ObjectiveThe aim of this study was to develop and validate quantitative diagnostic biomarkers and artificial intelligence (AI) models for Behçet disease uveitis (BU). MethodsWe conducted a single-center retrospective study using AI-based algorithms to evaluate OCT angiography (OCTA) fundus scans from 62 BU patients and 61 healthy individuals. ResultsFrom the 837 radiomic features derived from the OCTA scans, the top 20 features were utilized for AI modeling. The neural network (NN) model displayed the highest area under the receiver operating characteristic curve (AUC) of 0.900 in the test set. The integration of clinical data further enhanced the model's performance, leading to an AUC of 0.908 in the temporal validation cohort. ConclusionOur research demonstrates that the radiomic score and the clinical integrated score can independently serve as diagnostic biomarkers for BU for future potential clinical translational research. SignificanceThe AI models we developed show great promise for clinical and translational research, marking progress towards fulfilling the demand for quantitative imaging biomarkers in BU diagnosis. However, further validation and evaluation are crucial to confirm their real-world applicability in clinical settings.