Artery/vein classification of retinal vessels using classifiers fusion

Area of Science:

• Ophthalmology and Medical Imaging
• Cardiovascular Health
• Computational Biology

Background:

• Morphological changes in retinal blood vessels are indicators of cardiovascular diseases.
• These vascular changes can lead to ocular complications like Hypertensive Retinopathy.
• Accurate classification of retinal arteries and veins is crucial for assessing vascular alterations.

Purpose of the Study:

• To develop and evaluate an automated system for classifying retinal blood vessels into arteries and veins.
• To improve the assessment of cardiovascular disease indicators through precise vascular analysis.
• To enhance the detection of ocular complications related to cardiovascular health.

Main Methods:

• Implementation of feature ranking strategies for identifying significant vascular characteristics.
• Utilizing a multiple classifiers decision-combination scheme for robust artery/vein classification.
• Validation using three datasets: a local dataset (44 images), INSPIRE-AVR (40 images), and VICAVR (58 images), including images with pathological conditions.

Main Results:

• The proposed system achieved high classification performance across all datasets.
• Overall accuracy rates were 90.45% for the local dataset, 93.90% for INSPIRE-AVR, and 87.82% for VICAVR.
• Experimental results were compared favorably against gold standard estimations and previous methodologies.

Conclusions:

• The developed approach demonstrates a promising and effective method for automated retinal artery/vein classification.
• This technique can aid in the early detection and monitoring of cardiovascular diseases through ocular imaging.
• The system's performance suggests its utility in clinical settings for assessing vascular health and related ocular complications.