AI's Foresight: Predicting the Course of a Degenerative Eye Disease

Artificial intelligence could soon help clinicians predict the progression of keratoconus, a condition where the cornea thins and bulges outwards. A systematic review of ten studies, covering 10,940 eyes, found that AI models can forecast the disease's course with accuracies ranging from 77.5% to 84.9%.

These algorithms, which use machine and deep learning, consistently identified younger age and specific corneal shape metrics as key predictors. The modelling performance improved when clinical risk factors, such as eye rubbing habits and deficiencies in vitamins D and B12, were included in the classification framework.

However, the technology is not yet ready for the clinic. The review highlights a critical weakness: none of the studies validated their models using independent, multi-centre data. Until there are standardised definitions for progression and more rigorous, real-world testing, these promising tools will remain in the research phase.