Artery-Vein Segmentation from Fundus Images using Deep Learning

TL;DR

The paper tackles automatic artery–vein segmentation in fundus images, a key step for retinal vessel analysis and cardiovascular risk assessment. It introduces Attention-WNet, a hybrid architecture that combines attention mechanisms with a dual-UNet (WNet) framework and trains separate models for arteries and veins to reduce arterio-venous confusion. The method uses z-score normalization, CLAHE on the green channel, and focal loss to address severe class imbalance, and it is evaluated on DRIVE and HRF with cross-dataset experiments showing competitive accuracy and robust generalization. The work demonstrates improved A/V segmentation performance and provides a path toward clinically useful retinal vasculature analysis across varied imaging conditions.

Abstract

Segmenting of clinically important retinal blood vessels into arteries and veins is a prerequisite for retinal vessel analysis. Such analysis can provide potential insights and bio-markers for identifying and diagnosing various retinal eye diseases. Alteration in the regularity and width of the retinal blood vessels can act as an indicator of the health of the vasculature system all over the body. It can help identify patients at high risk of developing vasculature diseases like stroke and myocardial infarction. Over the years, various Deep Learning architectures have been proposed to perform retinal vessel segmentation. Recently, attention mechanisms have been increasingly used in image segmentation tasks. The work proposes a new Deep Learning approach for artery-vein segmentation. The new approach is based on the Attention mechanism that is incorporated into the WNet Deep Learning model, and we call the model as Attention-WNet. The proposed approach has been tested on publicly available datasets such as HRF and DRIVE datasets. The proposed approach has outperformed other state-of-art models available in the literature.

Figures (7)

• Figure 1: The flow Diagram of the Proposed Model
• Figure 2: Channel-wise histogram of artery and veins for original, red, green, blue, and grayscale fundus images of a single patient.
• Figure 3: Image enhancement through CLAHE on the green channel from the input fundus image(ronneberger2015u)
• Figure 4: Proposed approach using Attention-WNET Deep Learning Model for AV segmentation
• Figure 5: Evaluation Procedure: Illustration of the different metrics used in the evaluation of A/V segmentation, from left to right (1) All vessel pixels (2) Vessel centerline pixels (3)) Centerline pixels of vessels wider than two pixels (hemelings2019artery)