U-Net Based Optic Cup and Disk Segmentation from Retinal Fundus Images via Entropy Sampling

Accurate identification of optical disk and cup regions plays a vital role in clinically detection of retinal diseases. Diagnosis of severe ophthalmic pathologies can be made by proper detection of optical disk and cup. The most common method used for ocular screening is color fundus image (CFI). From this CFI, the cup-to-disk (CDR) ratio can be calculated correctly after proper detection of disk–cup regions, and this CDR plays important clues for glaucoma detection. This paper presents an entropy-based deep learning approach to perform such accurate segmentation from digital retinal images. The performance of the proposed approach exhibits promising results in comparison to prior existing methods on two dataset Drishti-GS and RIM-ONE V3.