Self-Supervised Domain Adaptation for Diabetic Retinopathy Grading using Vessel Image Reconstruction

TL;DR

This paper introduces a self-supervised domain adaptation method for diabetic retinopathy grading that leverages vessel image reconstruction, outperforming existing strategies and achieving competitive accuracy with standard architectures.

Contribution

The paper proposes a novel self-supervised task based on vessel image reconstruction for domain adaptation in DR grading, demonstrating superior performance over current methods.

Findings

Our approach outperforms existing domain adaptation strategies.

Using entire target domain data, we achieve competitive accuracy with standard networks.

The method effectively leverages vessel reconstruction for invariant feature learning.

Abstract

This paper investigates the problem of domain adaptation for diabetic retinopathy (DR) grading. We learn invariant target-domain features by defining a novel self-supervised task based on retinal vessel image reconstructions, inspired by medical domain knowledge. Then, a benchmark of current state-of-the-art unsupervised domain adaptation methods on the DR problem is provided. It can be shown that our approach outperforms existing domain adaption strategies. Furthermore, when utilizing entire training data in the target domain, we are able to compete with several state-of-the-art approaches in final classification accuracy just by applying standard network architectures and using image-level labels.