ORIGA(-light): an online retinal fundus image database for glaucoma analysis and research

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Science

Background:

• Retinal fundus images are crucial for diagnosing diseases like glaucoma and diabetic retinopathy.
• Current retinal data is often stored locally, limiting the exploitation of clinical knowledge.
• There is a need for accessible, annotated datasets for developing and validating diagnostic algorithms.

Purpose of the Study:

• To establish ORIGA(-light), an online repository for sharing clinical ground-truth retinal images.
• To provide open access for researchers to benchmark computer-aided segmentation algorithms.
• To facilitate the development of objective methods for analyzing retinal image features.

Main Methods:

• Development of an in-house image segmentation and grading tool.
• Creation of a quantified objective benchmarking method focused on optic disc and cup segmentation.
• Annotation of retinal images by trained professionals, including glaucoma diagnostic signs.

Main Results:

• ORIGA(-light) currently houses 650 annotated retinal images.
• The database includes detailed annotations for optic disc and cup segmentation and Cup-to-Disc Ratio (CDR).
• The system is designed for continuous updates with more clinical data.

Conclusions:

• ORIGA(-light) offers a valuable public resource for retinal image analysis research.
• The platform supports the benchmarking of computer-aided diagnostic tools for ocular diseases.
• Open access to annotated data promotes advancements in automated glaucoma detection and other retinal condition diagnoses.