Artificial intelligence (A) is reshaping how glaucoma is detected and managed, but its clinical impact hinges on the quality of the data behind it. Speaking at the 2nd International Glaucoma Symposium on 31 January 2026, in Mainz, Germany, Prof. Dr. Luís Abegão Pinto underscored the need for robust, population-based normative databases to ensure AI tools reflect real-world biology rather than referral-driven datasets.¹

In this Q&A conversation, the Eye Care Network spoke with Prof. Pinto about how AI-enabled normative data could redefine screening, diagnosis, and clinical decision-making in the years ahead.

Note: Transcript edited for clarity and length.

The Eye Care Network: Could you please summarise the key themes of your presentation and why building robust normative databases with AI is so crucial for the future of glaucoma care?

Prof. Dr. Luís Abegão Pinto, MD, PhD, FEBOS-GL: The key theme is that the performance of AI in glaucoma is fundamentally limited by the reference data we use. Many current systems are trained on hospital-based datasets with high disease prevalence and local definitions of glaucoma. Robust, population-based normative databases allow us to anchor AI to real-world biology rather than referral patterns. That is essential if we want AI to move from promising prototypes to tools that genuinely support clinical decision-making at scale.

Population-based data collection can be complex. What advantages does AI bring to this process compared with traditional methods, particularly in terms of scale, accuracy, or bias reduction?

Prof. Pinto: AI makes scale possible. It allows automated analysis, standardised feature extraction, and continuous quality control across very large datasets, something that would be impractical with traditional approaches. It also enables us to capture and model real-world variability across populations and devices. Although AI does not eliminate bias on its own, it allows bias to be identified and handled systematically, rather than implicitly embedded in small or highly selected cohorts.

From your experience, what are the biggest challenges in deploying AI tools in real-world population setting —such as data quality, device variability, or patient diversity—and how is your work addressing these obstacles?

Prof. Pinto:The biggest challenge is defining what we are asking AI to detect. Concepts like “glaucoma” or “referable glaucoma” are not absolute; they depend on the goals of the screening programme and the health care context. Our work addresses this by combining population-based screening with clinic-level validation, including optical coherence tomography, visual fields, and structured phenotyping. This allows us to stratify disease severity and functional relevance, and to define clear, purpose-driven targets for AI rather than relying on a single, overly simplistic label.

Looking ahead, how do you envision AI-enhanced normative databases transforming clinical decision-making or screening programmes for glaucoma in the next 5 to 10 years?

Prof. Pinto: As the technology matures, I expect AI-enhanced normative databases to become part of routine clinical infrastructure rather than research tools. Adoption will likely accelerate once these systems are trusted and widely deployed, and that increased use will, in turn, generate larger and more credible datasets. This creates a positive feedback loop: better data lead to better models, which support broader clinical uptake. Over time, this should deepen our understanding of glaucoma as a disease—how it presents, how it progresses, and how risk varies across populations—and ultimately, lead to more informed and effective care.

Prof. Dr. Luís Abegão Pinto, MD, PhD, FEBOS-GL
E: [email protected]
Prof. Pinto is professor of ophthalmology - Faculty of Medicine of Lisbon University and head of the glaucoma unit of ULS Santa Maria in Lisbon, Portugal.

Reference

1. Pinto LA. The prowess of deploying AI for normative database collection in a population-based setting. Presented at: 2nd International Glaucoma Symposium; 31 January 2026; Mainz, Germany.