Deciphering age-related macular degeneration by deep phenotyping and machine learning

Age-related macular degeneration (AMD) is a very common cause of blindness. It is difficult to predict who will progress to the stage of the disease where eyesight is threatened as some patients progress slowly or not at all and others progress quickly.

We can teach computers to analyse high resolution images of the inside of the eye. We have access to hundreds of thousands of such images from patients with AMD as well as those who don’t. These images will allow us to train computers to identify what eye changes appear in patients with AMD. Once the computers have learnt this, we expect they will identify new changes.

This approach will enable us to better predict the patients who are most at risk of sight loss. This should help us develop better treatments and enter the most appropriate patients into clinical trials. It should also allow us to better understand why AMD develops.