Hybrid Approach for Effective Segmentation and Classification of Glaucoma Disease Using UNet++ and CapsNet

Abstract

Glaucoma is an incapacitating eye disease that can result in total vision loss if left untreated, making early treatment through screening more likely to prevent irreversible visual impairment and delay. Unfortunately, however, due to complex testing procedures and healthcare professional shortages often resulting in delays- contributing to an increasing global incidence of blindness. Also, with an estimated 76 million people affected by glaucoma worldwide, recent statistics highlight the need of addressing this prevalent condition. To combat these challenges and enhance manual methods further it is urgently required that a reliable framework be created for early detection of lesions on Optic Cup and Optic Disc with characteristics overlapping with variations in eye colour making accurate diagnosis even more challenging. At present, we present an automated system for the detection of Glaucoma. The model we present begins by pre-processing retinal images using advanced techniques like histogram equalization and Contrast Limited Adaptive Histogram Equalization (CLAHE), designed to improve image quality and analysis. Next U-Shape Network technique (UNet++) used for discs and cups segmentation separately; this segmentation technique offers accurate identification. Segmented optic disc and optic cup images are then utilized in our glaucoma diagnosis stage. We employ Capsule Network, a deep learning technique renowned for recognizing complex patterns, to detect glaucoma and it handles of complicated spatial relationships in an effective manner, solving such restrictions that are present in CNN. Comprehensive performance evaluation metrics were devised and 5-fold Cross validation was performed to verify our method, which demonstrated its superiority over existing methods with an overall accuracy rate of 97.89%. This high rate of precision highlights how automated framework can detect early signs of glaucoma for early intervention as well as reduce risks such as irreversible loss of vision for patients.