Glaucoma Identification Using Deep Convolutional Neural Networks

Abstract

Glaucoma is a leading cause of irreversible blindness, and its early detection is crucial for effective management and treatment. In this study, we present a novel approach to glaucoma identification that leverages Deep Convolutional Neural Networks (DCNN) to enhance diagnostic precision beyond traditional CNN methods. Our methodology integrates several advanced steps: initially, retinal images from a publicly available Kaggle dataset are pre-processed using a median filter to reduce noise and enhance image quality; subsequently, the Fuzzy C-Means (FCM) algorithm is employed for precise segmentation of the retinal regions. Texture features critical for glaucoma detection are then extracted using the Gray Level Co-occurrence Matrix (GLCM). These features are fed into the DCNN, which automatically learns hierarchical representations for accurate classification. Experimental results demonstrate significant performance improvements over conventional CNN approaches. For malignant cases, the proposed method achieved an accuracy of 94.53% compared to 91.48%, with corresponding enhancements in specificity (89.30% vs. 86.96%), sensitivity (91.91% vs. 89.22%), and precision (90.61% vs. 88.09%). Similarly, for benign cases, the DCNN approach achieved an accuracy of 94.93% versus 88.81%, with specificity, sensitivity, and precision improvements as well. These findings underscore the potential of DCNNs in advancing the early detection of glaucoma, thereby facilitating timely and more effective clinical interventions.