Enhancing fine retinal vessel segmentation: Morphological reconstruction and double thresholds filtering strategy.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Diabetic retinopathy causes progressive changes in retinal vessels, complicating disease analysis and treatment.
• Existing computerized retinal vessel segmentation algorithms struggle with tiny vessels, impacting overall performance.

Purpose of the Study:

• To develop an improved image processing technique for accurate retinal vessel segmentation.
• To enhance the analysis and diagnosis of progressive eye diseases through better vessel segmentation.

Main Methods:

• Utilized enhancement filters, coherence filters, binary thresholding, morphological techniques, homomorphic methods, and Wiener filtering.
• Addressed center light reflex, insufficient/varying contrast, and vessel coherence issues.
• Applied double thresholding with image reconstruction for precise segmentation.

Main Results:

• Achieved high accuracy (0.96) and sensitivity (0.81) on STARE and DRIVE datasets.
• Demonstrated improved performance over existing retinal vessel segmentation methods.
• Successfully segmented fine retinal vessels, overcoming common algorithm limitations.

Conclusions:

• The proposed technique offers a robust solution for retinal vessel segmentation.
• This method can aid ophthalmology experts in diagnosing ocular abnormalities and guiding treatment.
• The improved segmentation capability supports better management of progressive eye diseases.