At EURETINA 2025 in Paris, Dr David Eichenbaum presented late-breaking data on the use of structural biomarkers to predict exudative AMD in geographic atrophy (GA). Leveraging real-world datasets and AI-powered analysis through the EvidenceEngine platform, his research explored how multimodal imaging and validated biomarkers can generate scalable, rapid insights into disease progression. In this Q&A, Dr Eichenbaum discusses how AI compares with traditional trial approaches, the key biomarkers linked to exudation risk and the potential for AI-driven biomarker research to shape the future of retinal disease management and clinical research.
AI is able to leverage extremely large amounts of data – similar to our real-world studies – but has the advantage of processing and modelling this large amount of data very quickly. Parameters can be added, deleted or altered depending upon the research question you are trying to answer, allowing studies in large populations to be conducted with tremendous efficiency.
In our real-world population, the biomarkers most strongly associated with an increased likelihood of conversion to exudation were hyper-reflective foci (HRF) burden, retinal pigment epithelium to drusen complex (RPEDC) thickness, inner retinal thickening and the presence of subretinal drusenoid deposits (SDD).
Amaros’ EvidenceEngine AI links EHR, claims data, and multimodal imaging from a broad network of US retina clinics. For the model presented at EURETINA, data from 90 sites was incorporated. The AI harmonizes the data from the disparate inputs in the real world and leverages its own library of validated biomarkers which are extracted from the images and used to build our model. While we presented retrospective findings based on the large image dataset reviewed for this presentation, the longer-term goal is to apply this technology prospectively to advance research and support clinical decision-making.
AI-driven biomarker research offers a number of potential advantages. It allows for scale and representativeness through a broad, multi-input network, as well as providing speed and reproducibility with rapid, repeatable analyses across multiple sites as new data become available. In clinical research, this can allow for modelling clinical trial-like populations with existing data, and then identify eligible subjects based on the tuned set of potential inclusion/exclusion criteria. In disease management and clinical practice, retrospective analyses of large clinical datasets can help stratify risks and predict more or less likely scenarios of disease progression and potentially treatment responses, to truly individualize healthcare decisions and personalize treatment.
Disclosures: Dr Eichenbaum is an Investigator and Consultant to Amaros, BitFount and RetinAI. No fees or funding were associated with this article.
Citation: David Eichenbaum. EURETINA 2025: The use of structural biomarkers to predict exudative AMD in geographic atrophy. touchOPHTHALMOLOGY.com. 19 September 2025.
Register now to receive the touchOPHTHALMOLOGY newsletter!
Don’t miss out on hearing about our latest peer reviewed articles, expert opinions, conference news, podcasts and more.
