Fundus2Angio: A Conditional GAN Architecture for Generating Fluorescein Angiography Images from Retinal Fundus Photography

Carrying out clinical diagnosis of retinal vascular degeneration usingFluorescein Angiography (FA) is a time consuming process and can posesignificant adverse effects on the patient. Angiography requires insertion of adye that may cause severe adverse effects and can even be fatal. Currently,there are no non-invasive systems capable of generating Fluorescein Angiographyimages. However, retinal fundus photography is a non-invasive imaging techniquethat can be completed in a few seconds. In order to eliminate the need for FA,we propose a conditional generative adversarial network (GAN) to translatefundus images to FA images. The proposed GAN consists of a novel residual blockcapable of generating high quality FA images. These images are important toolsin the differential diagnosis of retinal diseases without the need for invasiveprocedure with possible side effects. Our experiments show that the proposedarchitecture outperforms other state-of-the-art generative networks.Furthermore, our proposed model achieves better qualitative resultsindistinguishable from real angiograms.