Attention W-Net: Improved Skip Connections for better Representations

TL;DR

The paper introduces Attention W-Net, a novel architecture for retinal vessel segmentation that incorporates attention mechanisms and regularisation to improve compatibility, prevent overfitting, and achieve state-of-the-art performance.

Contribution

It presents a new U-Net based architecture with attention blocks and enhanced regularisation measures for improved retinal vessel segmentation.

Findings

Achieved F1 score of 0.8407 on DRIVE dataset

Achieved AUC of 0.9833 on DRIVE dataset

Outperformed its backbone and was competitive with state-of-the-art methods

Abstract

Segmentation of macro and microvascular structures in fundoscopic retinal images plays a crucial role in the detection of multiple retinal and systemic diseases, yet it is a difficult problem to solve. Most neural network approaches face several issues such as lack of enough parameters, overfitting and/or incompatibility between internal feature-spaces. We propose Attention W-Net, a new U-Net based architecture for retinal vessel segmentation to address these problems. In this architecture, we have two main contributions: Attention Block and regularisation measures. Our Attention Block uses attention between encoder and decoder features, resulting in higher compatibility upon addition. Our regularisation measures include augmentation and modifications to the ResNet Block used, which greatly prevent overfitting. We observe an F1 and AUC of 0.8407 and 0.9833 on the DRIVE and 0.8174 and 0.9865 respectively on the CHASE-DB1 datasets - a sizeable improvement over its backbone as well as competitive performance among contemporary state-of-the-art methods.