Federal Grant Award Summary

The University of North Carolina at Charlotte received a $617,244 Project Grant award from the National Eye Institute (NEI) under the Vision Research program (CFDA 93.867), effective May 1, 2026 through March 31, 2030. The project, titled "Distributed Foundational Models for Multi-Task Learning in Diabetic Retinopathy," develops and validates distributed federated learning (FL) approaches for training foundational machine learning (ML) models to address diabetic retinopathy (DR), a leading cause of severe vision loss. The research utilizes multi-modal optical coherence tomography (OCT) data from multiple institutions to create algorithms capable of performing multi-task functions including disease stage classification, disease progression prediction, and treatment outcome assessment.

The project delivers scientific and technical innovations that address critical barriers to ML model development in ophthalmic research, specifically the challenge of limited access to large, diverse, well-curated datasets due to privacy concerns and data-sharing restrictions. The primary deliverable is the development and validation of a novel Ophthalmic Adaptive Personalized Federated Learning (OPTHO-APFL) technique across four independent institutions, which enables collaborative model training while preserving patient data privacy and accommodating heterogeneous data distributions across different clinical sites and imaging protocols. Clinical deployment experiments will validate the federated learning framework's effectiveness in real-world ophthalmologic settings, establishing foundational ML models that support improved early detection, intervention, and treatment monitoring for diabetic retinopathy patients.