Artificial Intelligence and Ophthalmology

I recently read an article by Rui Fan and colleagues about the detection of glaucoma by a computer in patients with high eye pressures.

By way of background, the internal pressure of the eyeball is typically somewhere in the range of 10-20 mmHg above atmospheric pressure. When the pressure in the eye goes up, we call it ocular hypertension, similar to “general” hypertension where the pressure of blood in the arteries of the whole body is elevated, which is what is measured in a typical blood pressure check. Having ocular hypertension is a risk factor for damage to the optic nerve. Often, when that damage occurs, we call it glaucoma.

However, not all eyeballs that have higher than typical pressures develop glaucoma. There is a famous study that attempted to investigate the implications of ocular hypertension and its treatment as far as development of glaucoma is concerned, which was called the Ocular Hypertension Treatment Study (OHTS).

In the article that I read by Rui Fan, automated deep learning algorithms were trained on photographs from OHTS. The authors suggest that the artificial intelligence model that was developed had high diagnostic accuracy in detecting glaucoma from optic nerve photographs.

This is one of many ways in which artificial intelligence models are being used on images of eyes. If computers can aid in this kind of diagnosis, there are many implications, one of which is that computers can be used widely to obtain information about eyes of the large number of individuals who are not regularly getting eye exams from eye doctors. Hopefully, this will help us in future to decrease the global burden of eye disease by getting patients who need eye care to providers in a more efficient manner.