Learning to Segment Vessels from Poorly Illuminated Fundus Images

Segmentation of retinal vessels is important for determining various disease conditions, but deep learning approaches have been limited by the unavailability of large, publicly available, and annotated datasets. The paper addresses this problem and analyses the performance of U-Net architecture on DRIVE and RIM-ONE datasets. A different approach for data augmentation using vignetting masks is presented to create more annotated fundus data. Unlike most prior efforts that attempt transforming poor images to match the images in a training set, our approach takes better quality images (which have good expert labels) and transforms them to resemble poor quality target images. We apply substantial vignetting masks to the DRIVE dataset and then train a U-net on the resulting lower quality images (using the corresponding expert label data). We quantitatively show that our approach leads to better generalized networks, and we show qualitative performance improvements in RIM-ONE images (which lack expert labels).