Detection and Classification of Glaucoma in the Justraigs Challenge: Achievements in Binary and Multilabel Classification

Philippe Zhang, Yihao Li, Jing Zhang, Weili Jiang, Pierre-Henri Conze, Mathieu Lamard, Gwenolé Quellec, Mostafa El Habib Daho

Abstract: Glaucoma is a significant global health issue that may lead to irreversible blindness if left undetected. This condition affects millions worldwide, necessitating improved diagnosis strategies for early detection and intervention. This paper details the LaTIM team’s approaches in the JustRAIGS challenge, addressing the two proposed tasks: the binary classification of referable versus non-referable glaucoma (Task 1) and the multi-label classification of ten additional disease-related ocular features (Task 2). By leveraging state-of-the-art pre-trained models and employing advanced machine learning techniques, we have developed methodologies that enhance the diagnostic accuracy for glaucoma. These methods not only aim to identify the presence of glaucoma with high reliability but also strive to classify its severity through the detailed analysis of additional ocular indicators.