An Automated Glaucoma Image Classification model using Perceptual Hash-Based Convolutional Neural Network

Abstract

Generally, identification of Glaucoma in color fundus images is a crucial process, which needs more knowledge and experience. An efficient spatial hashing-based data structure for facilitating the investigation of 3D shapes by the use of CNN. This model makes use of the sparse occupancy of 3D shape boundary and constructs the hierarchical hash tables for an input model under dissimilar resolutions. This paper designs an automated Glaucoma image classification model utilizing Perceptual Hash-Based Convolutional Neural Network (PH-CNN) model. The presented classification model operates in different stages namely feature extraction, feature reduction and classification. Initially, feature extraction process takes place via Discrete Wavelet Transform (DWT). Next, selection of features or reduction of features is carried out by the Principal Component Analysis (PCA) technique. Finally, PH-CNN model is applied for the classification of Glaucoma images. For validating the effective results of the presented PH-CNN approach, a benchmark dataset is applied and the results are assessed under several dimensions. These maximum values attained from the experimentation indicated that the projected model can be applied to diagnose the Glaucoma disease in real time.