Hybrid Framework for Explainable Glaucoma Detection: Integration of U-Net, DeepLabV3+, and Semi-Supervised Learning

View at publisher Summarise with VIRAT
Authors: Neha Dhavale, Arvind Kumar Sharma
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In this paper, we present an image based hybrid methodology which is combination of convolution methodologies for diagnosing glaucoma via retinal fundus images. It combines U-Net and DeepLabV3+ for multiscale segmentation with strong consistency, and a semi-supervised classifier involving Shepard Convolutional Neural Network (ShCNN) and Deep Kronecker Network (DKN) for accurate and robust classification in the presence of labeled and pseudo-labeled data. The hybrid combination obtains better generalization and robustness rather than individual models. It remedies a number of generic challenges, including dataset noise, scarce labels, and clinical explainability. The proposed framework reaches the accuracy, sensitivity and specificity of 98.7%, 98.2% and 99.1% respectively, performing superior than the original models of CNNs and SVMs. Explainable AI methods such as Grad-CAM also offer clinical interpretability to be safe for routine use.

Fields of Study & Subjects

Retinal Imaging and Analysis Ophthalmology and Visual Health Research Advanced Neural Network Applications

Publication details

Journal / source
CRC Press eBooks
Pages
406–413
Publication date
25 March 2026
Language
EN
Document type
Book chapter