Enhanced glaucoma detection using ensemble based CNN and spatially based ellipse fitting curve model

Glaucoma is one of the leading causes of blindness for people over the age of 60. For the automatic glaucoma detection using fundus images, several approaches have been recently developed. However, the extraction of optic cup boundary considered as critical work due to the blood vessels interweavement and also the significant challenges in accurate localization of affected optical disc based on size of the cup which is either too small or large. To overcome this challenges efficient technique needs to be considered. To accurately detect the glaucoma and correspondingly the optic cup OC and optic disk OD boundary the ensemble based CNN classification with spatially based ellipse fitting curve model (SBEFCM) has been proposed. In this study, the glaucoma or diabetes retinopathy DR classification has been performed by the ensemble CNN classification. The optic disc and optic cup boundary detected by new spatially weighted ellipse fitting model SBEFCM. For the proposed SBEFCM, which enhances and prolongs the multi-ellipse fitting technique. The ensemble based CNN classification is performed for various datasets such as LAG, RIM-ONE and local retinal glaucoma dataset. Previously the input fundus image is pre-processed and further blood vessel segmentation performed to avoid the interweavement with surrounding tissues and blood vessels. The new SBEFCM accurately detected the ratio of OC and OD boundary prediction which are depicted and by using the ensemble CNN classification, various performance metrics such as sensitivity, specificity, accuracy and AUC values has been identified for glaucoma detection and by comparison the proposed ensemble CNN outperforms the existing methods.