RVO-ME: A Dual-Task OCT Dataset for Segmentation and Detection of Macular Lesions in Retinal Vein Occlusion

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Retinal vein occlusion (RVO) is a leading cause of vision loss, often complicated by macular edema (ME).
• Optical coherence tomography (OCT) is crucial for diagnosing and monitoring RVO-ME.
• Automated RVO-ME analysis is limited by a lack of high-quality annotated datasets.

Purpose of the Study:

• To create a manually annotated dataset of OCT B-scans for RVO-ME analysis.
• To provide a benchmark for evaluating automated segmentation algorithms.
• To advance artificial intelligence (AI) models for RVO-related diseases.

Main Methods:

• Compiled 3,012 OCT B-scans from 146 eyes (130 patients) with RVO-ME.
• Provided segmentation labels for subretinal fluid, intraretinal fluid, ellipsoid zone, and external limiting membrane.
• Included point annotations for detecting highly reflective foci.

Main Results:

• Developed a comprehensive RVO-ME OCT dataset.
• The dataset includes detailed annotations for key retinal structures and fluid indicators.
• Facilitates quantitative analysis and algorithm development.

Conclusions:

• The new dataset addresses the critical need for annotated data in RVO-ME research.
• It serves as a valuable resource for developing and validating AI-driven diagnostic tools.
• Expected to accelerate progress in automated analysis of RVO-ME using OCT imaging.