Artificial intelligence in orbital diseases: opportunities, challenges, and future horizons

Orbital diseases-including thyroid eye disease, tumors, inflammation, fractures, vascular lesions, infections, and congenital abnormalities-are uncommon but clinically consequential. Because orbital diseases are heterogeneous and generate multimodal data, artificial intelligence (AI) is well-positioned to support their evaluation and management. We explore the current relationship between AI and orbital diseases and propose future directions. Four questions are addressed. First, why do orbital diseases need AI? Current orbital care remains limited by manual, observer-dependent measurements, multimodal imaging that complicates decision-making, and disparities in access to orbital expertise. Second, why is the orbit technically ready for AI? Orbital diseases generate rich visual and multimodal data; most core tasks are also image-based, anatomically defined, and measurable. Third, what has AI already achieved? Applications are summarized in diagnostic assistance, segmentation, and measurement, ocular sign detection, activity and severity assessment, treatment response prediction, surgical assistance, and medical education. Finally, what are the gaps in orbital AI, and where should the field go next? Barriers related to data scarcity, privacy, and multimodal integration are discussed. For future development, we propose a three-tiered roadmap comprising OrbitBank, OrbitFM, and OrbitAgent. These represent short-, medium-, and long-term directions, progressing from multicenter data infrastructure and privacy-preserving collaboration to multimodal foundation models and, ultimately, clinician-supervised agentic systems for more standardized and precise orbital care.