A Datasheet for the INSIGHT Moorfields Cornea Anterior Segment Dataset (CADMUS).

PURPOSE: To describe Cornea Anterior Segment Dataset from Moorfields via INSIGHT (CADMUS), a large multimodal anterior segment imaging dataset developed through the Eye and Oculomics Health Data Research Hub (INSIGHT) Health Data Research Hub to support corneal research and artificial intelligence (AI) development.

DESIGN

A dataset description.

PARTICIPANTS

Anonymized anterior segment imaging and linked clinical metadata from 22 482 patients attending Moorfields Eye Hospital National Health Service (NHS) Foundation Trust between December 2019 and September 2024 (refreshed every 2 years).

METHODS

The CADMUS was derived from routine clinical care using INSIGHT infrastructure. It comprises raw Digital Imaging and Communications in Medicine MS-39 Placido images, OCT scans, grayscale photographs, derived tomography indices, and linked electronic health record metadata including demographics, diagnoses, procedures, clinic activity, refraction, visual acuity, and follow-up. Data were pseudonymized, standardized, and curated within the INSIGHT Secure Research Environment under established governance and ethical approval.

MAIN OUTCOME MEASURES: Dataset scale, multimodal integration, longitudinal follow-up, disease representation, surgical data capture, and availability of linked imaging and clinical metadata.

RESULTS

The CADMUS includes 945 243 image acquisitions from 22 482 patients, with 278 101 tomography records, 514 524 clinic visits, and 40 471 surgical records from 12 079 patients. Additional data include 52 463 refraction measurements (13 235 patients) and 543 902 visual acuity records (22 368 patients). Longitudinal follow-up is present in 96.1% of patients (mean 6.1 visits). The dataset captures diverse anterior segment pathology, including keratoconus (42 194 records), nuclear cataract (11 800), corneal scar (3,430), Fuchs' dystrophy (2652), and corneal decompensation (1264). Common procedures include phacoemulsification (8209), corneal cross-linking (7115), Descemet membrane endothelial keratoplasty (1739), penetrating keratoplasty (1308), and laser superficial keratectomy/phototherapeutic keratectomy (1189).

CONCLUSIONS

The CADMUS is a large, longitudinal, multimodal anterior segment dataset linking imaging with detailed clinical metadata in a secure NHS environment. It provides a robust platform for reproducible research, longitudinal modeling, surgical outcomes analysis, and AI development in anterior segment disease.

FINANCIAL DISCLOSURES: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.