AI used to predict retinal disease progression

A new study published in Nature Medicine has shown that artificial intelligence (AI) performs as well as, or better, than clinicians in predicting whether a patient with age-related macular degeneration (AMD) will develop the more serious exudative AMD (exAMD) within the next six months.

The study’s authors made use of the Moorfields Eye Hospital AMD dataset of retinal images to create a bank of anonymised retinal scans from patients with exAMD in one eye and at high risk of developing it in the other.

An AI system comprised of two deep convolutional neural networks was then fed high-dimensional volumetric eye scans, which the system then segmented into thirteen anatomical categories. Overlaid onto raw scans, the segmented images were then fed into the prediction model.