Geographic atrophy: When lesion borders cannot be graded

The availability of FDA-cleared complement inhibitors for geographic atrophy (GA) has increased the clinical importance of reliably identifying and measuring GA lesion size over time. Although blue-light fundus autofluorescence (FAF) has been the traditional standard for GA monitoring, ultrawidefield (UWF) green-light FAF (GAF) is increasingly used in routine retina practice. A retrospective study published in Graefe’s Archive for Clinical and Experimental Ophthalmology by Abbasgholizadeh and colleagues at the Doheny Eye Institute and the David Geffen School of Medicine at the University of California, Los Angeles, examines how often GA lesion borders cannot be reliably delineated on UWF GAF imaging and identifies the factors associated with ungradability.¹

Study design and methods

This retrospective, cross-sectional study included 88 patients (138 eyes) with GA secondary to nonexudative age-related macular degeneration (AMD), identified from electronic health records at the Doheny–UCLA Eye Centers between 2019 and 2025. Patients were required to have concurrent UWF GAF, UWF pseudocolor, and OCT imaging at the same visit. Eyes with exudative AMD, non-AMD causes of atrophy, or significant media opacity were excluded.¹

Two independent, masked, certified Doheny Image Reading and Research Lab (DIRRL) graders classified UWF GAF images as gradable (clearly delineated GA borders amenable to measurement) or ungradable (indistinct borders). Discrepancies were adjudicated by the senior investigator. Choroidal thickness (CT) was measured on central horizontal OCT B-scan at the foveal center and at points 1.5 mm nasal and temporal to the fovea.¹

Results: Ungradability rate and choroidal thickness

UWF GAF images were deemed ungradable for GA size in 7 of 138 eyes (5.1%; 95% CI, 2.5%–10.1%). At the patient level, 6 of 88 patients (8.0%; 95% CI, 3.2%–14.1%) had at least 1 ungradable eye. Inter-reader agreement for gradability was excellent (κ = 0.867; P < .001). In the DICOM subset, inter-grader reliability for GA area measurement was near-perfect (ICC = 0.995 for single measures; 0.998 for average measures).¹

The most striking finding was that ungradable eyes had markedly thinner choroids at all measurement locations compared with gradable eyes. Mean subfoveal CT was 74.71 ± 37.04 μm in ungradable eyes versus 177.41 ± 91.16 μm in gradable eyes (P = .0002). The difference was similarly significant at the nasal point (58.14 ± 14.73 μm vs 132.70 ± 88.28 μm; P = .0013) and temporal point (85.29 ± 34.54 μm vs 177.40 ± 87.07 μm; P = .018).¹

Additional trends were observed, though they did not reach statistical significance given the small number of ungradable eyes. Ungradable images were more frequent in phakic eyes than pseudophakic eyes (16.7% vs. 3.3%; P = .068), and among eyes with foveal GA involvement compared with those without (8.8% vs. 0%; P = .055). Among evaluated OCT biomarkers, none differed significantly between gradable and ungradable eyes, though pigment epithelial detachment (PED) showed a nonsignificant trend toward higher frequency in ungradable eyes (14.3% vs < 1%).¹

Among the 7 ungradable eyes on UWF GAF, 6 showed sharp demarcation of atrophy borders on companion UWF pseudocolor imaging and were deemed gradable for lesion size. The authors note that this reduces the overall proportion of eyes not amenable to GA monitoring using UWF-based imaging to approximately 1%.¹

Discussion

The authors propose that the loss of lesion-border contrast in eyes with very thin choroids is likely driven by green autofluorescence from the sclera, which reaches the detector with less attenuation as choroidal tissue diminishes. The reduced confocality of UWF systems compared with conventional field devices is proposed to further compound this effect. The same thin-choroid mechanism that impairs UWF GAF gradability, however, is described as working in favor of UWF pseudocolor assessment: the whitish appearance of the underlying sclera within the bed of atrophy provides enhanced contrast for delineating GA lesion extent on color imaging.¹

The authors compare their findings with the GATHER-1 trial, in which GA gradeability on standard-field blue-light FAF was approximately 90%, and note that their real-world UWF GAF ungradability rate of approximately 5% compares favorably. They also note that the study did not evaluate UWF blue-light FAF, the availability of which on UWF platforms may have additional implications for the scleral autofluorescence contribution and ungradability rate that remain to be studied.¹

Limitations acknowledged by the authors include the retrospective design, the tertiary care academic practice setting, the small number of ungradable eyes limiting subgroup comparisons, and the lack of standardized acquisition timing for choroidal thickness measurements. The authors also note that no specific choroidal thickness threshold for UWF GAF ungradability can be inferred from these data. Future directions identified include evaluation of AI-assisted or semi-automated GA segmentation workflows for UWF GAF and companion UWF pseudocolor images.¹

The authors conclude that UWF GAF imaging supports reliable GA monitoring in the large majority of retina clinic patients, with ungradability confined to a small minority of cases characterized by very thin choroids, and that companion UWF pseudocolor imaging provides a practical and accessible alternative in most such cases.¹

Reference

1. Abbasgholizadeh R, Quarta A, Alhelaly M, et al. Factors impacting the assessment of geographic atrophy on ultra‑widefield fundus autofluorescence imaging. Graefes Arch Clin Exp Ophthalmol. Published online June 11, 2026. doi:10.1007/s00417-026-07314-0