GlaucoFusionNet: A Causal Inference Aware Multimodal Explainable AI Framework with Counterfactual Reasoning for Glaucoma Risk Prediction
Year: 2026
Type: preprint
Abstract: Glaucoma is one of a top global cause of irreversible vision loss. But timely diagnosis and thorough risk stratification are the keys to interrupting the progression of blindness. But they are correlation-driven, lack causal reasoning, and provide limited interpretability for clinical decision-making. Furthermore, conventional multimodal approaches are combined retinal imaging and clinical data for diagnosis purpose. But they are not addressing the confounding effects and provide actionable explanations for disease progression. To overcome limitations, study proposes GlaucoFusionNet based on a multimodal explainable AI framework with causal inference for automated glaucoma risk prediction and progression analysis. GlaucoFusionNet is designed for images and tabular data. It integrates retinal fundus images and structured clinical data through a feature fusion architecture. ResNet-based CNN extracts deep retinal image representations. An MLP network model is implemented to accumulate the nonlinear relationships among clinical variables. The extracted embeddings are fused to construct multimodal features for glaucoma classification. SHAP provides feature-level explanation and Grad-CAM visualizes prediction of model on images. Directed Acyclic Graphs used to identify meaningful relationships between clinical factors with reduced spurious correlations. Beyond classification, the framework provides a risk stratification score for identifying high risk patients for disease progression.
Source: Research Square
Language: English
Topic: Retinal Imaging and Analysis
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7210263142