Machine Learning and Convolutional Neural Network (ML-CNN) Based System for Early Identification of Glaucomatous Optic Neuropathy

Abstract

Glaucoma, a leading cause of irreversible blindness, often goes undetected in its early stages due to the absence of symptoms and limitations of traditional diagnostic methods like manual cup-to-disc ratio (CDR) assessment. Early and accurate detection is essential to prevent permanent vision loss, yet conventional techniques are time-consuming and error-prone. This study proposes a hybrid machine learning framework that integrates Convolutional Neural Networks (CNN) with Support Vector Machines (SVM) for automated glaucoma detection using retinal fundus images. The images are pre-processed and segmented to isolate the optic disc and cup, enabling precise CDR calculation. High-level features extracted by the CNN are classified by the SVM into normal or glaucomatous categories. To enhance model robustness, data augmentation and transfer learning are employed, addressing challenges such as limited datasets and overfitting. The proposed framework demonstrates high accuracy and scalability, offering a reliable and efficient solution for early glaucoma screening and real-world deployment in ophthalmology.