Comparison of Geographic Atrophy Measurements Between Blue-Light Heidelberg Standard Field and Green-Light Optos Ultrawide Field Autofluorescence

Abstract

Purpose: This study compared geographic atrophy (GA) measurements in the macula using standard 30° field and ultrawide field (UWF) fundus autofluorescence (FAF) imaging.

Methods: Participants from Age-Related Eye Disease Study 2 (AREDS2) and Optos PEripheral RetinA (OPERA) studies with GA were included for comparison between standard field FAF with Heidelberg Spectralis and Optos 200Tx UWF FAF. Two time points 5 years apart were evaluated. GA area (mm²) was recorded in the macular area for both imaging types and in the peripheral field for UWF.

Results: Of 102 paired images (73 subjects), the mean (SD) baseline GA area was 5.32 (6.36) mm² with standard and 4.79 (5.87) mm² with UWF FAF (P < 0.001). The mean difference between the two modalities was 0.52 mm² (95% confidence interval, –2.41 to 1.37). Progression of GA in 25 eyes over 5 years showed a median annual growth rate of 1.28 mm² (range, 0.02 to 4.7) for standard and 1.34 mm² (range, 0.04 to 5.3) for UWF FAF (P = 0.49).

Conclusions: The measurement of GA is larger on standard than on UWF FAF imaging. The observed difference may be due to image averaging and the use of blue versus green FAF. Similar GA progression with standard and UWF FAF suggests either may be used longitudinally, although not interchangeably. Further investigation is required with updated UWF technology.

Translational Relevance: With the increasing adoption of UWF imaging modalities, this study suggests that Optos UWF FAF may be used longitudinally as an alternative to standard field FAF to monitor GA.

Introduction

Age-related macular degeneration (AMD) is among the leading causes of vision loss in the modern world. In 2019, AMD was estimated to affect more than 18 million people over the age of 40 in the United States.¹ Worldwide, the disease is projected to affect 288 million people by 2040.² While the introduction of intravitreal anti–vascular endothelial growth factor treatments has proven instrumental in the management of neovascular AMD, the nonneovascular or dry form of the disease continues to cause severe vision loss through the development of geographic atrophy (GA). GA, a progressive loss of retinal pigment epithelium, photoreceptors, and choriocapillaris,³ is viewed as an indicator of late-stage AMD. As treatments for GA start to emerge,^4,5 the importance of understanding its pathogenesis and developing techniques to detect and monitor it cannot be understated. 

Fundus autofluorescence (FAF) is a key imaging technique that has been recommended in the assessment and monitoring of GA.⁶ Change in GA area as measured by FAF imaging is considered a primary end point in clinical trials. The current standard of care in FAF, Heidelberg Spectralis (Heidelberg Engineering, Heidelberg, Germany), is a scanning laser ophthalmoscope that offers a 30° view of the retina with an attachment to extend the view to 55°. Spectralis uses a blue excitation wavelength of 488 nm with a 500-nm barrier filter to produce FAF images.⁷ The high-contrast images along with semiautomated methods of GA measurement make Spectralis the preferred imaging modality for monitoring GA in clinical trials. 

A more recent advancement in imaging technology, Optos 200Tx UWF FAF (Optos PLC, Dunfermline, Scotland), offers visualization of up to 200° of the retina in a single image. Due to the expanded field of view and versatility, ultrawide field (UWF) imaging systems are increasingly popular. In contrast to Spectralis, Optos uses green excitation wavelengths of 532 nm and 633 nm coupled with a barrier filter of 540 nm.⁷ While GA manifests as hypoautofluorescence in both imaging modalities, it is currently unknown if there are differences in measurements between the two systems. Primarily, this study serves as a comparison of commercially available blue versus green light FAF systems. However, beyond field of view and wavelength, several imaging system additional factors may influence the accuracy of GA measurements.⁸ These include image resolution, image-processing algorithms, and the sensitivity of each system to FAF signals, among others. Variations in these parameters can significantly impact the detection and quantification of GA areas, suggesting a need for careful consideration when comparing data obtained from different FAF imaging technologies. 

Limited cross-sectional studies are available to compare measurements of GA area between UWF FAF and standard FAF. This study aims to systematically compare GA measurements using Spectralis and UWF Optos FAF imaging. In addition, the progression of GA over time will also be compared between the two camera modalities. 

Methods

Study Design

This study performed post hoc imaging analysis of Age-Related Eye Disease Study 2 (AREDS2) and Optos PEripheral RetinA (OPERA) participants with GA on FAF imaging. 

The AREDS2 study, which has previously been described in detail,⁹ was a multicenter, randomized trial designed to evaluate the safety and efficacy of nutritional supplements in reducing the risk of advanced AMD development. The study enrolled 4203 participants between the ages of 50 and 85 years with bilateral large drusen or unilateral advanced AMD. The AREDS2 FAF study¹⁰ obtained annual FAF images from a subset of these subjects at participating clinical sites as equipment became available. At the 5-year conclusion of the AREDS2 study, participants from certain clinical sites received additional imaging, including ultrawide field autofluorescence, as part of the OPERA study.¹¹ The AREDS2 10-year follow-up study¹² included a subgroup of participants who returned for an in-clinic visit at the 10-year time point, which included repeat ultrawide field and standard FAF imaging. 

For the purpose of this analysis, participants with paired standard and UWF FAF images at the 5-year time point were included if images were graded to have definite GA on one or both modalities. Pairs of standard and UWF FAF images at the 10-year time point were included for progression analysis when available. Institutional review board (IRB) approval was obtained at each site, and written informed consent was obtained from all participants. Post hoc image analysis was approved at the University of Wisconsin IRB. The research was conducted under the tenets of the Declaration of Helsinki and complied with the Health Insurance Portability and Accountability Act. 

Imaging and Grading Protocol

Standard 30° blue-light FAF images were obtained with Heidelberg Spectralis, and UWF green-light FAF images were obtained with Optos 200Tx (Fig. 1). Images were evaluated by certified graders at the Wisconsin Reading Center. GA was defined as an area or areas of well-defined hypoautofluorescence with a minimum size of 250 microns in the smallest diameter or area larger than 0.05 mm². All FAF images were calibrated based on the proprietary scale parameter embedded in the images or in the DICOM files. Images were evaluated using manual planimetry where graders manually segmented borders of all the GA lesions. 

Figure 1.

 

Fundus autofluorescence images of participant with geographic atrophy captured with Heidelberg 30° standard field (A) and Optos 200° ultrawide field—square outline represents a 30° field (B).

Graders used calibrated monitors and could adjust the brightness and contrast of each image. GA area (mm²) was recorded in the macular area for both modalities as well as in the peripheral field for UWF. A minimum 1-week grading washout period was included between FAF modalities. A team of two graders participated in the assessment (CPF, TFS), but each image was graded by a single grader. A subset of images was regraded for assessment of intergrader variability. 

Key GA phenotype grading included characterization as subfoveal or foveal sparing and unifocal or multifocal. GA was considered subfoveal when hypoautofluorescence was noted to overlap the center of the macula on the FAF image. Optical coherence tomography scans were unavailable for confirmation of foveal involvement with either modality. The images were also evaluated for the presence of reticular pseudodrusen (RPD) and pattern characterization in the junctional zone of GA. 

Images were considered ungradable if image quality was insufficient to delineate the margins of GA, GA extended beyond standard field image borders, or GA merged with peripapillary atrophy (Fig. 2). Peripheral UWF FAF hypoautofluorescence lesions in a nummular pattern consistent with cobblestone degeneration were not annotated as GA (Fig. 3). For follow-up visit grading, graders had access to previous image annotations. 

Figure 2.

 

Ungradable fundus autofluorescence images: due to geographic atrophy extension beyond standard field (A) with corresponding ultrawide field (B) and confluence with peripapillary atrophy in standard (C) and ultrawide field (D).

Figure 3.

 

Ultrawide field fundus autofluorescence image demonstrating peripheral nummular autofluorescence pattern consistent with cobblestone degeneration (arrow).

Consensus review (CPF, TFS) was performed in cases of >10% difference in measured GA area between modalities and the reason for the difference categorized (Fig. 4). 

Figure 4.

 

Fundus autofluorescence grading review of >10% area difference between standard field (left) and ultrawide field (right) demonstrating no obvious difference (A, B) and image quality and visualization (C, D and E, F).

Results

Of 328 paired standard and UWF images available at the 5-year time point, 113 had GA on standard and 113 on UWF FAF. Six eyes had GA in only one modality, leaving a total of 110 paired images (Table 1). Of these, five were considered ungradable for standard field FAF, two due to GA merging with peripapillary atrophy, two with GA extending beyond standard field, and one with poor image quality. A total of two UWF FAF images were considered ungradable due to GA merging with peripapillary atrophy (Fig. 2). The remaining 102 pairs with measurable GA were included in the 5-year baseline analysis. A total of 27 paired images were available at the 10-year time point, of which 2 were ungradable due to GA extending beyond the standard field. The remaining 25 paired images were included for a 5- to 10-year progression analysis (Fig. 5). 

Table 1.

 

Geographic Atrophy Presence Grading

 

Figure 5.

 

Geographic atrophy progression between 5-year (A, B) and 10-year (C, D) time points observed on standard field (A, C) and ultrawide field (B, D).

Baseline Analysis

Of the 102 paired images (73 subjects) at the 5-year time point, the mean (SD) baseline GA area was 5.32 (6.36) mm² (range, 0.078 to 27.07) with standard and 4.79 (5.87) mm² (range, 0.057 to 26.40) with UWF FAF (P < 0.001). The mean difference between the two modalities was 0.52 mm² (95% confidence interval [CI], –2.41 to 1.37) (Fig. 6). GA was subfoveal in 35 (34%) of standard and 35 (34%) of UWF FAF images, with 98% exact agreement between modalities (κ = 0.96). Although this reflects a high level of agreement between modalities, OCT was not available for confirmation of foveal involvement, and results should therefore be interpreted with potential limitation. A subanalysis of extrafoveal GA showed a mean difference of 0.32 mm². GA was multifocal in 69 (68%) of standard and 66 (65%) of UWF FAF images, with 97% exact agreement between modalities (κ = 0.93). RPD was present in 37 (36%) of standard and 14 (14%) of UWF FAF images with an exact agreement of 75% (κ = 0.39). GA junctional zone pattern assessment included no pattern, focal/patchy, or banded/diffuse and showed an exact agreement of 80% (κ = 0.61) between the two modalities. The region outside the standard field showed GA in 3 (3%) images on UWF FAF with a mean (SD) area of 0.53 (0.55) mm² (Supplementary Fig. S1). Cobblestone degeneration was present in 16 (16%) images on UWF FAF (Table 2). 

Figure 6.

 

Bland–Altman plots: difference in GA measured on FAF between standard field and UWF FAF at (A) 5-year time point (n = 102 eyes) and (B) 10-year time point (n = 25 eyes). Grader reproducibility in GA measured on FAF with (C) standard field and (D) UWF.

Table 2.

 

Fundus Autofluorescence Features

 

Progression Analysis

The mean (SD) baseline area of 25 eyes in the progression analysis between 5- and 10-year time points was 5.54 (7.48) mm² with standard and 4.95 (6.57) mm² with UWF. The median annual growth rate was 1.28 mm² (range, 0.02 to 4.7) for standard and 1.34 mm² (range, 0.04 to 5.3) for UWF FAF (P = 0.49), with a median square root transformation of 1.13 mm (range, 0.14 to 2.2) and 1.16 mm (range, 0.2 to 2.3), respectively (P = 0.68). 

Reproducibility Analysis

Twenty percent of images were regraded for assessment of intergrader variability with a mean difference of 0.1 mm² (95% CI, –2.16 to 1.96) for standard and 0.19 mm² (95% CI, –0.92 to 0.54) for UWF FAF (Fig. 6). 

Grading Review

A review was performed for cases with greater than 10% difference in measured GA area between standard and UWF FAF images. Sixty eyes (58%) from the 5-year time point underwent review with 23 (38%) classified as image quality/visualization and 37 (62%) as no obvious reason for the difference. Sixteen eyes (64%) from the 10-year time point underwent review with 8 (50%) classified as image quality/visualization and 8 (50%) as no obvious reason for difference. 

Discussion

In this study, we performed a comparison of GA measurements from standard blue-light and UWF green-light FAF imaging modalities. We aimed to evaluate the suitability of UWF FAF for use in the measurement and monitoring of GA and to assess the level of agreement between standard and UWF FAF. 

This study demonstrates comparable image quality in standard and UWF FAF images with similar rates of ungradable images, high agreement in subfoveal and multifocal status, fair agreement on RPD status, and substantial agreement of junctional zone pattern. The measured GA area was 0.52 mm² smaller in UWF FAF with an increasing difference with larger area. Rates of annual GA progression, however, were comparable, with a median difference of 0.06 mm² between modalities. The intergrader agreement was excellent within each modality. 

There have been limited comparisons of standard and UWF imaging for the area of GA to date. 

A previous study comparing the two modalities predated the development of calibration and measurement tools in Optos.¹³ Other studies have compared the green versus blue spectrum with images obtained using Spectralis imaging and found varying results. Wolf-Schnurrbusch et al.¹⁴ showed a mean difference of 0.49 mm² with blue light measuring larger, whereas Pfau et al.¹⁵ found negligible difference between the two. A recent study by Abbasgholizadeh et al.¹⁶ compared measurement of GA using Heidelberg standard FAF and Optos UWF FAF at a single time point using both manual planimetry and Heidelberg semiautomated annotation software. This group observed significant but small differences in area measurements between the three methods compared, with Heidelberg semiautomated annotation being the smallest, followed by Optos, and largest by Heidelberg using manual planimetry. The mean difference between Heidelberg and Optos using manual planimetry was similar to our study (0.71 mm² vs. our observed 0.52 mm²) with similarly high levels of intergrader agreement. 

Our data suggest that both modalities can reproducibly measure GA and evaluate associated phenotypic parameters. Given the similar growth rate measurements, data from both devices can be pooled in clinical trials for broader analysis. However, the mean difference in area between the two modalities is greater than intergrader variability, suggesting that these devices are not interchangeable, and utilizing a consistent modality is important for accurate longitudinal assessment of a patient's GA progression. Although graders are more accustomed to using Heidelberg FAF, which typically provides a better outline of the GA area, our reproducibility results show tighter limits with Optos than with Heidelberg. Junctional zone patterns described using blue-light Heidelberg FAF were adopted for green light–based Optos FAF.¹⁷ Our study corroborated with other recent publications and found a strong association between patterns observed using Heidelberg Spectralis blue-light FAF and Optos green-light FAF.¹⁸ 

Advantages of Heidelberg Spectralis fundus autofluorescence imaging include high-contrast images and a semiautomated method for GA quantification.¹⁹ Limitations include a field of view that results in an upper GA size limit for study enrollment, typically around 20 mm².²⁰ Optos UWF does not face this constraint, and including it as a viable option for capturing GA measurements allows a potential for increased site participation and a greater range of GA burden in these studies. While Spectralis can be equipped with a 55° attachment to extend its view, it is not the standard configuration used in most studies. 

While 58% of imaging pairs showed a difference of >10% in area, some of the largest differences were in eyes with GA area >18 mm². Such large GA areas are not seen in clinical trial data sets due to inclusion criteria restrictions. The difference observed between modalities may be due to differences in image calibration, image averaging, and visualization between green and blue FAF wavelengths. Overcoming the inherent distortion of ultrawide field images requires complex image calibration utilizing adjustments for participant refraction and corneal curvature.²¹ Intraocular implants have previously served as “intraocular rulers” to test projection models, with Sagong et al.²² finding an average difference between ultrawide field measurement and known intraocular implant size of 10% without correcting for patient axial length. Additionally, Spectralis employs image averaging, which enhances the contrast of hypoautofluorescent areas, making GA lesions appear more pronounced. This high-contrast imaging can lead to a more definitive delineation of GA. Conversely, Optos does not utilize averaging, resulting in some lesions appearing inconclusive or less distinct for GA. The absence of averaging may affect the clarity and contrast of hypoautofluorescent areas, potentially leading to variability in GA measurement. Furthermore, the Optos system uses green autofluorescence, which allows for the clearer visibility of the fovea. This is an advantage in eyes with GA close to the fovea and helps measure the distance between GA and the foveal center. On the other hand, Spectralis uses blue FAF, wherein the blue light is readily absorbed by the macular pigment, making the borders of GA difficult to distinguish from the fovea.¹⁰ This often necessitates multimodal imaging with OCT to accurately delineate the GA margins. Despite the lack of OCT, we did not observe a significant difference in foveal involvement evaluation between the two modalities. 

The difference in detection rates of RPD between the two modalities is significant. This could be due to both green-light penetration and limited resolution of the peripheral retina with Optos. RPD is situated in subretinal space and is most visible using OCT and infrared reflectance or FAF.²³ The deeper penetration of green wavelength might not be as effective in highlighting RPD, as they are superficially located. Limited studies are reporting RPD prevalence from UWF FAF images and have similarly low rates ranging from 5.6% to 15%.^24,25 Most studies reporting peripheral retinal changes have reported on reticular pigmentary change, which is distinct from reticular pseudodrusen.²⁵ Yoon et al.²⁶ demonstrated the feasibility of longitudinal UWF measurements of eyes with RPD using Optos pseudocolor and red-free UWF photography. Over the mean follow-up of 6.8 years, Yoon et al.²⁶ showed an increase in RPD area and an association of large RPD size and rapid increase in RPD area with the development of late AMD. 

Among peripheral findings in our data set, the occurrence of atrophy outside the macula was infrequent, aligning with the broader understanding that such manifestations are rare. Cobblestone degeneration, distinct from GA due to AMD, is relatively common and observed in 16% of our data set. Our protocol specifically differentiates cobblestone degeneration from GA due to AMD, highlighting the importance of careful lesion characterization, especially when utilizing different imaging systems. The OPERA study showed cobblestone degeneration in 12% to 30% of eyes with AMD.¹¹ The Reykjavik study, a population-based study of 1138 eyes in Iceland, found uncategorized peripheral FAF abnormalities in nearly 25%.²⁷ Only a small number of eyes had macular atrophy, and they were not quantified. Other studies have shown cobblestone changes to vary between 16% and 23% of eyes with AMD.²⁸ 

Our study has several limitations that are important to consider. While FAF has been an important modality for monitoring GA progression in clinical trials, OCT is emerging as a key tool for measuring GA area and assessing novel endpoints, such as ellipsoid zone. This retrospective study compared an early version of Optos FAF, and further research is required with more contemporary models and newer UWF FAF modalities. Spectralis images were also obtained in tiff format and not in their native e2e format, which prevents measurement using Region Finder.¹⁹ However, manual planimetry was used on both modalities, making them comparable. Further, this study observed few cases of peripheral GA, which prevents the authors from making any conclusions about the utility of UWF FAF in tracking these lesions. Lastly, graders were not masked to the imaging modality being utilized. Strengths include images obtained by certified photographers from many sites across the United States and evaluation by certified reading center graders with robust reproducibility. 

Conclusions

Measurement of GA area was shown to be larger in Heidelberg standard field versus Optos UWF, which may be related to differences in image averaging and use of blue versus green FAF, respectively. Comparable GA progression measurements between modalities suggest that either may be used longitudinally in clinical trials, although not interchangeably. While most features associated with GA are comparable between the two modalities, RPD is less visible on Optos images. Given advancements in FAF technology, further investigation is indicated to compare the measurement and monitoring of GA using the latest modalities. 

Acknowledgments

The authors thank Rick Voland at the Wisconsin Reading Center for statistical support. 

Supported in part by an Unrestricted Grant from Research to Prevent Blindness, Inc. to the UW-Madison Department of Ophthalmology and Visual Sciences. 

This article was presented at the Association for Research in Vision and Ophthalmology Annual Meeting, May 5, 2024. 

Disclosure: C.P. Froines, None; T.F. Saunders, None; J.A. Heathcote, None; J.W. Pak, None; E.Y. Chew, None; B.A. Blodi, None; A. Domalpally, None