Convolutional neural network-based differentiation of intraocular inflammatory cells with ultrahigh-resolution OCT

The differential diagnosis of uveitis is broad and often challenging. A key indicator of intraocular inflammation is the presence of cells in the aqueous or vitreous humor. We developed a lightweight convolutional neural network (CNN) model capable of classifying intraocular inflammatory cell types using ultrahigh-resolution optical coherence tomography (OCT). The model was trained and optimized using OCT images of known cell types-mononuclear cells and granulocytes-and achieved an accuracy of 88.4 ± 0.6%, with an area under the receiver operating characteristic (ROC) curve of 94.3 ± 0.4%. The inflammatory cell compositions predicted by the model were consistent with clinical diagnoses in uveitis patients. This approach offers a promising tool for the diagnosis and monitoring of intraocular inflammation in uveitis.