Fluorescence Lifetime Imaging Ophthalmoscopy, Vision, and Chorioretinal Asymmetries in Aging and Age-Related Macular Degeneration: ALSTAR2

Abstract

Purpose: Eyes with age-related macular degeneration (AMD) and some healthy aged eyes exhibit risk-indicating delays in rod-mediated dark adaptation (RMDA) and prolonged long spectral channel (LSC) lifetimes by fluorescence lifetime imaging ophthalmoscopy (FLIO) in the Early Treatment Diabetic Retinopathy Study (ETDRS) outer ring, especially nasally. To learn FLIO's potential for AMD detection, we correlate FLIO to RMDA.

Methods: The ALSTAR2 follow-up cohort underwent FLIO, color fundus photography, two-wavelength autofluorescence (for macular pigment optical density [MPOD]), visual function testing, including RMDA (rod intercept time [RIT]). AMD was staged by the Age-Related Eye Disease Study (AREDS) 9-step at baseline and follow-up. In pseudophakic eyes with high-quality FLIO, mean intensity maps and meridian plots were created. Vision data were analyzed using linear regression and Spearman's r.

Results: Of 155 eyes (155 participants [75 ± 5.0 years; 60.7% female participants]), 67 eyes were healthy, 38 had early (e)AMD, and 50 had intermediate (i)AMD (P = 0.02). LSC lifetimes were longest in iAMD in all ETDRS regions (P < 0.01) and short spectral channel (SSC) lifetimes in inner and outer rings (P < 0.01). The LSC pattern manifested in 65 of 88 AMD eyes and 30 of 67 healthy eyes. Lifetimes were longest on the nasal meridian and shortest on temporal. LSC lifetimes in the inner and outer rings correlated strongly with RIT (r = 0.68). A stable subgroup had short LSC lifetimes and short RIT. SSC correlated weakly with MPOD.

Conclusions: Prolonged lifetimes in AMD exhibit spatial asymmetry, suggesting mechanisms beyond retinal cells and including choroid. Lifetimes correlate with delayed RMDA, potentially indicating risk for AMD onset and early progression. Further research into SSC signal sources is warranted.

Age-related macular degeneration (AMD) is the most common cause of legal blindness, leading to a $50 billion burden annually in the United States alone.¹ By 2040, 288 million affected patients globally are anticipated.² Treatments are approved for late stages of geographic atrophy^3,4 and macular neovascularization.^5–7 Treatment options targeting early disease stages are sought. Beneficial effects on disease progression accrue from the oral supplementation of antioxidants.^8–10 Photobiomodulation is a recently approved mitochondrial rejuvenant.¹¹ As AMD pathophysiology is still being learned, we aim at the functional validation of imaging biomarkers for AMD onset and early progression. 

The Alabama Study on Early Age-Related Macular Degeneration 2 (ALSTAR2) prospectively studies retinal structure and visual function during the transition from aging to AMD, in a 3-year follow-up period.^12,13 The underlying hypothesis is a center-surround model of cone resilience and rod vulnerability uniting deposit-driven AMD progression with a foveal singularity^12,14; this is defined by marked, reciprocal changes in the cone and rod spatial densities within 0.5 mm of a high cone peak.¹⁴ Around this peak, photoreceptors and support cells distribute with nearly radial symmetry, within a 3-mm-diameter high-risk area for AMD.^15,16 A key functional assay is rod mediated dark adaptation (RMDA), measuring the efficiency of retinoid resupply across tissues affected by AMD, at the rim of this area.¹⁷ A 2016 study, found that pathologic RMDA in healthy eyes according to the Age-related Eye Disease (AREDS) 9-step grading system¹⁸ (approximately 25% of the sample) increases risk for AMD onset 2-fold over 3 years.¹⁹ 

Fluorescence lifetime imaging ophthalmoscopy (FLIO) is an innovative autofluorescence-based imaging technique using time-correlated single photon counting to determine how long a retinal fluorophore stays energetically elevated after excitation by short wavelength (473 nm) light.^20–22 Researchers unveiled distinct FLIO distribution patterns across different retinal diseases.^23–32 In eyes with early AMD (eAMD) and intermediate (iAMD), Sauer et al. described a distinct prolongation of lifetimes “between the large vessels with a diameter [of] approximately 3 mm to 6 mm centered on the fovea [fitting] in the area of the [ETDRS] outer ring,^26,33 that is especially prominent nasally” (Fig. 1). This prolongation is also more noticeable in the long spectral channel (LSC) than in the short spectral channel (SSC).^33,34 The LSC pattern manifested in 20 (35%) of 57 clinically healthy aged eyes. Interestingly, this region resembles that in which subretinal drusenoid deposits (SDDs) first appear.^35,36 These extracellular deposits between photoreceptors and retinal pigment epithelium (RPE) associate with especially slow RMDA.^37–41 

Figure 1.

 

Lengthened fluorescence lifetimes in aging retina. Two eyes graded as healthy by the AREDS 9-step scale for color fundus photography are shown. The top row shows FAF, LSC, and SSC images of the left eye of a 77-year-old woman. In LSC and SSC images, the dotted line outlines an annulus roughly coincident with the outer ring of the ETDRS grid (not shown) which is better visible in the LSC image. The Jet256 color scale artificially amplifies the visibility of this pattern, which appears like a lopsided annulus that we herein refer to as a horseshoe. The bottom row shows the left eye of a 75-year-old woman. Blue FAF, blue (488 nm excitation wavelength) fundus autofluorescence; LSC, long spectral channel (560 nm–720 nm); SSC, short spectral channel (498 nm–560 nm).

FLIO is a projection image of all retinal layers. Consequently, its utility as a metabolic imaging technique for the high-risk macula lutea necessitates more information about cellular and subcellular signal sources. Ideally, new data will come from ongoing microscopic studies of tissue samples.⁴² Well-designed observational studies can inform a cellular basis. First, FLIO images of pathologies, in combination with optical coherence tomography (OCT)-based multimodal imaging, reveal cellular sources of FLIO signal in the absence of strongly emitting RPE.^43–45 Second, vision function testing offers the opportunity to distinguish fluorophores associated preferentially with cone- or rod-mediated visual pathways and supporting cells¹⁴ and with known value to AMD assessment, such as RMDA.^46,47 Third, longitudinal studies can help indicate the prognostic value for any aspect of the FLIO signal. 

This report addresses the last two points with participants of the ALSTAR2 follow-up visit, whose AMD status at the baseline visit is known. We test the hypothesis that impaired RMDA correlates with prolonged LSC lifetimes with comprehensive visual function testing and FLIO imaging, including the SSC. We use new analytics to quantify aspects of the AMD-related spatial pattern. As signal sources of SSC lifetimes include xanthophyll carotenoids,^48,49 we also compare SSC to two-wavelength autofluorescence (2WAF) imaging for macular pigment optical density (MPOD) in the same eyes. 

Methods

Patient Selection

Patient enrollment and AMD assessment are detailed in the Supplementary Material and outlined briefly here. Patients were selected from the 3-year follow-up visit of ALSTAR2 (NCT04112667).^12,50 At baseline, participants aged ≥60 years were identified from the comprehensive ophthalmology clinic of the Callahan Eye Hospital Clinic of the Department of Ophthalmology and Visual Sciences at the University of Alabama at Birmingham using electronic medical records to search for International Classification of Diseases, Tenth Revision (ICD-10) codes of eAMD and iAMD. Confounding ocular conditions (except early cataracts), neurological disorders, and conditions preventing informed consent were excluded. At baseline and 3-year follow-up visits, AMD status was determined by color fundus photographs according to the AREDS 9-step classification system¹⁸ by one experienced grader (author M.E.C.). Retinal imaging and visual function testing were conducted on two separate days; FLIO images were captured at follow-up only. 

For inclusion in the current report, FLIO images needed to be evenly illuminated, accurately focused, and lacking vitreous opacity shadows (assessed by author L.G.). All study eyes were pseudophakic to avoid the influence of the autofluorescent natural lens.^51,52 All participants had reliable RMDA data with <30% fixation losses. 

Image Acquisition

After pupil dilation using 1% tropicamide and 2.5% phenylephrine hydrochloride, participants underwent multimodal high-resolution retinal imaging (Heidelberg Engineering, Heidelberg, Germany, except where noted), including OCT (Spectralis HRA + OCT,) near-infrared reflectance imaging (NIR; same device), color fundus photography (CFP; 450+ Carl Zeiss Meditec, Jena, Germany), 2WAF (investigational module for the Spectralis HRA + OCT) for MPOD assessment, and FLIO imaging (details below). 

FLIO Image Capture and Analysis

FLIO images were captured and analyzed on eyes that had RMDA testing at baseline. As detailed in the Supplementary Material, FLIO, based on the Spectralis device, excites retinal autofluorescence using a 473 nm wavelength pulsed laser. Emitted photons are counted within a 30 degrees × 30 degrees retinal field in the SSC (498–560 nm) or the LSC (560–720 nm) using 2 hybrid photomultipliers. Native fluorescence lifetime data were further analyzed using SPCImage version 8.1 (Becker & Hickl GmbH, Berlin, Germany) to approximate the mean fluorescence decay time per pixel. Fitted images were then transferred to ImageJ software,⁵³ where previously described plug-ins were used to extract LSC and SSC lifetimes for each Early Treatment Diabetic Retinopathy Study (ETDRS) subfield. 

Grayscale LSC and SSC images were qualitatively assessed by one experienced grader (author L.G.) for the presence of an asymmetric pattern of long lifetimes that we refer to as a horseshoe. LSC and SSC images were displayed in ImageJ software using a standardized grey scale (see also below), where the lower (black) and upper (white) ends were defined as 100 ps and 700 ps, respectively, by referring to prior literature. Images were then reviewed, paying close attention to the initial description by Sauer and colleagues.^26,33 More specifically, every image was reviewed for the presence of distinctly prolonged lifetimes between the vascular arcades, approximately in the area of the outer ETDRS ring. If such a pattern was identified, artifact-related reasons for lifetime prolongation (e.g. large floaters) were ruled out and no other areas of lengthened lifetimes beyond the horseshoe were identified, the respective image was labeled as showing the pattern. 

Lifetime topography was investigated graphically with meridian plots using the “Grids OCT” plug-in (available from the ImageJ Update Site at https://sites.imagej.net/CreativeComputation/). Lifetimes were collected in 30-degree (clock-hour) wedges from 0.0 mm to 5.0 mm eccentricity. The four cardinal meridians were plotted using MatLab (version 9.5; The MathWorks, Natick, MA, USA; code downloadable at: https://www.mathworks.com/matlabcentral/fileexchange/26311-raacampbell-shadederrorbar). 

2WAF Imaging

As detailed in the Supplementary Material, 2WAF images were captured with the Spectralis device using our published methods^54,55 so that maps of SSC could be directly compared to the distribution of MPOD. 

Map Creation

As detailed in the Supplementary Material, we used our published methods⁵⁶ to visualize LSC lifetimes, SSC lifetimes, and MPOD topography as maps. In brief, we created stacks of all images of each modality separately for healthy, eAMD, or iAMD eyes, and then computed the mean values for each pixel. 

FLIO lifetimes have been conventionally shown as a Jet256 color scale with blue for long and red for short in both channels (see Fig. 1). Pseudo color maps such as these can be misleading,^57,58 in that transitions between one color and another may appear as large differences to the human observer while representing only small differences in underlying numerical values. Therefore, we used a unified 100 to 700 ps grayscale for both channels that was shown as shades of blue for LSC (ImageJ lookup table “Cyan Hot”) and shades of yellow for SSC (ImageJ lookup table “Yellow Hot”). The ImageJ lookup table “Jet256” was used for comparison color-coded maps. 

Visual Function Testing

As detailed in the Supplementary Material, we utilized our published methods^13,50 for separate tests of exclusively rod- (scotopic) or cone-(photopic) vision and of vision mediated by both rods and cones (mesopic). Best-corrected visual acuity (BCVA) and contrast sensitivity (CS) were assessed under photopic and mesopic conditions.^59,60 CS was determined using the Mars chart.^61,62 After dilation, RMDA was assessed using a computerized adaptometer (AdaptDx, LumiThera, Poulsbo, WA, USA) and expressed as rod-intercept time (RIT) in minutes. Scotopic light sensitivity (LS) was tested under dilation and dark adaptation using a microperimeter (S-MAIA; iCare, Fremont, CA, USA). Afterward, mesopic light sensitivity was assessed. 

Statistical Analysis

Demographics, visual function, FLIO, and MPOD values were summarized using means and standard deviations or number and percentages for continuous and categorical data, respectively. Linear regression was used to compare visual functions, LSC and SSC lifetimes, and MPOD values between disease groups. Age-adjusted Spearman correlations were used to assess correlations between LSC and SSC lifetimes and visual function parameters. A P value of ≤ 0.05 (2-sided) was considered statistically significant. All analyses used SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). 

Results

Cohort Composition

A total of 640 eyes of 331 participants were imaged with FLIO (Table 1). Of these images, 596 (93.1%) met image quality standards. To avoid the impact of crystalline lens fluorophores, only pseudophakic eyes (N = 331) underwent further analysis. One hundred fifty-five eyes of 155 patients (mean age 75.0 ± 5 years, 94 female patients, 60.7%) were eligible for this study (see Table 2 for details). Sixty-seven eyes were healthy, 38 eyes had eAMD, and 50 eyes had iAMD. Mean ages of healthy and eAMD participants were similar (74.5 ± 4.7 years versus 74.2 ±5.1 years), whereas participants with iAMD were significantly older (77.5 ± 4.6 years, P < 0.01). 

Table 1.

 

Cohort Composition

 

Table 2.

 

Cohort Characteristics and Visual Functions

 

Vision Testing

Table 2 details differences in visual functions between disease groups. RIT was significantly longer in iAMD eyes (35.00 ± 12.69 minutes) compared to eAMD (19.44 ± 9.70 minutes) and healthy eyes (14.33 ± 4.40 minutes, P < 0.001). Significant differences in the same direction (worst in iAMD, followed by eAMD and then healthy) were found for mesopic visual acuity (P < 0.01), photopic contrast sensitivity (P = 0.03), and scotopic light sensitivity (P < 0.01). Photopic visual acuity, mesopic contrast sensitivity, and mesopic light sensitivity did not differ significantly among disease groups. 

LSC Lifetimes in Aging and AMD

Table 3 compares mean LSC lifetimes of the ETDRS central subfield, inner, and outer rings, in healthy, eAMD, and iAMD eyes. We found significantly longer lifetimes in iAMD compared with other groups in all ETDRS regions (all P < 0.01). This is further supported by Figure 2, which shows maps of LSC lifetimes averaged per pixel for all eyes included in the respective disease group. Comparison of individual panels reveals brighter colors in iAMD eyes (see Fig. 2C) compared to eAMD eyes (see Fig. 2B) and healthy eyes (Fig. 2A). Dark grey dotted lines in Figure 2E and Figure 2F highlight a lopsided annulus resembling a horseshoe. Supplementary Figures S1A, S1B, and S1C show the same maps using the Jet256 color scale, which artificially amplifies the visibility of the horseshoe pattern. Individual LSC images showed this pattern in 95 of 155 total eyes (61.2 %), in 65 of 88 (73.9%) eyes with AMD, and in 30 of 67 healthy eyes (44.7 %), no complete annuli were observed. 

Table 3.

 

Comparison of Fluorescence Lifetime Imaging Ophthalmoscopy Lifetimes Captured in the Long Spectral Channel Between Disease Groups Per ETDRS Subfields

 

Figure 2.

 

Standard retinas by disease group of fluorescence lifetime imaging ophthalmoscopy captured in the long spectral channel. All LSC lifetimes images per disease group are stacked upon one another, using the foveal center and the optic nerve head as a reference. Mean FLIO lifetimes are calculated and displayed on a pixel level for all images providing FLIO data at that pixel. (A) Standard retina of 155 study eyes graded as healthy according to the AREDS 9-step classification system. (B) Standard retina of all eyes graded as eAMD. (C) Standard retina of all eyes graded as iAMD. (D, E, F) Show the same data as A, B, and C. FLIO lifetimes are shortest in healthy eyes, longer in eAMD, and longest in iAMD eyes. The grey dotted lines in E and F indicate a truncated annulus resembling a horseshoe pattern of prolonged LSC lifetimes described by Sauer et al.³³ eAMD, early age-related macular degeneration; FLIO, fluorescence lifetime imaging ophthalmoscopy; iAMD, intermediate age-related macular degeneration; LSC, long spectral channel; ps, picoseconds.

In Figure 3A, the asymmetrical pattern suggested above is captured and quantified by plotting LSC lifetimes along horizontal and vertical meridians. Error bands of the temporal meridian separate at approximately 1.4 mm eccentricity. Error bands of the nasal meridian separate at approximately 2.3 mm eccentricity and then increase sharply beyond 3 mm eccentricity, near the optic nerve head. Lifetimes along the nasal meridian are longest. In contrast, lifetimes along other meridians smoothly decrease, with the shortest lifetimes along the temporal meridian. 

Figure 3.

 

Comparison of FLIO lifetimes along horizontal and vertical meridians between long and short spectral channels. (A) LSC lifetimes along the nasal (blue), superior (green), temporal (black), and inferior (red) meridians centered on the fovea. The lines indicate the lifetimes averaged across all study eyes and shaded bands indicate standard errors. In the LSC, error bands of the temporal meridian separate at approximately 1.4 mm (red arrowhead). Error bands of nasal meridian separate at approximately 2.3 mm eccentricity (blue arrowhead). Lifetimes remain lower on the temporal meridian compared to other meridians, and lifetimes on the nasal meridian lengthen considerably as the optic nerve head is approached. These meridional asymmetries can give rise to an en face impression of a horseshoe pattern. (B) SSC lifetimes along the same meridians. FLIO, fluorescence lifetime imaging ophthalmoscopy; LSC, long spectral channel; ps, picoseconds; SSC, short spectral channel.

Correlation of LSC Lifetimes With Visual Functions

Table 4 shows the relationship of LSC lifetimes with vision function tests. Longer LSC lifetimes overall correlate significantly (P values ranging from < 0.01 to 0.03) with longer RIT (longer RIT meaning worse scotopic vision). This relationship shows subfield- and stage-specific differences, that is, statistically significant in the central subfield of all study eyes (r = 0.51), healthy eyes (r = 0.29), and eAMD eyes (r = 0.44), and not significant for iAMD eyes. A significant, stronger correlation was found in the inner ring for all (r = 0.68), healthy (r = 0.33), eAMD (r = 0.61), and iAMD (r = 0.43) eyes. The correlation was significant and highest in the outer ring for all (r = 0.68), healthy (r = 0.27), eAMD (r = 0.61), and iAMD eyes (r = 0.49). 

Table 4.

 

Correlation Between LSC Lifetimes Averaged Across ETDRS Subfields and Visual Functions Per Disease Group

 

Correlations of LSC lifetimes with RIT (see the bold P values in Table 4) were stronger and more consistent than those of other vision tests. Significant correlations varied in direction and are briefly presented here; for details, see Table 4. A significant negative correlation with scotopic LS was found for all ETDRS regions in all eyes. For mesopic CS, we found a positive correlation (higher CS and longer lifetimes) in the central subfield and inner ring of healthy eyes, but a negative correlation for the inner ring of iAMD eyes. A positive, weaker correlation compared to RIT was found for mesopic VA in the inner ring of iAMD eyes. These discrepant directions of small effects are harder to interpret and diverge from consistently unidirectional correlations found for RIT. No significant correlations were found for photopic VA, photopic CS, or mesopic LS. 

Figure 4 depicts the relation between RIT and LSC lifetimes of all eyes in the ETDRS central subfield, inner ring, and outer ring (see Figs. 4A, 4B, 4C, respectively). Although r values of linear fits indicate a moderate correlation, a subgroup of eyes with short RIT and short LSC lifetimes, almost all healthy, are apparent in the bottom left corner of all panels. Other fits did not improve the correlations or capture this subgroup. 

Figure 4.

 

Association between LSC lifetimes and RIT per ETDRS subfield for healthy, eAMD, and iAMD eyes. (A) Correlation between LSC lifetimes and RIT in the central subfield of the ETDRS grid in all study eyes. Healthy eyes are labeled in light blue, eAMD eyes in blue, and iAMD eyes in light red. The X-axis displays RIT in minutes, and the Y-axis displays LSC lifetimes values in picoseconds. (B) Correlation between LSC lifetimes and RIT in the inner ring of the ETDRS grid. (C) Correlation between LSC lifetimes and RIT in the outer ring of the ETDRS grid. The black line represents the linear fit. The respective equation, R, and R² values are provided in the bottom right corner of each panel. The correlation is strongest for the outer ring, followed by the inner ring and the central subfield. The bottom left corner of each panel shows a group of eyes characterized by slow RIT and low LSC lifetimes. eAMD, early age-related macular degeneration; ETDRS, Early Treatment Diabetic Retinopathy Study; FLIO, fluorescence lifetime imaging ophthalmoscopy; iAMD, intermediate age-related macular degeneration; LSC, long spectral channel; ps, picoseconds; RIT, rod intercept time.

Stable Eyes Show Short LSC Lifetimes and Short RIT

FLIO was performed for the follow-up visit only of this prospective study. Nevertheless, it is possible to use the difference in AREDS grades at baseline and follow-up to gain insight into FLIO's utility in identifying eyes with notable 3-year progression characteristics. Supplementary Figures S2A and S2B show these characteristics of 155 FLIO-imaged eyes and 400 vision-tested eyes, respectively. This analysis shows that most eyes either stay stable or progress by one or more AREDS grades (even within a stage) over 3 years. Very few regress in AREDS grade. Many eyes graded as healthy at baseline (42.5% in Supplementary Fig. S2A and 47.0% in Supplementary Fig. S2B) stay healthy after follow-up. Thus, the progression characteristic of FLIO-imaged eyes is representative of the overall cohort. 

Figure 5 plots LSC lifetimes, RIT, and increment in AREDS grades (–1 for regression up to +4 for progression), with respect to baseline AREDS stage, within ETDRS subfields (see Figs. 5A, 5B, 5C for central subfield, inner, and outer ring, respectively). A subcohort with short RIT and LSC lifetimes is shown predominantly in green, meaning that these mostly healthy eyes did not progress during follow-up. Overall, eyes that progressed minimally or remained stable had shorter RIT and shorter LSC lifetimes at the 3-year follow-up. 

Figure 5.

 

Correlation among LSC FLIO lifetimes and RIT per disease group, ETDRS subfield, and progression during follow-up. The X-axis shows rod intercept time in minutes, and the Y-axis shows mean LSC lifetimes in picoseconds. Eyes graded as healthy at baseline are represented as circles, eAMD eyes as upright triangles, and iAMD eyes as diamonds. The change in AREDS grades during follow-up is color-coded: stable eyes are labeled in green, eyes that progressed by one stage are in yellow, those by two stages are in orange, by three stages are in red, and by four stages are in black. Eyes that regressed by one stage are labeled in grey. The black line indicates the linear fit of the correlation between RIT and LSC lifetimes. Equation, R, and R² values are provided in the bottom right corner. (A) Correlation between RIT and LSC lifetimes in the central subfield of the ETDRS grid. (B) Correlation between RIT and LSC lifetimes in the inner ring. (C) Correlation between RIT and LSC lifetimes in the outer ring. The correlation is strongest in the outer ring, followed by the inner ring and the central subfield. In all panels, a cohort of eyes is visible in the bottom left corner, characterized by short RIT and short LSC lifetimes. Hardly any of those progressed (most are colored in green) during the follow-up period, suggesting that they are subject to protective factors. AREDS, Age-Related Eye Disease Study; CS, central subfield; ETDRS, Early Treatment Diabetic Retinopathy Study; FLIO, fluorescence lifetime imaging ophthalmoscopy; IR, inner ring; LSC, long spectral channel; OR, outer ring; RIT, rod intercept time.

SSC Lifetimes in Aging and AMD

Supplementary Table S1 shows the average SSC lifetimes in the ETDRS central subfield, inner, and outer rings by diagnostic group. Although no significant difference was found among groups in the central subfield, SSC lifetimes were longer in iAMD eyes compared to eAMD eyes and healthy eyes in the inner and outer rings. This finding is supported by Supplementary Figures S3A, S3B, and S3C, showing pixel-level SSC lifetime averages of healthy, eAMD, and iAMD eyes. Brighter images in iAMD represent longer SSC lifetimes compared to eAMD and healthy eyes. The grey dotted line in Supplementary Figure S3F highlights an asymmetrical distribution of lengthened lifetimes, like the horseshoe described above for LSC (see Figs. 2E, 2F). Supplementary Figures S1D, S1E, and S1F show the same maps using the Jet256 color scale, which highlights this asymmetry further. The central subfield appears stable and independent of the disease stage for both channels. Individual SSC FLIO images showed the pattern in 70 of 155 eyes (45.2%) overall, 50 of 88, 56.8%) AMD eyes, and 20 of 67 healthy eyes (29.9%); again, no complete annuli were observed. Figure 3B shows SSC lifetimes along the horizontal and vertical meridians. Like the LSC, the shortest lifetimes are temporal, and the longest are nasal. Like the LSC, error bands for the temporal and nasal meridian separate at approximately 1.4 mm and approximately 2.4 mm, respectively. Lifetimes increase rapidly beyond approximately 3 mm eccentricity nasally, toward the optic nerve head. 

SSC Lifetimes Correlate With Visual Functions

Correlations between SSC lifetimes and visual function are generally weaker than those with LSC lifetimes yet statistically significant (P values ranging from < 0.01 to 0.05), as shown in Supplementary Table S2. We did find significant and positive correlations between SSC lifetimes and RIT for all study eyes and healthy eyes in the central subfield. Correlations were stronger in the inner ring of all eyes and healthy eyes, and stronger still in the outer ring in all eyes but weaker in healthy eyes in the same area. Scotopic LS correlated negatively with SSC FLO in the inner and outer ring. Mesopic LS correlated negatively with SSC lifetimes in intermediate eyes in all ETDRS regions. No correlations were found with mesopic CS, mesopic VA, photopic CS, and photopic VA. 

Supplementary Figure S4 shows the correlation of RIT and SSC lifetimes (compare to Figure 4 for LSC). Note that the linear fit line for the central subfield is almost horizontal, indicating no association between the RIT and SSC lifetimes. The subcohort of eyes with short LSC lifetimes and short RIT, which was easily identifiable in Figure 4, is indistinguishable in the central subfield and barely visible in the inner and outer rings (Supplementary Figure S4A versus Supplementary Figure S4B, S4C, respectively. 

Association of SSC Lifetimes and MPOD

Because SSC lifetimes are strongly associated with MPOD in young adults,^48,49,63 we compared SSC lifetimes to MPOD determined with 2WAF in the same 155 study eyes. Supplementary Table S3 compares MPOD averaged across ETDRS regions, among healthy, eAMD, and iAMD eyes. We confirmed our previous findings, from ALSTAR2 baseline and a similar sample,^54,55 of significantly higher MPOD values in iAMD (20% higher) and eAMD (10% higher) compared to healthy in the central subfield. No significant differences were identified in the inner and outer rings. 

Supplementary Figure S5 shows pixel-level averages of MPOD in arbitrary units of eyes of each diagnostic group (see Supplementary Figs. S5A, S5B, S5C for healthy, eAMD, and iAMD eyes, respectively). Higher MPOD values are visible in iAMD compared with eAMD eyes and healthy eyes in the central subfield, whereas barely any difference is visible in the inner and outer rings. This is inverse to the appearance of SSC lifetimes (Supplementary Fig. S3). We used a scatterplot to further investigate the relation between SSC lifetimes and MPOD (Supplementary Fig. S6). The only significant correlation was found in the central subfield and it was relatively weak (r = –0.20, P = 0.01). No significant association was found between MPOD and LSC lifetimes in other ETDRS subfields (Supplementary Fig. S7) or within the entire grid for either SSC and LSC lifetimes (Supplementary Fig. S8). 

Stable Eyes Show Short SSC Lifetimes and Short RIT

Supplementary Figure S9 shows the correlation of RIT, SSC lifetimes, AREDS stages, and grades (compared to Fig. 5 for LSC). As in Figure 5, a cluster of eyes with stable AREDS grades (colored green) is visible in the bottom left corner of each panel. However, this cohort is more difficult to distinguish than in LSC (see Fig. 5) and is only really separable in Supplementary Figures S9B and S9C. 

Discussion

We present significantly longer LSC and SSC lifetimes in iAMD eyes and eAMD eyes compared with healthy aged eyes. LSC lifetimes were prolonged in all ETDRS regions, and SSC lifetimes in the inner and outer rings only. By investigating the symmetry of lifetimes with meridian plots, we showed that a previously seen annulus, especially for LSC, was better described as a horseshoe, with a closed end of long lifetimes near the optic nerve head and shorter lifetimes temporally. We found a strong positive correlation between RIT and LSC lifetimes (worse vision and longer lifetimes), especially in the outer ring (r = 0.68), coinciding with the horseshoe. For SSC, correlations were similar in direction and topography but weaker (r = 0.48). We identified a cohort of eyes with short LSC lifetimes and RIT who neither developed nor progressed in AMD grade in the 3 years before FLIO assessment. We discuss imaging analysis approaches, AMD pathophysiology, and prospects for identifying at-risk and protected groups at AMD onset. 

We confirm an annular pattern of prolonged FLIO lifetimes in aging and AMD initially described by Sauer et al. and extend with new analytics to show and quantify its asymmetry.³³ This group showed longer lifetimes in AMD eyes compared with healthy aged eyes, and in advanced AMD, compared to earlier stages. This pattern was identified in 35% of healthy eyes in their sample, compared with 44.7% of healthy eyes in ours. This discrepancy may be explained by the older age of our healthy subjects (74.5 ± 4.7 years) compared with theirs (66.5 ± 8.7 years). This interpretation is supported by an age series (n = 97, age ranging from 9 – 85 years) from Sauer et al.⁶⁴ Our healthy participants may be further along the trajectory from aging to AMD than this previous sample. Sauer et al. also noted, without further discussion or follow-up, a similar yet less striking horseshoe pattern in the SSC, which we also saw in our sample. 

Because our underlying hypothesis incorporates specific retinal cells, we used analytic approaches previously applied to anatomic studies of the neurosensory retina, that is, maps and plots along the cardinal meridians, to investigate the annular pattern.^{15,16,65–68} Previous studies on FLIO relied on analytic software requiring a manual determination of regions of interest.^69,70 Yet, many published figures show long lifetimes between the fovea and optic nerve head, indicating an asymmetry between the nasal and temporal side of the macula.^{20,34,44,70–74} In both channels, we found that the nasal meridian has longer lifetimes than the others between the foveal center and approximately 3 mm eccentricity, where lifetimes lengthen prominently near the optic nerve head. The temporal meridian has the shortest lifetimes. This pattern is only partly captured by ETDRS rings, which we use due to strong eccentricity effects in photoreceptor topography. The greater visibility of the horseshoe in LSC over SSC was due to a steeper fall-off outside the nasal sector in LSC (see Fig. 3). 

The radial asymmetry shown herein allows speculation about contributory mechanisms beyond the nearly symmetric distribution of central photoreceptors, on the grounds that related mechanisms should spatially colocalize. Of note, posterior pole choroid is the thinnest near the optic nerve, at all ages.^75–78 This is where SDD first manifests in this cohort, as determined by analyzing 55 degrees NIR images,³⁵ and longer lifetimes have been associated with the presence of SDD.^34,74 Indeed, we initially wondered if the LSC horseshoe in aged healthy eyes was related to the later formation of SDD.^37–41 We speculate that extracellular deposits (drusen, SDD) form where physiologic intercellular transport is high due to numerous cells, for example, under the fovea or rod ring, where supporting microvasculature is insufficient, for example, near the optic nerve head, or a combination. This model posits that an early step in AMD deposit formation is impaired transit across the choriocapillaris and Bruch's membrane under the fovea (leading to soft drusen material)^79–81 that spreads outward over time to involve rod-rich regions. RPE cell bodies may be initially relatively functional judging from maintained thickness in population-based studies.^82,83 We suggest that RPE cells eventually reach a threshold of hypoxia and metabolic insufficiency and become a barrier to extracellular lipid transfer, allowing SDD to form.⁸⁴ LSC lifetimes, attributed to bisretinoid fluorophores in RPE lipofuscin,⁴² may thus capture this metabolic stress before structural changes occur. Because SSC lifetimes are prolonged in the same regions and correlate with delayed RMDA, the same signal sources as LSC, and additional ones may be affected. 

By comparing the FLIO pattern to the underlying distribution of retinal cell populations, we can also speculate about signal sources. Multiple candidates for fluorophores responding to excitation with blue light (473 nm in our device) have been identified (summarized in Table 5). Differentiating signal sources exclusively captured by either LSC or SSC is difficult. The horseshoe is located inside and beyond the ETDRS outer ring,³³ and thus partially overlaps with the elliptical ring of highest densities of rods, which crests at 120,000 to 140,000/mm² just outside the ETDRS grid.^16,35 Rod photoreceptors are the most numerous cell type in the retina, with inner segments full of mitochondrial flavoproteins (see Table 5) that might contribute to SSC signal except in the fovea.^48,49,85 Hence, fluorophores attributable to rods may contribute to the pattern. A layer of abundant RPE mitochondria (approximately 700 per cell),^86,87 may be largely masked by strongly emitting lipofuscin, positioned anteriorly in the light path. Native flavoproteins have long lifetimes of 2 to 3 ns.⁸⁵ We recall that lifetimes depend not only on the fluorophore but also on its surroundings and conformational state. If flavoproteins are bound to another protein, lifetimes can be very short, expanding the range of measurable lifetimes.⁸⁵ 

Table 5.

 

Retinal Signal Sources for Fluorescence Lifetime Imaging Ophthalmoscopy

 

The correlation between FLIO lifetimes and RIT highlights FLIO's potential to indicate eyes at risk for AMD onset and early progression. RMDA is a functional biomarker of this risk.¹⁹ Using ALSTAR2 baseline data, we identified several structural biomarkers that significantly correlate with RIT, at varying strengths: the interdigitation zone discernibility (r = –0.56),⁸⁸ the SDD area (r = 0.27),³⁵ the ellipsoid (r = –0.33) and the interdigitation zone area (r = –0.59),⁸⁹ the interdigitation zone thickness (r = –0.43),⁸⁹ and the choriocapillaris flow deficits over 30 degrees × 30 degrees (r = 0.35) and below the fovea (r = 0.52).^81,90 Compared to these, the correlation presented for LSC lifetimes in this study is stronger, reaching up to 0.68 for the inner and outer rings, with the caveat that FLIO-tested participants are 3 years older than the baseline cohort. Unlike the structural biomarkers, however, FLIO depends on cellular metabolism.^91–94 FLIO can thus be compared to quantitative (short wavelength) autofluorescence, which reports intensities in a perifoveal annulus (qAF8).⁵⁶ In pseudophakic ALSTAR2 participants (n = 346), qAF8 did neither differ between disease stages nor correlate with visual function.⁵⁶ FLIO capturing such differences and correlations despite a smaller sample (n = 155) underscores its potential for assessing metabolic changes associated with incipient pathology before structural changes occur. 

Most eyes with short LSC lifetimes and short RIT were healthy at baseline and did not develop AMD over 3 years. FLIO lifetimes generally lengthen during the AMD disease course.^95,96 In eyes progressing to advanced AMD, drusen exhibit longer lifetimes than stable eyes at baseline.⁹⁷ The combination of short FLIO lifetimes and short RIT may thus indicate protective lifestyle or genetic factors. Consumption of antioxidants^8–10 or the Mediterranean diet, and supplementation with xanthophyll carotenoids lutein and zeaxanthin are beneficial.^98–103 The impact of ARMS2 risk alleles is of special interest due to a strong relationship of this gene with RMDA in healthy eyes.¹⁰⁴ Protective single nucleotide polymorphisms have also been found in CFH, C2, CFB, SKIV2L, SYN3, TIMP3,¹⁰⁵ APOE-ε4,^106,107 PELI3, C3, C9, COL4A3, and APOH genes.¹⁰⁸ Genotyping the cohort, analyzing dietary patterns, assessing lifestyle factors, and investigating systemic disease to address these questions are ongoing. 

We confirmed previous findings of higher MPOD in iAMD and eAMD compared with healthy aged eyes in the ALSTAR2 baseline sample.^54,55 Unexpectedly, we also found only a weak association between MPOD and SSC lifetimes within the central subfield (r = –0.20, P = 0.01). These findings contrast with, but do not necessarily contradict, previous findings of strong correlations in this subfield. Two studies compared MPOD, an imaging measure of xanthophyll carotenoids to FLIO, and found a strong negative correlation between MPOD and lifetimes (higher MPOD, shorter lifetimes; r = –0.61 to –0.76 in SSC and r = –0.66 in LSC, all P < 0.001).^48,49 Both of these prior cohorts were substantially younger than our sample (30.0 ± 6.9 years and 24.1 ± 3.6 years versus 75.0 ± 5.0 years, respectively). They were also more homogenous, having investigated exclusively healthy individuals, whereas we also included AMD eyes. Spectroscopy studies showed that xanthophylls in an aqueous detergent solution emits weak short wavelength autofluorescence and thus could contribute directly to FLIO imaging.⁴⁸ However, the extrapolation of in vitro data to in vivo hypotheses is not straightforward.⁹¹ Finally, additional SSC signal sources, such as mitochondria, may be present in xanthophyll-bearing cells of the central subfield (Müller glia, cones, and rods). 

Strengths of our study include the carefully selected sample of 155 pseudophakic eyes with visual function testing and FLIO imaging including healthy aging, eAMD, and iAMD. Custom review tools, available for download, permitted standardized analysis of LSC and SSC lifetimes and MPOD and quantification of distributional asymmetries. Combining associations with visual function and retinal topography led to insights into possible FLIO signal sources. ALSTAR2 design allowed exploration of FLIO's and RMDA's joint value in characterizing stable eyes. Limitations are an unequal distribution of AMD stages and ages, limited ethnic diversity, and a lack of information about the artificial lenses of pseudophakic eyes. Future analyses should clarify FLIO's utility to predict AMD onset and early development in a longitudinal setting and further elucidate signal sources via cross-sectional OCT imaging and histology. Associating the presence of the horseshoe pattern with structural changes on OCT may reveal further insights into the pattern's origin and its relation to AMD pathology. FLIO will eventually inform OCT via transfer learning. Confirmation that RMDA and FLIO together may identify protected populations will aid genetic studies leading to biologic target discovery and drug development. In conclusion, we report strong correlations between FLIO lifetimes and RMDA, further enhancing FLIO as an in vivo metabolic imaging technology. These findings should be further explored on a detailed topographic level using AO-FLIO^109,110 and OCT-linked FLIO as these modalities become available. 

Acknowledgments

Supported by NIH grants R01EY029595 (C.O. and C.A.C.), R01EY027948 (C.A.C.), P30EY03039 (institutional); Dorsett Davis Discovery Fund (C.O.), Alfreda J. Schueler Trust (C.O.); Deutsche Forschungsgemeinschaft (DFG) Grants GO 4009/1-1 (L.G.), and Ha 4430/5-1 (M.H.), Unrestricted funds to the Department of Ophthalmology and Visual Sciences (UAB) from Research to Prevent Blindness, Inc., and EyeSight Foundation of Alabama. 

Disclosure: L. Goerdt, Novartis Pharma AG (R), Bayer Healthcare AG (R) unrelated to this research, BioEQ/Formycon (C) (outside this project); M.E. Clark, None; T.N. Thomas, None; L. Gao, None; G. McGwin, None; M. Hammer, None; J.N. Crosson, None; K.R. Sloan, None; C. Owsley, is an inventor on the method and apparatus for the detection of impaired dark adaptation used in this research (P), Johnson & Johnson Vision (outside this project) (C); C.A. Curcio, Heidelberg Engineering (F), Genentech/ Hoffman LaRoche (C), Apellis (C), Astellas (C), Boehringer Ingelheim (C), Character Biosciences (C), Osanni (C), Annexon (C), Mobius (C), and Ripple (C) (outside this project)