VESSEL-NET: A VESSEL-AWARE ENSEMBLE NETWORK FOR RETINOPATHY SCREENING FROM FUNDUS IMAGE

Retinal diseases are one of the most popular malady in modern world with increasingly use of electronic screens. Previous studies on deep learning based automatic screening generally used global image of fundus. However, due to limited size of input image, global image lacks in information of medical-related region and has poor resolution in local details. This paper proposes a novel vessel-aware ensemble network for retinal disease detection. Specifically, the proposed network consists of three aspect of information which integrates deep diverse context of the global fundus image and the local medical-related regions, i.e. local disc-region and local vessel-related region. Finally, the output probabilities of different streams are fused as the final screening result. The experiment results on the AMD data of ODIR dataset show that our method outperforms other state-of-the-art algorithms and shows the significance of vessel-related regions.