Glaucoma Retinal Image Detection through the Segmentation of OD Using Modified Deep Learning Method

Classifiers are the important processing module in any type of classification systems. This paper uses the proposed Modified LeNET (MLNET) classification architecture along with the standard LeNET to classify the retinal pictures into healthy cases and cases of glaucoma. This research work develops an automated computer aided system which has the following modules as preprocessing, Optic Disk (OD) segmentation, Feature computations and MLNET classification. The Glaucoma classification system has been functioned in two processing phases as training and testing. The training processing phase trains both healthy and Glaucoma retinal images from the known dataset using preprocessing, OD segmentation and feature computations from the segmented OD region. These features from the OD region have been further trained by the proposed MLNET classifier. The testing processing phase tests the unknown retinal image into either Glaucoma or healthy class through the sub processing modules of preprocessing, OD region segmentation and feature computations. The features from the OD region in the unknown test retinal image have been fed into the proposed MLNET classifier with respect to the previous training results.