AGS 2025: Development and evaluation of an AI model to set target IOP with Jithin Yohannan, MD, MPH

Jithin Yohannan, MD, MPH, discussed the development and evaluation of an AI model designed to assist in setting target IOP for patients with glaucoma at the 2025 American Glaucoma Society (AGS) Annual Meeting, held February 26 to March 2 in Washington, DC.

Previous research has shown that deviations from a set target IOP correlate with glaucoma progression, making precise target-setting critical, noted Yohannan, of Wilmer Eye Institute, Johns Hopkins University School of Medicine in Baltimore, Maryland. However, nonspecialists may lack the expertise to establish accurate targets, prompting the need for an AI-driven approach.

Yohannan and his team developed a machine learning model trained on a large dataset of visual field, OCT, and clinical information to predict individualized target IOP values. In initial testing, the model performed well, with an average error of about 2 mmHg. Further validation using longitudinal data showed that the AI-generated target IOP was comparable to those set by clinicians in terms of its impact on glaucoma progression. The AI model also outperformed society guideline-based targets and the absence of target-setting.

Challenges remain in integrating this AI model into clinical workflows. One major issue is backend data extraction, as accessing necessary patient data across different electronic medical record systems is complex. Additionally, frontend user interface design must ensure that clinicians meaningfully incorporate AI-generated targets into their decision-making. Future steps include external validation at multiple institutions and refining the model’s integration into diverse clinical settings.

Yohannan acknowledged the contributions of Alex Pham, BS (ophthalmology resident, University of Maryland) and Edgar Robitaille (undergraduate, Johns Hopkins) in advancing this research.