A global review of publicly available image datasets for the anterior segment of the eye
This review, authored by the ESCRS Digital Health Special Interest Group, examines the current state of publicly available datasets for anterior segment imaging, focusing primarily on cataract, refractive, and corneal surgeries. Through a comprehensive search of PubMed and Google, the authors identified 26 accessible datasets. Most of these datasets are small, lack key demographic information, and primarily feature images of healthy eyes.
This highlights a significant need for more diverse and well-curated datasets to support advancements in artificial intelligence (AI) and machine learning (ML) applications in ophthalmology.
advancing research
A Review of Anterior Segment Imaging Datasets
The authors emphasise that improved dataset accessibility and design are crucial, as many current datasets are poorly described or difficult to access due to restrictive conditions. The limited scope of the available datasets could lead to biases, particularly when developing AI models intended for global use. Additionally, the review highlights the importance of international collaboration to enhance data sharing and ensure more representative data, particularly from underrepresented regions such as South America and Oceania.
Looking ahead, the authors propose the development of a regularly updated directory of anterior segment image datasets to increase transparency and foster collaboration. This initiative aims to improve access to data, enabling more robust AI-driven innovations in ophthalmology and ultimately leading to improved diagnostics, treatments, and outcomes in cataract, refractive, and corneal surgery.
