A Classification Model for Glaucoma Grading Using Multi-Modal Image Fusion Strategies

Glaucoma may lead to irreversible blindness, for which timely detection and treatment are significant. In recent years, automated diagnostic methods have been widely used to classify retinal diseases. Most automatic detection methods are based on a single imaging modality, such as fundus or Optical Coherence Tomography (OCT) images. These methods usually reflect the retinal diseases only to a certain extent, and modality-specific information between different imaging modalities have not been utilized. In this work, multi-modal image fusion strategy is introduced into a classification model for glaucoma grading. The proposed method is validated using a public clinical dataset. Experimental results show that the proposed method achieves 0.85 kappa score, which is better than baseline.