Artery-venous classification in fluorescein angiograms based on region growing with sequential and structural...

Area of Science:

• Ophthalmology
• Medical Imaging
• Computational Biology

Background:

• Fluorescein angiography (FA) is vital for retinal hemodynamics and vascular morphology analysis.
• Artery-venous classification in FA images is essential for clinical parameters like arterio-venous passage time (AVP) and arterio-venous width ratio (AVR).
• Manual classification is time-consuming and requires expertise, necessitating automated solutions.

Purpose of the Study:

• To propose a novel, automated method for artery-venous classification in FA images.
• To develop a region growing strategy with sequential and structural features (RGSS) for accurate classification.
• To overcome the limitations of manual classification in FA image processing.

Main Methods:

• Image registration using mutual information.
• Extraction of sequential dye perfusion features and vessel structural features.
• Region growing strategy seeded from the optic disc, propagating through the vessel network.

Main Results:

• The RGSS method achieved high classification accuracy: 0.91 ± 0.04 on the Duke dataset and 0.92 ± 0.03 on a private dataset.
• Demonstrated effectiveness in classifying thin arteries and veins, even at vessel crossings.
• Successfully classified arteries and veins within complex retinal vascular networks.

Conclusions:

• The proposed RGSS method enables accurate automatic artery-venous classification in FA images.
• This automation frees ophthalmologists from laborious manual marking.
• Facilitates precise measurement of AVP and AVR for improved clinical assessment of circulatory issues.