the upper boundary of the slab. The green line was drawn 150 pixels (∼22 μm with an estimated refractive index of 1.4) from the red line down toward the choroid. The qAF image of the albino rat (A) appears brighter than that of the pigmented rat (D).

Scale bars: 80 μm (horizontal) and 20 μm (vertical, for C and F).

The qAF image was obtained by dividing the pixel intensities of the original FAF image, I_FAF, with the averaged pixel intensity of the fluorescence reference target, I_ref-AF (qAF = I_FAF/I_ref-AF). The qAF image of the albino rat (Figure 3A) appears to be brighter than that of the pigmented rats (Figure 3D). The en face qOCT image (Figures 3B and 3E) was calculated by first manually segmenting the RPE layer on each cross-sectional VIS-OCT image (B-scan). To segment the RPE layer, a line was drawn between the IS/OS (the junction between the photoreceptor inner and outer segments) and the RPE as the upper boundary of a slab (the red lines in Figures 3C and 3F). The lower boundary of the slab (the green lines in Figures 3C and 3F) is drawn 150 pixels from the red line toward the choroid (∼22 μm with an estimated refractive index of 1.4 for the retina). After all B-scans were segmented, the mean OCT signal of each A-line in the slab was calculated and divided by the mean signals of the reflection reference target to obtain the qOCT value of one pixel of the en face qOCT image. The green dashed line in qOCT image in Figures 3B and 3E marks the locations of the OCT B-scans shown in Figures 3C or 3F, respectively.

Quantification of FAF images in INEQ

To quantify qAF/qOCT of each eye, the qAF and qOCT values were calculated as the mean of the pixel intensities of the qAF and qOCT images excluding the signals in the areas of blood vessels and the optic disc. The boundaries of the recognizable blood vessels and the optic disc were outlined manually (Figure 4) using custom-developed software.

Figure 4.

Blood vessel and optic disc segmentation

(A) The original qAF image of a rat eye.

(B) The blood vessels and the optic disc were segmented manually.

The qAF, qOCT, and qAF/qOCT for each of the imaged rat eyes were calculated and plotted versus the corresponding fluorescence intensities of total fluorophores in the lipids extracted from the RPE-choroid (Figure 5). Both the qAF (Figure 5A, blue stars) and qOCT (Figure 5A, red circles) values of the pigmented rats appear much lower than those of the albinos. Also, the qAF values are not correlated with the fluorescence intensities of total fluorophores (Figure 5A, higher qAF values do not correspond to higher fluorescence intensities of total fluorophores). In fact, the low qAF points of the pigmented rats (Figure 5A, qAF on the right side of the dashed line) have high fluorescence intensities of total fluorophores as compared with those of albino rats (Figure 5A, qAF on the left side of the dashed line). However, the qAF/qOCT values correlate well with the fluorescence intensities of total fluorophores measured in the total lipids extracted from the same eye (Figure 5B). The qAF/qOCT values and the fluorescence intensities of total fluorophores of the pigmented rats are higher than that of the albino rats, consistent with our previous study showing that the amounts of A2E in the pigmented rats are higher than that in the albinos.¹⁰ The relationship between qAF/qOCT and INEQ can be established using the relationship shown in Figure 5B and Equation 1. The data related to Figure 5 can be found in Table S1.

Figure 5.

The calculated qAF, qOCT, and qAF/qOCT vs. the fluorescence intensity of total RPE fluorophores of each eye

(A) qAF and qOCT vs. fluorescence intensity of total RPE fluorophores. The vertical lines connect the qAF and qOCT of the same eyes.

(B) qAF/qOCT vs. fluorescence intensity of total RPE fluorophores. The dashed line was fitted excluding the two outliers with an R2 value of 0.95. See also Table S1.

Age-related changes in qAF/qOCT and INEQ

An age-related increase in qAF/qOCT and INEQ was observed in pigmented rats (Figure 6), indicating the accumulation of lipofuscin with age, a phenomenon also observed in human subjects.¹⁶ However, no significant increase in qAF/qOCT and INEQ was observed in the albino rats. In fact, there was even a decrease between the 5-month and 8.4-month-old albino rats (Figure 6). The data related to Figure 6 can be found in Table S1.

Figure 6.

Age-related changes in compensated FAF intensities for albino (SD) and pigmented (LE) rats

The qAF/qOCT values in the eyes of pigmented rats are all higher than those in the albino rats. The age-related increase in pigmented rats was rapid, but not in the albino rats. There was even a decrease in qAF/qOCT in the albino rats between 5 and 8.4 months. Dashed lines are trendlines. See also Table S1.

Discussion

We have demonstrated that VIS-OCT-FAF is capable not only of measuring FAF intensity free of imaging device fluctuations and signal attenuations but also of translating the compensated FAF intensities measured in vivo to amounts of RPE lipofuscin fluorophores with the intensity-equivalent of A2E or the INEQ concept. FAF carries information about RPE lipofuscin that could be used to assess the health status and pathological conditions of the retina related to the RPE. How to read the information FAF carries, however, is a technical challenge.

The development of the VIS-OCT-FAF technology made it possible to compensate attenuations caused by the ocular media in the optical path for both the excitation and fluorescence signals, which cannot be measured directly and vary among individuals, such as absorption and scattering by the pre-RPE media and the RPE melanin. In VIS-OCT-FAF, the FAF and VIS-OCT images are simultaneously acquired using the same illuminating light source. We have proved that the compensated FAF, qAF/qOCT, is free of not only the fluctuations of the detector and illuminating light but also the effects of attenuation caused by the ocular media,^9,10 thus representing the true FAF intensities. The qAF/qOCT linearly correlates with the amount of A2E in rat eyes measured directly by LC/MS, as we previously reported.¹⁰ In the current study, we further confirmed that qAF/qOCT linearly correlates with the amount of total RPE lipofuscin fluorophores represented by the fluorescence intensities measured from the lipid extracts of RPE samples (Figure 5B).

We introduced the concept of INEQ to represent the total fluorophores in RPE lipofuscin that contributes to the fluorescence intensities. Using this concept, the compensated FAF intensities qAF/qOCT can be converted to the INEQ (in the unit of concentration, mg/mL) representing total RPE fluorophores and hence the amount of RPE lipofuscin. The reasons we used A2E to generate a standard curve of INEQ were that A2E is a major fluorophore contributing to FAF^11,12,13 and thus the excitation and emission spectra of A2E are similar to those of FAF and more importantly, A2E can be chemically synthesized^17,18 to the quantities that can be weighed accurately. With the INEQ, we can “measure” the amount of RPE fluorophores in vivo and hence the amount of lipofuscin using the VIS-OCT-FAF. Assessment of the quantity of INEQ based on in vivo FAF imaging would facilitate quantitative lipofuscin assessment in patient care. qAF/qOCT and the INEQ are related in Figure 5B together with Equation 1, and both provide true quantification of the RPE lipofuscin.

To calculate qOCT, we segmented the OCT B-scans with a slab thickness of 150 pixels in depth containing both the RPE layer and the choroid. Since the incident light goes through the RPE to reach the choroid and the reflected light from the choroid passes the RPE, the light signal from the choroid detected by the VIS-OCT also carries attenuation information of the RPE melanin and ocular tissues in the optical path. We selected 150 pixels for the slab thickness to fully cover the reflections from the choroid (Figure 4C), analogous to using integrating spheres to measure light transmission of a sample. Light reflected from the choroid in albino animals would also excite fluorophores in the RPE, unlike in the pigmented rats. That is another reason why we have included reflections from the choroid. The values of the calculated qOCT with this segmentation strategy are more stable than segmenting only the RPE layer in our previous study.^8,9,10 However, more studies about slab thickness selection are needed when imaging human subjects.

The power of the VIS-OCT-FAF technology is demonstrated when we measured FAF intensities in vivo, showed the correlation of qAF/qOCT with the fluorescence intensities of the total RPE fluorophores, and calculated the INEQ in pigmented and albino rats. Although the uncompensated measurements of FAF intensities (qAF) in the albino rats are higher than those in the pigmented rats (Figure 5A, qAF, left vs. right), the compensated FAF intensities (qAF/qOCT) are higher (Figure 5B) in the pigmented rats. The higher qAF/qOCT values in the pigmented rats are confirmed by the measured fluorescence intensities of total fluorophores in the corresponding eyes (Figure 5B). The findings of higher amounts of RPE lipofuscin in pigmented rats as compared with albino animals are consistent with not only our previous studies¹⁰ but also studies reported by other groups.^19,20

We have also observed a significant age-related increase in compensated FAF intensities in pigmented rats. Interestingly, no such increase was observed in albino rats. Since wild-type pigmented rodents should have a normal visual system, the accumulation of lipofuscin in pigmented rats means a normal aging process. The absence of a significant age-related increase of lipofuscin in albino rats is likely due to light-mediated photooxidation and photodegradation of bisretinoids in albino animals that lack melanin.¹⁹ Whether there are other mechanisms involved requires more studies. This observation is limited to the age range of the albino animals. The sample size of the albino rats in our experiments is also limited.

In summary, we have demonstrated that the VIS-OCT-FAF technology can quantify the true RPE lipofuscin fluorophores, and therefore the amount of lipofuscin. The development of this technology is an important step toward expanding the use of FAF in clinical research and patient care.

Limitations of the study

We have clearly demonstrated the strength of the VIS-OCT-FAF technology and the intensity-equivalent of A2E, INEQ, in the current study. This study, however, has several limitations. First, validation was performed in rat models and human-specific calibration will be required to account for differences in ocular optics and lipofuscin composition. Second, sample size and age distribution, particularly in albino animals, were limited, which may influence the interpretation of age-related trends such as those shown in Figure 6. Third, the INEQ represents an optical surrogate for total lipofuscin fluorophores rather than direct molecular quantification. Finally, while the attenuation compensation model has been previously validated, it relies on assumptions regarding wavelength-dependent transmission, which may need further investigation.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Shuliang Jiao (shjiao@fiu.edu).

Materials availability

All materials will be shared by the lead contact upon request.

Data and code availability

• •Any data reported in this paper will be shared by the lead contact upon request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported by NIH grants R01EY031492, R01EY026643, and P30EY14801 (Core Grant to the Bascom Palmer Eye Institute) and in part by an unrestricted grant from Research to Prevent Blindness to Bascom Palmer Eye Institute.

Author contributions

S.J. initiated and designed the research; R.Z., Y.L., R.W., and S.J. performed the research and analyzed the data; S.J., R.Z., and R.W. wrote the paper.

Declaration of interests

The authors declare no conflicts of interest related to this article.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
Synthetic A2E	This paper	N/A
Experimental models: Organisms/strains
Rat: Sprague Dawley	Charles River, USA	RRID:RGD_737891
Rat: Long Evens	Charles River, USA	RRID:RGD_2308852
Software and algorithms
MATLAB	MathWorks	RRID:SCR_001622
LabVIEW	National Instruments	RRID:SCR_014325
MATLAB analysis script	This paper	N/A
LabVIEW control script	This paper	N/A
Other
Home built VIS-OCT-FAF imaging system	This paper	N/A

Experimental model and study participant details

Four Sprague Dawley and seven Long Evens rats (male, adult) were used for the study. Animals were euthanized immediately after imaging, and eyes were collected. All animal procedures were carried out in accordance with the ARVO (Association for Research in Vision and Ophthalmology) Statement for the Use of Animals in Ophthalmic and Vision Research and with the guidelines and regulations of Florida International University (FIU). All experimental protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at FIU (Approval No. IACUC-21-042-CR02). All methods are reported in accordance with the ARRIVE guidelines to ensure transparency and reproducibility in animal research.

Method details

The VIS-OCT-FAF imaging system

The VIS-OCT-FAF imaging system for the present work was upgraded from that used in our previous studies^8,9,10 with faster line scan CMOS cameras (OCTOPLUS, Teledyne e2V, France) for the OCT spectrometers, which can acquire OCT images at a maximum line rate of 70 KHz. The configuration of the imaging system remains the same, including the light sources for the single mode optical fiber-based spectral domain VIS-OCT (supercontinuum laser, SuperK Extreme EXB-6, NKT Photonics, Denmark) and NIR-OCT (SLD-371, Superlum), the dichroic mirrors for coupling the two OCT light in the sample arms of the OCT systems (DM1: DMLP505, Thorlabs, DM2: NT43-955, Edmund Optics), the filters and the photomultiplier tube module for detecting the fluorescence signal (LPF, FGL515M, cut-on wavelength: 515 nm, Thorlabs; SPF, FESH0750, Cut-Off Wavelength: 750 nm, Thorlabs; PMT, H10723-20, Hamamatsu), and the lens combination for delivering the illumination light into the eye while collecting the reflected and fluorescence light from the eye for imaging (L1, f = 75 mm, achromatic; L2, Volk lens, 60D).

The light scanning was controlled by the analogue output (AO) channels of a multifunction DAQ card (NI PCIe-6361), which digitized the signal of the PMT at a sampling rate of 2M samples/second using one of its 16-bit analogue input (AI) channels. The sampling clock of the AO was used to synchronize the data acquisition of the FAF and the image acquisition of the OCT. The cameras of the two spectrometers were controlled by two Cameralink image acquisition cards (NI PCIe-1433). With an OCT A-line rate of 40K lines/second, at each scanning point, 50 samples were acquired and average to generate one pixel for the fluorescence image.

The NIR-OCT was designed for aligning the eye with the optics before image acquisition, which is important for imaging human eyes. In animal imaging, the VIS-OCT was also turned on during alignment to optimize the image quality by tunning the focus of the ocular lens and the polarization controller.

Reference targets in the intermediate retinal imaging plane

Two reference targets were used for in vivo imaging, a fluorescence and a reflection reference target. The fluorescence reference target was fabricated by staining optical epoxy (EPO-TRK 301-2, Epoxy Technology, Billerica, MA) with A2E (100 μg/mL). Synthesized A2E was dissolved in epoxy part B (100 μg A2E in 260 μL epoxy part B), and then mixed with 740 μL of epoxy part A. The A2E-epoxy mixture was placed in a mold and cured at 80°C overnight. The hardened A2E-epoxy was demolded and cut to the size of 30 mm × 15 mm x 1 mm.

The reflective reference target was a silver coated ground glass diffuser (DG10-220-P01, Thorlabs). The reflectivity of the ground glass diffuser is relatively stable in the incident angle of −15°–15°. When the illuminating light is scanned on the reference targets, the shutter in the reference arm of the VIS-OCT is closed. Thus, only the reflected light from the two reference targets is detected. When the VIS light power in the sample arm was set to 450 μW before entering the eye, the detected light with the VIS spectrometer is well below the saturation level.

It was a significant improvement using the silver coated ground glass diffuser as a reflection reference target for imaging in vivo. We used an optical quality PMMA slide as the reflection reference in our previous studies and found that the flection of the PMMA slide was sensitive to the incident angle and the focus of the illuminating light. When the PMMA target is in focus and well aligned with the scanning light, the VIS-spectrometer could be saturated.

Animals and in vivo imaging

Sprague Dawley and Long Evens rats (Charles River, USA) were used for in vivo study. All animal procedures were carried out in accordance with the ARVO (Association for Research in Vision and Ophthalmology) Statement for the Use of Animals in Ophthalmic and Vision Research and with the guidelines and regulations of Florida International University (FIU). All experimental protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at FIU (Approval No. IACUC-21-042-CR02). All methods are reported in accordance with the ARRIVE guidelines to ensure transparency and reproducibility in animal research.

For imaging in vivo, an animal was anesthetized by an intraperitoneal injection of a mixture of ketamine (54 mg/kg body weight) and xylazine (6 mg/kg body weight). The eyes were topically anesthetized with 0.5% proparacaine hydrochloride eye drops and their pupils were dilated with 0.5% tropicamide ophthalmic solution. A hard contact lens was placed to cover the cornea, and the animal was secured in an animal mount with 5° of freedom. The alignment of the eye with the optical axis was adjusted by using the animal mount and monitoring images in real time. The system had a real-time display of the NIR- and VIS-OCT images in the horizontal (X axis) and vertical (Y axis) directions using a cross-scan pattern. It also provides either an OCT fundus imaging mode or a 2-dimensinoal FAF imaging mode to help select the imaging area through coordination of all the five-axis of the animal mount. To image rodent eyes, we place the optic disc at the center of the imaging area as a default. When the eye is well aligned and the real-time display showed good imaging quality for both the VIS-OCT and the FAF, the image acquisition is activated.

Calculation of qAF/qOCT

The mathematical model was modified from our previous analysis to accommodate the changes of the reflection reference target. The quantified FAF intensity qAF is calculated by dividing the fluorescent intensity of the fundus (I_FAF) with that of the A2E reference target (I_ref-AF), qAF = I_FAF/I_ref-AF, and can be expressed as¹⁰:

$$qAF=\frac{{\tau }_{pre}^2\left(\lambda \right){\left[1-{\rho }_{pre}\left(\lambda \right)\right]}^2{\xi }_{RPE}\frac{\pi }{4}{\alpha }^2}{{\xi }_{ref-A2E}\frac{\pi }{4}{{\alpha }^{\prime }}^2},{\xi }_{RPE}=C_L{Q}_L{{\epsilon }_{L}d}_{RPE}$$ (Equation 2)

where τ_pre and ρ_pre are transmittance and reflectance of the pre-RPE media. $\frac{\pi }{4}{\alpha }^2$ and $\frac{\pi }{4}{{\alpha }^{\prime }}^2$ are solid angles comprising the light from the retina and fluorescence reference target to the detector, respectively. C_L, Q_L, ε_L, and d_RPE are concentration, quantum yield, extinction coefficient, and the effective thickness of lipofuscin in the RPE layer. ξ_RPEis defined as the fluorescence efficiency of RPE lipofuscin.

The quantified OCT (qOCT) is calculated by dividing the OCT signals from the segmented RPE layer (R_RPE-OCT) with that of the reflection reference target (R_ref-OCT), qOCT = R_RPE-OCT/R_ref-OCT, which can be expressed as¹⁰:

$$qOCT=\frac{{\tau }_{pre}^2\left(\lambda \right){\left[1-{\rho }_{pre}\left(\lambda \right)\right]}^2{\rho }_{RPE}\left(\lambda \right)\frac{\pi }{4}{\alpha }^2}{{\rho }_{ref}\left(\lambda \right)\frac{\pi }{4}{{\alpha }^{\prime }}^2}$$ (Equation 3)

where ρ_RPE and ρ_ref are the reflectance of the RPE layer and the reflection reference target, respectively. We thus have

$$\frac{qAF}{qOCT}=\frac{C_L{Q}_L{{\epsilon }_{L}d}_{RPE}/C_{A2E}{Q}_{A2E}{\epsilon }_{A2E}{d}_{ref-A2E}}{{\rho }_{RPE}/{\rho }_{ref}}$$ (Equation 4)

The equation can be further expressed as

$$\frac{qAF}{qOCT}=K_{A2E}\times \frac{C_L{Q}_L{{\epsilon }_{L}d}_{RPE}}{{\rho }_{RPE}}$$ (Equation 5)

where K_A2E is the calibration factor related to the fluorescent dye A2E in the reference target.

The assumption that the same attenuation factors apply to excitation and emission wavelengths was based on the analysis of the total transmission of the ocular media and was addressed in our previous publication.⁹

Synthesis of A2E

A2E was synthesized by adding 1 g of all-trans retinal to a reaction mixture of 31.3 mL of ethanol, 3.8 mL of ethanolamine, and 4.5 mL of acetic acid. The reaction was carried out at room temperature in the dark for 48 h with gentle rocking, as described previously.^17,18 To purify A2E, the reaction mixture was dissolved in acetonitrile and washed 5 times with hexane and 1 M sodium acetate (1:1). Purified A2E was dried and stored under argon at −20°C in the dark. The purified A2E was used for establishing the fluorescence calibration curve and for making the fluorescence reference target. The emission spectrum of the synthetic A2E was measured and found close to that of lipofuscin extracted from the RPE of human donor eyes.^18,21

RPE-choroid preparation and lipid extraction

Animals were euthanized immediately after imaging, and eyes were collected. To dissect the RPE-choroid, the anterior segment of an eye was removed to obtain an eye cup. The retina was carefully detached and discarded, and the RPE-choroid was collected. Total lipids were extracted from the RPE-choroid samples as described previously.^22,23 The tissue sample from each eye was homogenized with100 μL of H₂O in a Bullet Blender (Next Advance, Troy, NY). The homogenate was then mixed with 100 μL of methanol in the Bullet Blender, followed by adding 100 μL of chloroform (CHCl₃) and mixing again. The lipid-containing chloroform was separated from the rest of the mixture by centrifugation and was transferred to a collection tube. Extraction was repeated 4 times with 100 μL of fresh chloroform added each time. Collected lipids in chloroform from each eye were pooled and dried in a SpeedVac (Savant Instruments, Holbrook, NY). The dried lipids were flushed with argon and stored at −20°C in the dark until use.

Quantification and statistical analysis

Image processing, segmentation, and data analysis were performed using custom scripts written in MATLAB (MathWorks). The data points shown in Figure 2 are the mean values of each measurement. The data shown in Figures 5 and 6 were calculated using the mean values of the pixel intensities (a.u.) within the segmented regions of interest in the images. Linear regression (MATLAB built-in first-degree polynomial model) was performed to build the curves shown in Figures 2, 5, and 6. The strength of the linear correlations was assessed using the coefficient of determination (R²), yielding R² values of 0.99 and 0.89, respectively, in Figures 2 and 5B.

Published: February 3, 2026