Artificial Intelligence in the Detection and Risk Stratification of Choroidal Melanoma: A Critical Comparative Synthesis and Future Directions.

The early differentiation of benign choroidal naevi from malignant melanoma remains one of the most nuanced challenges in ophthalmic oncology, with profound implications for patient survival. Conventional diagnostic pathways rely on multimodal imaging and expert interpretation, but inter-observer variability and the rarity of melanoma limit timely and consistent detection. Recent advances in artificial intelligence (AI) offer a promising adjunct to conventional ophthalmic practice. This review provides a critical comparative synthesis of the studies to-date which have looked at AI's use in the detection, risk stratification, and longitudinal monitoring of choroidal melanoma. While early results are promising-with some models achieving an accuracy comparable to expert clinicians-significant challenges remain regarding generalisability, dataset bias, interpretability, and real-world deployment. We conclude by outlining practical priorities for future research to ensure that AI becomes a safe, effective, and equitable tool for improving patient outcomes.