Unlocking anterior segment research with the world's largest front-of-eye dataset
Anterior segment conditions affecting the front of the eye, including cataract, keratoconus, and loss of corneal transparency, are leading causes of blindness globally, yet fewer than 10% of ophthalmic imaging datasets include them. In response, INSIGHT has curated a landmark cornea anterior segment dataset, CADMUS.
Comprising nearly 1 million eye images with rich associated metadata from 22,482 patients, CADMUS can support research across several domains:
- Disease progression: longitudinal imaging enables analysis of how conditions evolve over time
- Multimodal predictive modelling: combined image and clinical metadata support predicting outcomes such as the need for surgery or treatment response
- AI/ML validation: provides an external test bed to validate and generalize algorithms trained elsewhere
- Health equity research: diverse data allows for the study of demographic bias (e.g., age, ethnicity, comorbidities) in device/algorithm performance
CADMUS was developed by clinical researchers at Moorfields Eye Hospital, London and UCL Institute of Ophthalmology, led by ophthalmic surgeon and NIHR Doctoral Fellow Shafi Balal and consultant ophthalmic surgeon Bruce Allan. The development work was supported by European Society of Cataract and Refractive Surgeons Digital Research funding awarded to Bruce. The data was collected across 514,524 appointments at Moorfields in multiple subspecialty clinics including external disease, contact lens, glaucoma, keratoconus, and refractive services.
Reflecting the demographic diversity of the Moorfields patient population, the dataset has broad representation across ethnicities and age groups typical of London and the UK catchment area. It also spans a wide spectrum of anterior segment pathology and systemic comorbidities, from keratoconus and cataract to corneal dystrophies, hypertension, and diabetes.
Watch Shafi Balal and Bruce Allan discuss the CADMUS dataset, below.
