A Multi-Anatomical Retinal Structure Segmentation System for Automatic Eye Screening Using Morphological Adaptive...

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Ophthalmologists rely on accurate segmentation of retina fundus images for diagnosis.
• Segmentation accuracy of retinal vessels, optic disc, and lesions directly impacts clinical decisions.
• Early detection of systemic diseases like diabetes can be facilitated by retinal screening.

Purpose of the Study:

• To propose an automated retinal fundus image segmentation system.
• To achieve high-accuracy segmentation of multiple retinal structures (vessels, optic disc, exudates) in a single session.
• To develop a system robust to pathological image variations.

Main Methods:

• A hybrid segmentation algorithm combining adaptive fuzzy thresholding and mathematical morphology.
• Three subsystems utilizing the same core algorithm for segmenting different retinal features.
• Validation on four benchmark datasets: DRIVE, STARE, DRISHTI-GS, and DIARETDB1.

Main Results:

• The proposed system achieved competitive segmentation performance across all tested datasets.
• It successfully segmented retinal vessels, optic discs, and exudate lesions with high accuracy.
• The system demonstrated superior performance compared to several existing state-of-the-art methods.

Conclusions:

• The automated system provides accurate and comprehensive segmentation of key retinal structures.
• It offers a valuable tool for compact diagnosis and enhanced clinical insight, particularly for diabetic retinopathy detection.
• The system's adaptability suggests potential for segmenting other heterogeneous anatomical structures in medical imaging.