Local Binary Patterns and Modified Red Channel for Optic Disc Segmentation

Glaucoma is one of the ocular eye diseases that can cause gradual vision loss and permanent blindness if it is not treated in the early stage. Current screening test such as intraocular pressure (IOP) assessment is not efficient since eye pressure is not the only symptom of glaucoma. The most suitable assessment of the glaucoma is by analyzing the health of the optic nerve head. In order to quantify the severity level of glaucoma, an automated detection system is developed by examining the optic disc and optic cup size. This paper explores two methods for optic disc segmentation, a part of modules in automated detection system, which are local binary patterns (LBP) and modified red channel (MRC). Both methods utilized only the red channel of RGB format fundus image as it alone is enough to achieve good performance in term of image contrast as compared to the other channels. For each method, preprocessing is first performed to enhance the quality of the input fundus image and post-processing is performed to smooth out the segmented boundary of the optic disc. RIM-One database is used to validate the simulation results for both tested methods. The results show that MRC performance is more stable in wider conditions compared to LBP. In conclusion, both methods segment the optic disc boundary with high accuracy, which can be used to calculate cup-to-disc ratio to determine severity of glaucoma.