Abstract
Despite major advances in ocular oncology, early diagnosis and risk stratification of uveal melanoma (UM) remain complicated, particularly in cases involving small or indeterminate lesions. These challenges are exacerbated by limited access to subspecialist care and the constraints of existing diagnostic frameworks. Artificial intelligence (AI) has emerged as a powerful tool in oncology and ophthalmology, capable of analyzing complex imaging, cytologic, and genomic data to support clinical decision-making. In UM, AI models have shown promise in improving lesion classification, predicting metastatic potential, and augmenting post-treatment surveillance. This review examines the current landscape of AI applications in UM, including tools for triage, prognostication, and post-treatment surveillance. We also focus on steps that must be taken to achieve end-stage clinical rollout, addressing critical barriers to implementation, such as model generalizability, explainability, workflow integration, and ethical considerations. Moving forward, the convergence of multimodal data, privacy-preserving model development, and patient-centered innovations may help translate these technologies into real-world practice. By addressing current limitations and aligning AI development with clinical needs, these tools could ultimately support earlier detection, more personalized care, and greater equity in access to specialist-driven management for UM.