OCTA-500: A retinal dataset for optical coherence tomography angiography study.

Area of Science:

• Ophthalmology
• Neuroscience
• Medical Imaging

Background:

• Optical Coherence Tomography Angiography (OCTA) is crucial for visualizing retinal vasculature.
• Publicly available OCTA datasets are limited, hindering research and development.

Purpose of the Study:

• Introduce OCTA-500, the largest and most comprehensive OCTA dataset to date.
• Provide rich annotations for diverse OCTA imaging analysis.
• Establish a benchmark for microvascular segmentation tasks.

Main Methods:

• Collected OCTA imaging from 500 subjects across two fields of view.
• Developed extensive annotations including modalities, projections, text, and segmentation labels.
• Proposed the Capillary, Artery, Vein, and FAZ (CAVF) segmentation task.
• Optimized the 3D-to-2D Image Projection Network (IPN) to IPN-V2.

Main Results:

• OCTA-500 contains OCT/OCTA volumes, multiple projections, and detailed segmentation labels.
• IPN-V2 demonstrated a ~10% mIoU improvement over IPN on the CAVF task.
• Analysis explored dataset characteristics' impact on segmentation performance.

Conclusions:

• OCTA-500 is a valuable, publicly available resource for OCTA research.
• The proposed CAVF task and IPN-V2 advance retinal microvascular segmentation.
• The dataset facilitates further studies on OCTA imaging and analysis.