Bruce Allan has published widely in corneal transplantation, cataract, and refractive surgery. He is currently leading the European Society of Cataract and Refractive Surgery (ESCRS) Digital Health Special Interest Group (DH-SIG).

Clinical research questions are answered by analysis of data collected either in clinical trials (prospectively) or in routine clinical practice (retrospectively). As digital transformation and protocol driven clinical care continue to develop, the quality distinction between prospective and retrospective data is becoming less relevant.

Clinical trials research is valuable but has clear limitations. Clinical trials are slow and very expensive. Many are never completed, and data collected is often never re-used. Datasets from clinical trials are also generally too small for analyses using deep learning techniques.

Deep learning has revolutionised pattern recognition and prediction in modern life. Existing applications in ophthalmology include more accurate formulae to guide lens implant selection in cataract surgery (biometry formulae) and automated screening for diabetic retinopathy. Deep learning does not require specialist knowledge for successful application: “code free” deep learning is freely available. The missing piece is the data, particularly for research in the front compartment (anterior segment) of the eye. In a recent Lancet review, less than 10% of all open access ophthalmic image datasets addressed anterior segment eye disorders.  

The ESCRS DH-SIG is aiming to create a library of accurately labelled open access image datasets derived from routine clinical care, democratising access to the data required for deep learning analyses in the surgery and medicine of the anterior segment of the eye.