RBVS-Net: A Robust Convolutional Neural Network for Retinal Blood Vessel Segmentation

Retinal vascular diseases are the utmost cause of visibility loss and blindness where the blood vessels in the eyes somehow fail to circulate the appropriate level of blood flow. Early and correct detection of retinal blood vessels facilitates humans to take expedient remedy against most of the ophthalmic diseases which can significantly reduce possible vision loss. This paper presents a robust RBVS-Net (Retinal Blood Vessel Segmentation Network) which is inspired by the popular U-Net architecture. Proper utilization of transfer learning and data augmentation lead RBVS-Net to achieve to outperform the state-of-the-art accuracy. Extensive experiments have been conducted on three benchmark retinal fundus image datasets, where the proposed algorithm achieves more than 96% average accuracy for vessel segmentation. A comparison with other recent works also demonstrates the efficiency of the proposed approach to segment the blood vessel from the retinal color fundus image.