Meet the data & AI researchers: Pearse Keane's Lab at ARVO 2025

The medical Artificial Intelligence (AI) lab of Pearse Keane has helped to shape the direction of data and machine vision research in ophthalmology. Professor Keane’s lab brings together clinical researchers and data scientists spanning UCL Institute of Ophthalmology (IoO) and Moorfields Eye Hospital, which are ranked number 1 globally for research in ophthalmology and vision science. Much of the lab’s work has been enabled by the INSIGHT Health Data Research Hub, directed by Professor Keane. INSIGHT is the world’s largest bioresource of ophthalmic imaging linked to clinical records, and provides data curation and computing infrastructure for AI development. 

This month, members of the Keane AI lab shared their latest projects at the world’s largest gathering of eye and vision researchers, The Association for Research in Vision and Ophthalmology (ARVO) annual meeting, held this year in Salt Lake City. Here we showcase the rising stars to watch, highlighting the breadth of their data and AI-enabled research across four key areas. (Links below are to abstracts on ARVO's website).

Foundation Models

The Keane AI lab continues to innovate in developing and refining retinal foundation models, focusing on making these AI tools more accessible and clinically relevant.

Eden Ruffell is exploring how foundation models can be enhanced through federated learning while maintaining privacy and improving their application to broader settings. Eden's study, “Federated fine-tuning a foundation model for generalisable disease detection”, demonstrates how federated learning could maintain or even enhance model performance, matching the accuracy of centralised models. Link to abstract

Yukun Zhou, who led development of the world’s first foundation model, RETFound, is interrogating the crucial relationship between training data and model performance, asking the question, "How does training data impact medical foundation models?". Yukun is also anticipating the next trends in foundation model development. Link to abstract

Justin Engelmann has developed RETFound-Green and RETFound-Mobile, high-performance versions of the original RETFound that do not require a high level of computing power. RETFound Mobile, for example, can be run on a regular smartphone. Justin’s work demonstrates how advanced AI capabilities can be deployed on everyday devices. He is also leading work on RETSeg, a retinal foundation model that can perform complex segmentation tasks with minimal input. Link to abstract

Gatera Fiston Kitema, together with Justin Engelmann and others in the team, has adapted RETFound-Green to create a high-performance, smartphone-ready glaucoma AI model designed for use in low-resource settings. This exciting research has transformative potential for applications in areas of high patient need. The abstract for the project is “Developing a high-performance, smartphone-ready glaucoma AI model with RETFound-Green” Link to abstract

From left: Eden Ruffell and Yukun Zhou; Pearse Keane presents.

Biomarkers and Oculomics

The group continues to progress retinal biomarker analysis to expand understanding of how eye conditions progress, and to push boundaries in the field of Oculomics, using AI to discover linkages between retinal biomarkers and systemic disease.

Josef Huemer’s study, "Atrial fibrillation and retinal sublayer thickness: a cross-sectional analysis of the AlzEye study," is investigating connections between heart conditions and retinal structure. Link to abstract

Ana Paula Ribeiro Reis is studying links between the retina and women’s reproductive health. One aspect of Ana’s research examines how hormonal contraceptives may affect retinal vasculature, interrogating the association of oral contraceptive pill use with vascular biomarkers from colour fundus photography in pre-menopausal women. Link to abstract

Owen Sweeney leads research that uses AI to identify retinal changes associated with obesity. Loss of peripheral sympathetic neurons is an important pathological process in obesity but measuring these changes non-invasively and at scale is challenging. Oculomics—analysing retinal imaging for systemic health insights—may offer a solution. Owen’s study investigates the hypothesis that obesity is associated with changes in retinal morphology. Link to abstract

Sophia Ghauri is investigating potential retinal biomarkers in psychiatric conditions - in particular looking at what clues Optical Coherence Tomography (OCT) can provide in understanding Bipolar Affective Disorder, using data from UK Biobank. Sophia’s project is entitled “Exploring Optical Coherence Tomography in Bipolar Affective Disorder Using Data from the UK Biobank”. Link to abstract

Rohan Misra is looking at how retinal

vascular properties and retinal layer thicknesses may be associated with Inflammatory Bowel Disease, exploring potential ocular manifestations of systemic inflammatory conditions. Emerging evidence highlights gut-retina crosstalk, where microbial alterations may contribute to intraocular inflammation. Link to abstract

David Merle has analysed OCT biomarkers using AI to predict fellow eye conversion in neovascular (wet) Age-related Macular Degeneration (AMD). Using a large, real-world dataset from Moorfields Eye Hospital, the study has identified and confirmed key structural biomarkers associated with progression risk. These findings provide a foundation for improved risk stratification and personalized monitoring strategies. Link to abstract.

Hagar Khalid has evaluated changes in visual acuity (VA) in the fellow eye of patients with neovascular age-related macular degeneration (nAMD) before the eyes transition to nAMD. Hagar examined functional changes that may predict disease progression, assessing changes from the date of observation of new signs of nAMD to the date of injection. Hagar’s project is entitled “Visual acuity changes prior to transition to neovascular Age-related macular degeneration” Link to abstract

Oscar Earle is working on a cross-sectional UK study, assessing the retinal age gap—calculated as the difference between retinal age (predicted using deep learning algorithms) and chronological age—as a potential biomarker in individuals with ulcerative colitis. Link to abstract

Shubhank Verma’s investigation is set against the backdrop of increasing interest in retinal imaging oculomics as markers of accelerated aging and systemic health. In particular, Shubank is interrogating the interplay around socioeconomic factors in his study “Eyeing Inequities: Socioeconomic Deprivation in Retinal Age Gap”. Link to abstract

Clockwise from top left: Hagar Khalid, Sophia Ghauri, Ana Paula Ribeiro Reis, Rohan Misra.

Technical Advancements

Another research theme  for the Keane AI lab  is advancing fundamental AI methodologies for ophthalmic applications.

Robbert Struyven is developing tools for improved OCT image annotation. He has led work on a user-friendly tool that lets clinicians and researchers quickly view, validate, and edit AI-generated segmentation masks of Optical Coherence Tomography (OCT) scans. The result is ”OCT-AVE: OCT Annotation, Validation & Editing for AI-enabled segmentation analysis".

Peter Woodward-Court is investigating how Large Language Models can be used to enrich healthcare data while ensuring patient privacy. Peter’s abstract is entitled "Large Language Models for Clinical Letter Anonymisation: A comparison to traditional approaches". Link to abstract.

Lie Ju is addressing out-of-distribution detection challenges in clinical AI with his study "Evaluation of Generalized OOD Detection in Deep Learning-Based Retinal Disease Diagnosis." This technical work is tackling the issue of detecting poor-quality images and another image anomalies that could compromise the robustness and safety of AI systems in clinical settings. Link to abstract

Dominic Williamson is a computer scientist with expertise in optimising anonymised retinal data for research. Dominic has led development of an AI system, trained on retinal scans, for more precise clinical trial recruitment. Link to press release

Data Integration and Implementation

Tackling the challenges of implementing AI in healthcare systems is an important objective for the research team as they look to bring "code to clinic".

Ariel Ong focuses on bridging technical and regulatory aspects of AI. Ariel’s studies include: “Towards automated and reliable fundus image quality assessment" and "A scoping review of artificial intelligence as a medical device (AIaMD) for ophthalmic image analysis in Europe, Australia and the US". Link to abstract.

Siegfried Wagner, who co-led work that originated the field of Oculomics in 2020, is investigating how integrated health data can power predictive models. Siegfried’s current studies include "Scaling UK Health Data Linkage: Implications for Oculomics" and "Prediction of incident cardiovascular events using clinically interpretable features from multimodal retinal imaging." Link to abstract

Pearse Keane is Professor of Artificial Medical Intelligence at UCL Institute of Ophthalmology, consultant ophthalmologist at Moorfields Eye Hospital and Director of the INSIGHT Health Data Research Hub. Pearse leads the AI Lab in developing the next generation of vision-language foundation models, optimising the use of anonymised data which has been collected as part of routine clinical care. In his presentation "The Role of Cloud Computing and Generative AI in Implementation Science," Pearse shared learnings on how these technologies can bridge research and practice, leveraging the INSIGHT bioresource for ophthalmic imaging. Link to abstract

From left: Ariel Ong, Siegfried Wagner

These diverse projects underscore the remarkable range of research happening within Professor Keane's group, from technical AI innovations to direct clinical applications. Their work spans the entire spectrum from foundation model development to clinical deployment, showcasing how AI is transforming ophthalmology. 

The group's presence at ARVO 2025 highlights their position at the forefront of using artificial intelligence to improve ophthalmic care and better understand the relationship between eye health and systemic disease.

For more detail on each of the projects outlined above, you can find the abstracts on the ARVO planner, using the abstract titles or researcher names above: https://eppro02.ativ.me/src/EventPilot/php/express/web/planner.php?id=ARVO25

---

To find out more about how INSIGHT works with researchers, visit our Researcher overview.