An Improved Machine Learning-Driven Glaucoma Diagnostic Framework

Glaucoma is a progressive eye condition that impairs the optic nerve and is often called a 'silent thief of sight.' The increasing number of individuals losing their vision due to this condition is a cause for concern. Research is being conducted to develop computer-aided systems that can help ophthalmologists in the early diagnosis of glaucoma. OD and OC are often compared to calculate the cup-to-disc (CDR) ratio, indicating whether the eye is glaucomatous. This study presents a novel methodology for segmenting the optic disc (OD) and optic cup (OC). We propose a modified version of the ResUNet architecture. ResUNet is a convolutional neural network architecture that combines residual learning and U-Net architecture for semantic segmentation tasks, offering improved performance in capturing fine details and spatial context. Our proposed model's performance was evaluated on the DRISHTIGS publicly available dataset, and an IOU score of 94.5% and 89.1% were achieved for OD and OC segmentation, respectively.