Export and Integration of Ophthalmic Optical Coherence Tomography Data Across Institutions (OCT) - Project Summary/Abstract Optical Coherence Tomography (OCT), a three-dimensional high-resolution in-vivo imaging technology, has be- come the gold standard of ophthalmic diagnostics, indispensable for detecting and monitoring the leading causes of blindness worldwide. At Mass. Eye and Ear (MEE) alone, over 1,500 OCT scans are performed every day, adding to millions of existing scans that represent an untapped treasure for “big data” machine learning research. Yet this potential remains locked: manufacturers store OCT data in over ten proprietary, undocumented formats, and even when official export tools exist, they are prohibitively slow. For example, exporting MEE's full Heidelberg Spectralis dataset would take an estimated 11 years. As a result, OCT imaging data is effectively inaccessible for large-scale research and AI applications. This project will change that. We will deliver a comprehensive, open-source solution that unlocks OCT data for the entire research community. Our team brings together world-class expertise in ophthalmology, machine learning, and system-level programming, including leaders of the IRIS and SOURCE registries that together cover over 70 million patients but currently lack OCT imaging due to these barriers. Aim 1 will establish a unified, open OCT data format with dual implementations: a DICOM version for seamless integration into clinical IT systems, and an HDF5 version engineered for machine learning speed and efficiency. High-performance Rust-based converters will translate all major vendor formats into this standard. Aim 2 will deliver turnkey repository software that automates data extraction from Picture Archiving and Communication Systems (PACS) or devices, supports flexible de-identification, and provides researcher-friendly programming interfaces in C, Python, and R. Aim 3 will deploy this infrastructure across 17 U.S. institutions within the project team and the SOURCE consortium as the initial launch sites, creating the first-ever distributed, federated OCT imaging repository with automated updates and seamless cross-site integration. The repository software is designed for broad, nationwide adoption: any ophthalmic center with a supported Picture Archiving and Communication System (PACS) will be able to implement its own local repository without specialized expertise. This means that in the long run, virtually every OCT scan performed in the U.S. can be- come searchable, shareable, and AI-ready. Once accomplished, this project will transform ophthalmic research by turning fragmented, inaccessible imaging into a unified national resource. It will enable unprecedented large-scale machine learning studies, accelerate discovery and reproducibility, and directly fuel innovation in the diagnosis and treatment of blinding eye diseases.
