Part one dealt with the wet form of age-related macular degeneration and how researchers are looking for ways to identify and assess what disease stage the patient is in and develop the correct treatment plan. Usually, the first signs that someone has age-related macular degeneration are visible via fundus photos. How can they be utilized for the diagnosis and treatment of age-related macular degeneration? It has to do with something that has been in the news a lot lately, namely Artificial Intelligence.
AI & Fundus Photos
Fundus photos shows the progression of age-related macular degeneration. Yet, interpreting them takes a great deal of clinical experience and there are only a few thousand retinal specialists in the U.S. On top of all of that, there can a lot of variability in how the images are construed. The reason for that has to do with the fact that patients fall along a spectrum where it is hard to categorize the risk. So, what can be done to interpret the fundus photos in order to assess when to start appropriate treatment?
Enter an artificial intelligence/machine learning (AI/ML)-based system. This system is being developed by researchers with a funding from National Eye Institute (NEI). It will not only screen for age-related macular degeneration, it will also predict which patients will progress to the late stage of the disease within two years and it will evaluate a patient’s risk for developing the late wet (neovascular) version of the disease from one’s risk for developing the late dry (geographic atrophy) version.
This is an important distinction since the treatment approaches for these two versions are very different. For example, the wet version needs to be diagnosed quickly, since a delay in getting receiving the anti-VEGF therapy can lead to a poor outcome.
This system was developed by Alauddin Bhuiyan, PhD, the chief scientist and founder of iHealthScreen. He calls his system iPredict. It is an autonomous system, meaning that it can deliver both clinical and diagnostic information on its own, without the need for expert interpretation. The system obtains information from patient retinal images. (Bhuiyan trained iPredict using over 90,000 color fundus photos from the NEI funded Age-related Eye Disease Study (AREDS).) This information includes the size and shape of drusen (small yellow deposits made up of fats and proteins) and pigment abnormalities. The data is combined with other information, such as patient age, smoking status, genetic profile and disease outcome. With all this data, the system can detect patterns that can help with screening and disease prediction.
In under a minute, the iPredict system generates a report classifying the patient as referable or not for age-related macular degeneration, along with recommendations to visit an ophthalmologist. It also generated a prediction score on a 0 to 100 percent scale that quantified the patient’s risk for developing late age-related macular degeneration within a year. Using the AREDS dataset, iPredict predicted the progression to late age-related macular degeneration with 86 percent accuracy.
If all of that wasn’t good enough, iPredict is currently being studied in non-specialist settings in the New York City area. The plan is to market this to primary care practices by the end of 2023, once it gets Food & Drug Administration (FDA) clearance.
Once again, research shows both that there are ways to treat age-related macular degeneration that can help to preserve vision for as long as possible and that technology can be utilized to diagnose age-related macular degeneration with a high degree of accuracy.