To determine classification criteria for tubercular uveitis.

Machine learning of cases with tubercular uveitis and 14 other uveitides.

Cases of noninfectious posterior uveitis or panuveitis, and of infectious posterior uveitis or panuveitis, were collected in an informatics-designed preliminary database, and a final database was constructed of cases achieving supermajority agreement on the diagnosis, using formal consensus techniques. Cases were analyzed by anatomic class, and each class was split into a training set and a validation set. Machine learning using multinomial logistic regression was used on the training set to determine a parsimonious set of criteria that minimized the misclassification rate among the intermediate uveitides. The resulting criteria were evaluated on the validation sets.

Two hundred seventy-seven cases of tubercular uveitis were evaluated by machine learning against other uveitides. Key criteria for tubercular uveitis were a compatible uveitic syndrome, including (1) anterior uveitis with iris nodules, (2) serpiginous-like tubercular choroiditis, (3) choroidal nodule (tuberculoma), (4) occlusive retinal vasculitis, and (5) in hosts with evidence of active systemic tuberculosis, multifocal choroiditis; and evidence of tuberculosis, including histologically or microbiologically confirmed infection, positive interferon-γ release assay test, or positive tuberculin skin test. The overall accuracy of the diagnosis of tubercular uveitis vs other uveitides in the validation set was 98.2% (95% confidence interval 96.5, 99.1). The misclassification rates for tubercular uveitis were training set, 3.4%; and validation set, 3.6%.

The criteria for tubercular uveitis had a low misclassification rate and seemed to perform sufficiently well for use in clinical and translational research.