What area of ophthalmology can benefit the most from AI?
Click Here to Manage Email Alerts
Click here to read the Cover Story, "AI in ophthalmology: From code to clinic."
Glaucoma
I am a retina specialist, so I am obviously partial to retinal diseases. However, I think we have already made a lot of headway in developing machine learning models for identifying diseases in the retina, and we are on our way to do more.
Considering all of the innovations we are seeing in retina, the area I think would benefit the most is glaucoma.
Glaucoma is one of the top causes of vision loss in America, and identifying the condition is paramount. We are not picking it up and not intervening early enough. I can envision a future in which patients can go to their primary eye care providers for their refractive exams and get routine imaging with built-in predictive technology for glaucoma. That technology could flag suspects for closer monitoring or referral to ophthalmologists. That would make great inroads to helping us identify more people with glaucoma earlier in the disease.
The reason we need something like this is because diagnosing glaucoma can be a challenge. It is a multimodal diagnosis, and it is not just one thing that gives a person glaucoma. It is not just high pressure, a visual field defect or the optic nerve appearance. You have to take all of those things into consideration along with factors such as family history, genetics, OCT imaging and a physical exam.
Because of all the important factors that go into diagnosis, creating a machine learning model to identify glaucoma has been challenging. We are, essentially, several years into the AI revolution in ophthalmology, and we have a lot of experience with training these models. I think it can be done effectively for glaucoma.
I am also looking forward to the day when we can choose the right therapy based on AI models. It could tell us what drop to start on, whether we should consider laser or whether someone needs to progress to incisional interventions such as a MIGS procedure or trabeculectomy. With all of these things, AI would be a helpful clinical decision-making tool.
Theodore Leng, MD, MS, is an associate professor of ophthalmology at Stanford University.
Retina
I am a retina specialist, and I am truly excited to harness the power of AI and retinal imaging in the benefit of clinical and scientific insights to better serve our patients.
Image analysis is crucial for everything we do as retina specialists, whether it is in the clinical practice setting or clinical research and drug development setting. That is where AI comes into play. Per current practices, the clinical science behind retinal research has been based on human expert graders. They are highly trained, but it is a time-consuming endeavor, and it has limitations.
Retina has been one of the driving specialties for advances in ophthalmology since we started moving toward digital data in the past couple of decades. It is actually one of the few fields of medicine that clinical decisions are based on digital imaging biomarkers.
In retina, for example, AI can fundamentally change the way we screen patients with specific conditions. We can find patients who need help the most as we become more efficient in early disease identification.
AI will also transform how we characterize retinal disease states and disease types. It is going to help us triage our patients. Who are the patients who need help by a health care provider, an ophthalmologist or a retina specialist, how fast and when? It will also allow us to identify which patients do not need a specific intervention and reduce unnecessary burdens on them and the health care system overall.
Thus, in the future, I believe that counseling and management of patients with retinal diseases may significantly benefit from AI-powered tools. These could help us identify new retinal imaging biomarkers and make predictions about a patient’s future outcome. We can use these AI-based predictive models to show us what a patient can expect based on disease natural history, or we could train models to show us how a patient will progress on a certain treatment. Either of these algorithms will be truly transformative for patients and clinical experts, whether they are in the clinical practice or clinical research settings.
Finally, I think AI will help close the gap between patients’ outcomes in clinical trials and clinical practice. We will see some of these algorithms currently working well in research start to become available in clinical practice, and hopefully that will close the gap between the good results we see in controlled settings vs. the sometimes suboptimal results in real-world settings.
Daniela Ferrara, MD, PhD, FASRS, is an assistant professor of ophthalmology at Tufts University.
Collapse
Read more about
Click Here to Manage Email Alerts