Multiscale Joint Optimization Strategy for Retinal Vascular Segmentation

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Accurate retinal vascular segmentation is vital for diagnosing conditions like diabetes, hypertension, microaneurysms, and arteriosclerosis.
• Existing methods struggle to segment deep and small blood vessels effectively, limiting diagnostic information.

Purpose of the Study:

• To develop a multi-scale joint optimization strategy for enhanced retinal vascular segmentation.
• To improve the detection of small and deep blood vessels for better disease diagnosis.

Main Methods:

• Utilized Multi-Scale Retinex (MSR) for uneven illumination correction in fundus images.
• Applied multi-scale Gaussian matched filtering to enhance retinal image contrast.
• Employed Particle Swarm Optimization (PSO)-optimized Otsu multi-threshold segmentation.
• Performed post-processing including binarization, morphological operations, and edge-contour removal.

Main Results:

• The proposed method successfully segmented more small blood vessels compared to existing techniques.
• Ensured the integrity of the vascular structure while improving segmentation performance.
• Achieved more obvious targets, higher contrast, and richer detailed information in segmented images.

Conclusions:

• The multi-scale joint optimization strategy significantly enhances retinal vascular segmentation.
• The method offers superior performance and provides more detailed diagnostic information for eye diseases.
• Demonstrated effectiveness and practicability on DRIVE and STARE datasets.