Optic disc and fovea detection using multi-stage region-based convolutional neural network

Detection of the optic disc (OD) and fovea in retinal images is an important step for automated detection of retinal disease in digital color photographs of the retina. Together with the vasculature, the optic disc and the fovea are the most important anatomical landmarks on the posterior pole of the retina. In this work, we presented a multi-stage region-based convolutional neural network for optic disc and fovea detection. In the first stage, standard faster-RCNN and SVM were employed for OD segmentation. In the second stage, the relative position information (RPI)-based faster-RCNN was proposed for fovea detection. The experimental result showed the average Euclidean distance with ground truth were 32.6 and 52 pixels for OD and fovea, respectively. The average Jaccard and dice index for OD segmentation were 0.8018 and 0.8873, respectively. The RPI-based faster-RCNN outperformed the standard network.