Automated Glaucoma Detection Using Advanced Retinal Imaging and Deep Learning Feature-Driven Classification

Abstract

This research describes an integrated methodology for automated glaucoma identification that makes use of retinal fundus imaging and advanced computational algorithms. The research begins with the diligent collection of a diversified Retinal Fundus Image Database, which includes both normal and glaucomatous cases. Image quality is improved by pre-processing techniques such as Median Filtering and pixel normalization, which are then followed by feature extraction utilizing Local Binary Pattern (LBP) to capture key glaucoma patterns. The use of a Long Short-Term Memory Convolutional Neural Network (LSTM CNN) refines the analysis even further by incorporating spatial and temporal data. Based on learning characteristics, a Support Vector Machine (SVM) classifier improves classification precision. The model's efficacy is evaluated using performance metrics such as accuracy and sensitivity, with segmentation algorithms refining the analysis and concluding in a glaucoma-detected image. This strategy appears to be promising for effective glaucoma screening, early intervention, and vision preservation in at-risk groups.