RBAD: A Dataset and Benchmark for Retinal Vessels Branching Angle Detection

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Accurate detection of geometrical features in retinal images, specifically branching points, is crucial for diagnosing various eye diseases.
• Current methods for retinal branching angle analysis are often coarse, lacking the fine-grained detail needed for efficient annotation and diagnosis.

Purpose of the Study:

• To propose a novel, self-configured image processing technique for precise detection and calculation of retinal branching angles.
• To introduce an open-source annotation tool and a benchmark dataset to facilitate research in this area.

Main Methods:

• A self-configured image processing technique was developed for detecting retinal branching angles.
• An open-source annotation tool and a benchmark dataset of 40 annotated retinal images were created.
• The proposed method was benchmarked against previous approaches for accuracy and efficiency.

Main Results:

• The novel method demonstrates high accuracy and robustness across various conditions in detecting retinal branching angles.
• The developed technique offers improved efficiency compared to existing methods for fine-grained retinal image analysis.
• The open-source dataset and tool provide valuable resources for the ophthalmic research community.

Conclusions:

• The proposed method for retinal branching angle detection is a valuable instrument for ophthalmic research and clinical applications.
• The availability of the dataset and source code promotes further development and validation in the field.
• This work addresses the limitations of coarse-level analysis, offering a more precise approach to retinal image interpretation.