The differentiation between Immunoglobulin G4-related ophthalmic disease (IgG4-ROD) and orbital lymphoma, particularly the mucosa-associated lymphoid tissue (MALT) subtype, presents a significant clinical challenge due to overlapping imaging features and similar presentations. Recent advances in artificial intelligence (AI), particularly radiomics and deep learning, have shown promising potential in enhancing diagnostic accuracy by extracting high-dimensional imaging features and constructing robust predictive models. This review systematically examines the current state of AI applications in distinguishing IgG4-ROD from orbital MALT lymphoma, highlighting key methodologies in image-based feature extraction, model development, and diagnostic performance evaluation. We explore various AI techniques applied to multimodal imaging data integration and discuss optimization strategies for deep learning architectures tailored to this clinical context. Additionally, the review addresses the practical challenges and limitations of translating AI-assisted diagnostic tools into routine clinical practice, including issues related to small sample sizes, retrospective single-center designs, data variability, interpretability, and the critical need for robust external validation. By synthesizing recent research findings, this review aims to provide a comprehensive overview of AI-driven diagnostic advances, critically assess current challenges, and propose future directions to improve the accuracy and reliability of orbital disease differentiation, ultimately supporting more precise clinical decision-making.
Weng et al. (2026) studied this question.