Towards Adversarial Retinal Image Synthesis

TL;DR

This paper introduces a data-driven method for synthesizing retinal images from vessel structures using adversarial image-to-image translation, enabling realistic eye fundus image generation without complex anatomical modeling.

Contribution

It presents a novel approach that learns to generate retinal images directly from vessel trees using adversarial learning, bypassing traditional complex anatomical models.

Findings

Generated images are visually distinct yet retain key features.

Quantitative analysis confirms high quality of synthetic images.

Method effectively learns from paired vessel and retinal images.

Abstract

Synthesizing images of the eye fundus is a challenging task that has been previously approached by formulating complex models of the anatomy of the eye. New images can then be generated by sampling a suitable parameter space. In this work, we propose a method that learns to synthesize eye fundus images directly from data. For that, we pair true eye fundus images with their respective vessel trees, by means of a vessel segmentation technique. These pairs are then used to learn a mapping from a binary vessel tree to a new retinal image. For this purpose, we use a recent image-to-image translation technique, based on the idea of adversarial learning. Experimental results show that the original and the generated images are visually different in terms of their global appearance, in spite of sharing the same vessel tree. Additionally, a quantitative quality analysis of the synthetic retinal images confirms that the produced images retain a high proportion of the true image set quality.