An Automated Glaucoma Detection Model to Estimate Glaucoma Abnormalities in Fundus Images Using CNN

Abstract

An eye situation glaucoma is analysed effectively with the damages occurred in the optic nerve of the eye and it will become for a long period. This becomes proactive in successive stages mainly pressure built up in the optic nerve inside the eye. The current Scenario state that most predominant glaucoma type is of primary open angle glaucoma (OAG) management Glaucoma cannot be identified easily until it becomes more serious in life. Most of the glaucoma type is of primary open angle glaucoma management. An attention is made on one of the eye diseases conditionally, glaucoma which hurts the optic nerve in addition after some time emerges as being severe. The condition is induced internal to the attention with the aid of way of the increase of intraocular stress, the internal tissue layer structure of a retina acts as foreground of a watch that senses light and sends photographs to the mind. This state of affairs will display up salient and might not display in some other time right away. If the disorder is detected early then it prevents field of vision distress. Primary Open Angle Glaucoma Administration is certainly one of the maximum essential as well as most difficult factors of the glaucoma discovery. The numerous findings which depend upon clinical analysis of glaucoma are intraocular pressure, visual view loss and also optical nerve cup. This incurred visual loss growth rate tremendously increases that is purely caused due to open anterior angles, visual abnormal characteristics identity and intraocular pressure of the eye. Glaucoma discovery can be determined through numerous methods: perimetry, tonometry, OCT, ophthalmoscopy, tachymetry and genotype. In this work, the author developed a novel algorithmic model to detect and provide a suitable diagnosis approach for glaucoma identification in the patients using 50 colour fundus images very accurately. In this proposed method Convolutional Neural Network (CNN) was targeted for segmentation and classification of the input fundus images to detect glaucoma by measuring cup and disc values. CNN algorithm is taken, and analysed to classify glaucoma images. The final effects of reminiscence are fed to an average clear out for pre-processing and wavelet sub-band decomposition for extraction of attributes that are fed to CNN classifier. Two ways the simulations were performed, one using GUI based simulation environment and second one the metrics evaluation using statistical properties. In this proposed work approximately 100% accuracy and CDR 0.80 values have been achieved with CNN novel approach for glaucoma detection using CDR values for 50 database eye fundus images obtained from the hospitals.