Blood Vessel Extraction in Color Retinal Fundus Images with Enhancement Filtering and Unsupervised Classification

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Retinal blood vessels are vital indicators for diagnosing systemic diseases like diabetic retinopathy, glaucoma, arteriosclerosis, and hypertension.
• Accurate extraction of retinal vasculature is essential for early disease detection and treatment planning.

Purpose of the Study:

• To develop and evaluate a novel automated method for retinal blood vessel network extraction.
• To improve the accuracy and reliability of retinal vasculature segmentation for diagnostic purposes.

Main Methods:

• The proposed method involves four stages: preprocessing, enhancement using Gabor, Frangi, and Gauss filters with a top-hat transform, hard/soft clustering (K-means, Fuzzy C-means) for binary vessel map generation, and postprocessing to remove false positives.
• The enhancement stage utilizes multiple filters to improve vessel visibility.
• Clustering algorithms are employed to segment the vessel structures.

Main Results:

• The Gabor filter combined with K-means clustering demonstrated high performance, achieving 95.94% accuracy on the STARE database and 95.71% on the DRIVE database.
• The developed method effectively extracts retinal blood vessel networks from color retinal images.
• The postprocessing step successfully removed falsely segmented regions.

Conclusions:

• The novel approach for retinal blood vessel extraction, particularly the Gabor filter and K-means clustering combination, is highly accurate and suitable for clinical diagnostic systems.
• This automated method offers a reliable tool for ophthalmologists and healthcare professionals.
• Further validation on diverse datasets could enhance its clinical applicability.