Dilated Deep Neural Architectures for Improving Retinal Vessel Extraction

Retinal vascular region is recognized as the promising anatomical for diagnosis of several commonly seen diseases including cardiovascular related and diabetes. In this paper we propose two novel deep neural architectures named Dilated fully convolved convolutional network (FCNN) dilated depth concatenated (DCNN) to segment retinal blood vessels. The proposed work evaluated both with without dilation. It observed from obtained results that dilation enhances performance. To eliminate non-uniform illumination low contrast differences effect preprocessed images are used training architectures. methodologies experimented on publicly available databases DRIVE STARE database. FCNN architecture can able obtain high accuracy about 95.39% which compared architecture. For DCNN also, 96.16% DCNN. experimental reveal significance operation in improving semantic segmentation