A multimodal paired slit-lamp photography and anterior segment optical coherence tomography dataset for anterior chamber biometry and cross-modality modelling

Abstract We describe a multimodal ophthalmic dataset comprising paired slit-lamp photographs, anterior segment optical coherence tomography (AS-OCT) images, and linked eye-level clinical and biometric measurements collected at a single tertiary eye hospital. The release contains 286 de-identified participants and 562 eye-level records, including 15,582 slit-lamp photographs and 9,092 AS-OCT images. Openly downloadable resources include de-identified tabular data, a data dictionary, and documentation, whereas imaging files are distributed through a separate controlled-access Zenodo record because ocular images may retain residual re-identification risk. Records are linked across modalities using non-meaningful patient_id and eye identifiers. We provide dataset-level descriptive statistics, repository structure, de-identification procedures, and quality-control checks, together with a compact baseline experiment showing that paired slit-lamp and AS-OCT records contain usable cross-modal signal for anterior chamber depth prediction. This resource is intended to support reproducible benchmarking, cross-modality modelling, and secondary research use under transparent access governance and versioned release practices.