[Uveitis diagnosis: Current approaches and future perspectives]

Uveitis encompasses a heterogeneous spectrum of etiologies, including systemic inflammatory diseases, infections, specific ophthalmologic entities, and pseudo-uveitis. The diagnostic approach relies on an algorithmic strategy aimed at the early identification of a curable etiology and/or an underlying condition suitable for targeted therapy, while taking into account the severity and relative frequency of the various causes. The prospective ULISSE study showed that a standardized diagnostic approach based on the anatomo-clinical classification of uveitis achieved a diagnostic yield comparable to unrestricted testing, while substantially reducing both the number and cost of investigations. Etiological diagnosis is primarily based on anatomo-clinical analysis of uveitis, integrating the site of inflammation, disease course, laterality, and specific ophthalmologic features. Subsequently, consideration of epidemiological and demographic factors, medical history, and extra-ophthalmologic clinical examination allows further refinement of the diagnostic workup. Complementary investigations should be prioritized and targeted, except for a minimal systematic baseline evaluation, as the diagnostic yield of non-directed testing is very low. Clinical decision-support tools based on artificial intelligence (AI) represent a major advancement. Machine learning models integrating clinical, ophthalmologic, laboratory, and imaging data available from the initial consultation achieve etiological prediction performances comparable to those of experts with access to a complete diagnostic workup. These multimodal AI models could transform diagnostic strategies by enabling a personalized etiological assessment based on the prioritization and dynamic adaptation of complementary investigations. However, their implementation in routine clinical practice will require prior medico-economic validation through controlled comparative studies.