Fundus Image-Based Glaucoma Classification using Feature Extraction and Classification Techniques

This research work presents a glaucoma classification method using structural and texture-based characteristics extracted from retinal fundus images. The optic disc and optic cup are typically segmented from retinal fundus images using adaptive thresholding, by processing the RGB color channels to extract key components. The Segmentation algorithm leverages statistical attributes such as mean and standard deviation to capture pertinent details from the red and green channels. Consequently, it generates images that isolate the optic nerve head zone in both channels. Subsequently, segmentation of the optic disc and cup is performed independently from the red and green channels, respectively. Additionally, the study addresses the challenges posed by glaucoma, a leading cause of blindness globally, and presents a novel algorithm for glaucoma diagnosis using retinal image analysis. Structural features such as the Rim- Width Ratio (RMW) and CDR along with GLCM (texture features), are utilized to classify retinal fundus images. The study includes a comparative analysis of classification algorithms, including k-Nearest Neighbors (k-NN), Random Forest (RF), and Support Vector Machine (SVM), based on metrics such as accuracy, F1, precision, and sensitivity. Overall, the proposed methodology offers the potential for enhancing glaucoma screening and diagnosis through automated image analysis and classification.