Deep learning in geographic atrophy: rethinking age-related macular degeneration progression and treatment

Geographic atrophy (GA) secondary to dry age-related macular degeneration (AMD) is a major cause of irreversible vision loss and remains a relentlessly progressive condition associated with substantial ocular morbidity (1). Until recently, management strategies were largely limited to visual rehabilitation and monitoring for complications, as no approved therapies were available to slow disease progression. Previous clinical trials exploring therapeutic options failed to achieve their primary outcome which was change in best corrected visual acuity (BCVA) (2-4). The recent regulatory approval of complement inhibitors has begun to shift this therapeutic landscape. The phase III OAKS and DERBY trials demonstrated that inhibition of complement factor C3 can reduce the rate of GA lesion growth (5,6). However, these trials relied on fundus autofluorescence (FAF)-based measurements of retinal pigment epithelium (RPE) loss as the primary outcome and did not demonstrate meaningful benefits in visual acuity or functional endpoints such as microperimetry. Interpretation of functional outcomes was further limited by spatial mismatch between fixed testing grids and the heterogeneous distribution of pre-atrophic and atrophic retinal changes. As complement inhibitors are increasingly adopted into clinical practice, concerns remain regarding the risk of conversion to exudative AMD and the modest functional efficacy observed to date (7). Together, these limitations have intensified the need for more sensitive, biologically relevant endpoints capable of capturing earlier disease activity and meaningful therapeutic response in GA.

In this context, Schmidt-Erfurth and colleagues provide a novel perspective aimed at perceiving GA as a photoreceptor centric disease. Traditionally, GA is perceived and monitored as a RPE centric pathology and disease management and monitoring strategies have relied on RPE morphology. However, the authors change this narrative by using a validated AI algorithm to segment the RPE and ellipsoid zone (EZ). The EZ indicates the integrity of the photoreceptor layer and can be identified on an optical coherence tomography (OCT) as a hyper-reflective band (8). It is increasingly being recognized as a valuable biomarker in various retinal degenerative conditions (9-12).

The authors did a post hoc analysis on images from the phase III OAKS and DERBY trials of pegcetacoplan. Their analyses demonstrate that photoreceptor degeneration, visualized as EZ loss on OCT, consistently precedes and exceeds RPE loss over time. Additionally, by stratifying patients into quartiles based on the magnitude of the EZ-RPE difference at baseline, the authors show that disease progression is faster, and treatment effects are greater in eyes with a larger EZ-RPE dissociation. These findings not only strengthen the biological rationale for complement inhibition but also suggest a promising approach to prognostication in GA, and its use as a meaningful therapeutic goal.

The role of EZ loss in GA

Preservation of EZ loss on OCT has been recently used as a therapeutic endpoint for clinical trials in MacTel2 involving the ciliary neurotrophic factor implant (13,14). This sets an important precedent for its use in other retinal degenerative conditions. The EZ represents the mitochondria-rich inner segments of photoreceptors, has been shown to correlate strongly with retinal sensitivity and visual function across multiple retinal diseases (10,15). This also overcomes the limitations of using FAF based measurements which are two dimensional and present late stage RPE loss rather than earlier pathological changes. The authors illustrate that EZ loss consistently encompasses a larger area than RPE loss at baseline and throughout follow-up. Over time, RPE atrophy appears to “fill in” the pre-existing zone of photoreceptor loss, suggesting a temporal and spatial sequence in which photoreceptor degeneration is an early and primary event.

Histopathological studies of GA lesion borders have demonstrated a transitional zone characterized by photoreceptor degeneration in areas outside the area of RPE atrophy (16). The current study provides in vivo confirmation of this concept, leveraging one of the largest imaging datasets available from successful phase III trials.

A key strength of this study is its use of validated, automated deep learning algorithms to segment EZ and RPE loss on OCT. Manual interpretation of OCT images is inherently limited by interobserver variability, susceptibility to human error, the time required to analyze large datasets, and difficulty in consistently identifying and quantifying subtle longitudinal changes. In contrast, AI-based volumetric OCT analysis enables precise, reproducible enface mapping of photoreceptor and RPE integrity across the macula, facilitating the derivation of the EZ-RPE difference metric.

The large-scale application of such algorithms should be preceded by rigorous validation on existing datasets and the use of high-quality images acquired from a single OCT platform to ensure consistency. When carefully developed and appropriately implemented, artificial intelligence (AI)-based retinal layer segmentation has the potential to be a powerful tool, offering deeper insights into the pathophysiology, monitoring, and treatment response of retinal degenerative diseases, as well as other retinal conditions.

Illustrating the therapeutic effects of pegcetacoplan

The study strengthens the therapeutic efficacy of pegcetacoplan in reducing the rate of progression of GA, as demonstrated by the OAKS and DERBY clinical trials. However, a key insight emerges when therapeutic effects are evaluated separately for RPE loss and EZ loss. While pegcetacoplan reduced RPE loss growth by approximately 20–27% compared with sham, the reduction in EZ loss was substantially greater, reaching 45–55% across trials and dosing regimens (8).

This differential effect suggests that complement inhibition may exert a more pronounced protective influence on photoreceptors than is apparent when relying solely on RPE-based endpoints. From a biological perspective, this finding is plausible, as complement activation has been demonstrated on photoreceptor outer segments, and C3 inhibition acts upstream in the complement cascade (17,18). By preserving photoreceptor integrity, therapy may delay the downstream structural collapse of the RPE and choriocapillaris. Preservation of photoreceptors, even in the absence of immediate changes in RPE atrophy, may have important implications for future visual function and disease progression.

Limitations and future directions

The authors appropriately acknowledge several limitations, including the post hoc nature of the analysis, the relatively large baseline lesion sizes, and potential ceiling effects for EZ loss over longer follow-up. OCT reflectivity-based biomarkers, while informative, are not equivalent to histology, and changes in EZ appearance may reflect alterations in reflectivity rather than true structural recovery. Nevertheless, the consistency of treatment-related EZ preservation across trials supports the biological relevance of this marker.

By providing precise maps of EZ loss, AI-based OCT analysis opens the door to morphology-driven patient monitoring and treatment in disorders like GA where therapeutic effect may not translate into a visual acuity benefit. Future studies should prospectively evaluate the EZ-RPE difference as a predictor of disease progression and therapeutic response, ideally in earlier stages of GA. Tailored microperimetry guided by EZ integrity and GA morphology could enable better structure-function correlations and improve the ability of trials to detect meaningful therapeutic effects. This represents an important direction for future GA research.

Acknowledgments

None.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, Annals of Translational Medicine. The article did not undergo external peer review.

Funding: None.

Conflicts of Interest: Both authors have completed the ICMJE uniform disclosure form (available at https://atm.amegroups.com/article/view/10.21037/atm-2026-1-0018/coif). J.C. reports support from Allergan (C), Novartis (C), Salutaris (C), OD-OS (C), Erasca (C), B&L (C), Iveric Bio (C), Ocular Therapeutics (I), AcuViz (I), Abbvie (I), Springer (R), and Elsevier (R). The other author has no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Wong WL, Su X, Li X, et al. Global prevalence of age-related macular degeneration and disease burden projection for 2020 and 2040: a systematic review and meta-analysis. Lancet Glob Health 2014;2:e106-16. [Crossref] [PubMed]
2. Zhang K, Hopkins JJ, Heier JS, et al. Ciliary neurotrophic factor delivered by encapsulated cell intraocular implants for treatment of geographic atrophy in age-related macular degeneration. Proc Natl Acad Sci U S A 2011;108:6241-5. [Crossref] [PubMed]
3. Jaffe GJ, Tao W, Stinnett S, et al. The Effect of CNTF on Retinal Thickness in Eyes With AMD Associated With Geographic Atrophy. Investigative Ophthalmology & Visual Science 2009;50:3919.
4. Shiromani S, Leef D, Jain N, et al. Ciliary Neurotrophic Factor Implant in Geographic Atrophy from Age-related Macular Degeneration: A Long-Term Case Study. Retin Cases Brief Rep 2025; Epub ahead of print. [Crossref]
5. Heier JS, Lad EM, Holz FG, et al. Pegcetacoplan for the treatment of geographic atrophy secondary to age-related macular degeneration (OAKS and DERBY): two multicentre, randomised, double-masked, sham-controlled, phase 3 trials. Lancet 2023;402:1434-48. [Crossref] [PubMed]
6. Liao DS, Grossi FV, El Mehdi D, et al. Complement C3 Inhibitor Pegcetacoplan for Geographic Atrophy Secondary to Age-Related Macular Degeneration: A Randomized Phase 2 Trial. Ophthalmology 2020;127:186-95. [Crossref] [PubMed]
7. Wykoff CC, Rosenfeld PJ, Waheed NK, et al. Characterizing New-Onset Exudation in the Randomized Phase 2 FILLY Trial of Complement Inhibitor Pegcetacoplan for Geographic Atrophy. Ophthalmology 2021;128:1325-36. [Crossref] [PubMed]
8. Schmidt-Erfurth U, Mai J, Reiter GS, et al. Disease Activity and Therapeutic Response to Pegcetacoplan for Geographic Atrophy Identified by Deep Learning-Based Analysis of OCT. Ophthalmology 2025;132:181-93. [Crossref] [PubMed]
9. Liermann YN, Behning C, Isselmann B, et al. Ellipsoid zone reflectivity as a functional imaging biomarker for age-related macular degeneration: a MACUSTAR study report. Sci Rep 2025;15:20093. [Crossref] [PubMed]
10. Retina Today. Bryn Mawr Communications; [cited 2026 Jan 29]. A Closer Look at Ellipsoid Zone Attenuation. Available online: https://retinatoday.com/articles/2024-nov-dec/a-closer-look-at-ellipsoid-zone-attenuation
11. Arepalli S, Traboulsi EI, Ehlers JP. Ellipsoid zone mapping and outer retinal assessment in stargardt disease. Retina 2018;38:1427-31. [Crossref] [PubMed]
12. Goerdt L, Rodriguez Garcia JL, Künzel SH, et al. Prognostic Relevance of Relative Ellipsoid Zone Reflectivity for Ellipsoid Zone Loss in Macular Telangiectasia Type 2. Invest Ophthalmol Vis Sci 2024;65:36. [Crossref] [PubMed]
13. Chew EY, Clemons TE, Jaffe GJ, et al. Effect of Ciliary Neurotrophic Factor on Retinal Neurodegeneration in Patients with Macular Telangiectasia Type 2: A Randomized Clinical Trial. Ophthalmology 2019;126:540-9. [Crossref] [PubMed]
14. l.ferguson@neurotechusa.com. neurotech. 2025 [cited 2025 Mar 13]. Neurotech’s ENCELTOTM (revakinagene taroretcel-lwey) Approved by the FDA for the Treatment in Adults of Idiopathic Macular Telangiectasia Type 2 (MacTel). Available online: https://www.neurotechpharmaceuticals.com/neurotechs-enceltotm-revakinagene-taroretcel-lwey-approved-by-the-fda-for-the-treatment-of-macular-telangiectasia-type-2-mactel/
15. Sarici K, Abraham JR, Sevgi DD, et al. Risk Classification for Progression to Subfoveal Geographic Atrophy in Dry Age-Related Macular Degeneration Using Machine Learning-Enabled Outer Retinal Feature Extraction. Ophthalmic Surg Lasers Imaging Retina 2022;53:31-9. [Crossref] [PubMed]
16. Pfau M, von der Emde L, de Sisternes L, et al. Progression of Photoreceptor Degeneration in Geographic Atrophy Secondary to Age-related Macular Degeneration. JAMA Ophthalmol 2020;138:1026-34. [Crossref] [PubMed]
17. Edmonds R, Steffen V, Honigberg LA, et al. Alternative Complement Pathway Inhibition by Lampalizumab: Analysis of Data From Chroma and Spectri Phase III Clinical Trials. Ophthalmol Sci 2023;3:100286. [Crossref] [PubMed]
18. Geerlings MJ, de Jong EK, den Hollander AI. The complement system in age-related macular degeneration: A review of rare genetic variants and implications for personalized treatment. Mol Immunol 2017;84:65-76. [Crossref] [PubMed]