Soul: An OCTA dataset based on Human Machine Collaborative Annotation Framework

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Branch retinal vein occlusion (BRVO) is a leading cause of vision impairment due to increased venous pressure.
• Optical Coherence Tomography Angiography (OCTA) provides high-resolution 3D retinal vasculature imaging.
• Existing datasets lack focus on BRVO and comprehensive annotation, hindering research.

Purpose of the Study:

• To introduce the Soul dataset, specifically curated for BRVO research.
• To propose a Human-Machine Collaborative Annotation Framework (HMCAF) for efficient data labeling.
• To facilitate machine learning applications in analyzing retinal vascular diseases.

Main Methods:

• Development of the Soul dataset, comprising original images, blood vessel labels, and clinical data.
• Categorization of the dataset into 6 subsets based on injection frequency and follow-up duration.
• Implementation of HMCAF for annotating scrambled retinal blood vessel data.

Main Results:

• Creation of a specialized BRVO dataset (Soul) with diverse subsets.
• Establishment of a collaborative framework (HMCAF) for efficient and accurate annotation.
• Provision of a valuable resource for machine learning model development in ophthalmology.

Conclusions:

• The Soul dataset and HMCAF offer a significant advancement for BRVO research.
• This resource enables more effective machine learning-driven analysis of retinal vascular diseases.
• Future studies can leverage this dataset for improved diagnostic and prognostic tools.