RFENet: Recurrent Feature Extraction Module with Attention for Retinal Vessel Segmentation

Abstract Retinal vessel segmentation is essential to assist doctors in diagnosing diabetic retinopathy, macular atrophy, glaucoma, and other ophthalmic diseases. Due to the small size and complex distribution of retinal vessel, traditional segmentation methods are prone to losing details. In order to improve the segmentation accuracy of retinal vessel and obtain the global contextual information in retinal images, this paper proposes a recurrent feature extraction network (RFENet) based model. First, the recurrent feature extraction (RFE) module extracts feature from the deep feature map in a recurrent manner to mitigate the problem of insufficient feature extraction when the model handles medical image segmentation. Next, this module performs a multi-scale fusion of the extracted features to obtain multi-scale features of the image. An attention module is used to further improve the feature extraction capability of the network before montaging the features at each resolution of the encoder with the corresponding features in the decoder. By applying the recurrent feature extraction module to the encoder-decoder structure for vessel segmentation on the color fundus image datasets DRIVE and STARE, experimental results show that the retinal vessel segmentation method based on the recurrent feature extraction module effectively improves the segmentation accuracy.