Moorfields Eye Hospital and University College London have published a dataset of nearly one million clinical eye images linked to patient records, with the goal of supporting AI development and research into anterior segment conditions. The release is notable because conditions affecting the front of the eye, including cataracts, are among the leading causes of blindness worldwide, yet they are significantly underrepresented in existing research datasets.

Despite their clinical importance, fewer than 10% of ophthalmic imaging datasets currently cover anterior segment conditions. The CADMUS dataset is designed to fill that gap, giving researchers access to rich, longitudinal data that tracks how eye conditions change over time, which is something most imaging datasets simply do not offer.

The dataset is already generating results. Researchers have used it to train deep learning models, one of which can predict a patient’s age and biological sex from routine anterior segment scans, suggesting that standard clinical images contain biological signals that go undetected by the human eye.

What the dataset contains

CADMUS holds 945,243 images from 22,482 unique patients, collected at Moorfields between December 2019 and September 2024. Because the data includes follow-up visits, it allows researchers to track disease progression and long-term outcomes over several years. The dataset includes three main types of information:

• Raw DICOM images from the MS-39 anterior segment OCT tomographer
• Derived measurements including keratometry values, pachymetry, wavefront aberrometry, and AI-generated classifier scores for keratoconus and related conditions
• Linked electronic health record data covering demographics, diagnoses, and clinical history

Lead author Shafi Balal said the data has already been used to establish precision limits for measuring keratoconus progression, giving researchers a clearer baseline for understanding how the condition develops. The age and sex prediction model further demonstrates that routine clinical scans carry more information than previously assumed.

How researchers can access the data

Access to CADMUS is managed through INSIGHT, Moorfields’ Eye and Oculomics Health Data Research Hub. Researchers must submit a Data Use Application, which is reviewed with oversight from an independent patient and public advisory board. The process also applies the internationally recognised Five Safes framework, which evaluates:

• Safe projects
• Safe people
• Safe data
• Safe settings
• Safe outputs

Moorfields hopes the dataset will support research into earlier disease detection, surgical outcome prediction, and the development of AI diagnostic tools. The full datasheet is published in Ophthalmology Science and available at doi.org/10.1016/j.xops.2026.101203.

The broader push to make health data useful

The CADMUS release is part of a wider trend in the UK and Europe toward structured, governed health data sharing. Several parallel developments point in the same direction.

The European Innovation Council and SMEs Executive Agency recently named three winners from an open call for projects that make health data usable and interoperable across fragmented systems. The initiative, part of the EU-funded UNITE Regional Innovation Valley project, received proposals from more than 1,000 organisations including universities, startups, hospitals, and healthcare providers. The three selected projects will move into an implementation phase in spring 2026.

In London, social enterprise consultancy PPL has released a free tool called the London Neighbourhood Public Data Explorer. It lets users compare neighbourhoods on a map, explore population health indicators by theme, and export reports in PDF format, aimed at supporting neighbourhood-level health planning across the capital.

Cambridge University Hospitals NHS Foundation Trust has also welcomed the launch of the Zenith supercomputer, funded by the Department for Science, Industry and Technology. Zenith is built to process health data at a scale not previously available to NHS researchers. The AI Centre for Value-Based Healthcare is working with the project to make sure the system is used responsibly and securely as it is applied to developing AI tools for patient care.