Visual field tests measure how wide of an area your eyes can see when you focus on a central point. Follow-up tests are compared to the initial baseline test to help assess whether there has been a reduction in the visual field, which can be indicative of glaucoma worsening. But visual field “worsening” remains notoriously difficult for physicians to detect, even with visual field testing.

Complicating matters, glaucoma is a very unpredictable disease: It progresses slowly in some patients, while in others it progresses very quickly. And when it comes to glaucoma, early diagnosis and treatment can preserve vision.

But what if a single, initial visual field test could help predict which patients are at risk for rapidly progressing glaucoma? That was the focus of a study by Jithin Yohannan, an assistant professor of ophthalmology at Wilmer Eye Institute, Johns Hopkins Medicine. That research, published last year in the journal PLOS ONE, showed that by using deep learning and other machine learning algorithms constructed by Yohannan and his colleagues, a computer could predict patient risk for rapidly progressing glaucoma with over 90% accuracy in some 210,000 visual field tests.

Since then, Yohannan has been researching what happens to the machine’s predictive abilities if it is also provided with information from optical coherence tomography scans along with clinical information such as intraocular pressure, age, gender and visual acuity testing. The short answer? Giving the machine this additional data improves its performance significantly.

The team presented the results of this research at the 2022 American Glaucoma Society meeting and at the 2022 annual meeting of the Association for Research in Vision and Ophthalmology. The next step, says Yohannan, will be to validate the findings using a different dataset, but he expects that such machine learning methods to forecast glaucoma risk may be in use within the next five to 10 years.

He and his colleagues are also working on a model that could predict the risk of needing glaucoma surgery, which could inform both the counseling of patients and the need for ophthalmologists to refer patients to a surgical glaucoma specialist earlier.

Yohannan says the unique thing about these tools is that they can predict future developments. And being able to predict what’s going to happen means being able to intervene before damage occurs. Knowing which patients are at risk could also help to inform clinical trial enrollment — which could speed up the process of identifying treatments not only for glaucoma, but for a wide variety of diseases.