Advanced Glaucoma detection using CNN models integrated with random forest classifier for enhanced diagnosis

Glaucoma is a prevalent retinal disease worldwide, leading to irreversible blindness if not diagnosed early. This research aims to distinguish glaucomatous eyes from normal eyes using an innovative glaucoma detection framework that integrates a Random Forest classifier with VGG-19 and ResNet50 models. The proposed method aggregates classification outcomes from Random Forest, VGG-19, and ResNet50 models through a post-processing decision block to deliver the final classification. This approach addresses limitations of existing techniques, significantly enhancing diagnostic accuracy and robustness by leveraging the strengths of these models. The framework, developed through ensemble modeling, achieves an accuracy of 96.10%, a precision rate of 99.10%, a recall rate of 89.00%, and an F1-score of 94.20%. Our research underscores the importance of integrating conventional optic disc measurement methods with image shape-derived features in machine learning models, enabling early and precise glaucoma diagnosis.