Ophthalmology at Crossroads for Information Technology, AI

The field of ophthalmology had already been moving toward telehealth and artificial intelligence (AI) before the COVID-19 pandemic, but these changes are being accelerated now, making it crucial for ophthalmologists to learn to adapt.

Ophthalmology, like many fields of medicine, is facing accelerated changes due to the COVID-19 pandemic. Prior to the pandemic, telehealth was only used for retinal screening, but it has blossomed to all aspects of ophthalmic care, explained Aaron Lee, MD, MSCI, associate professor, University of Washington, during a session on information technology at the American Academy of Ophthalmology 2021 annual meeting.

Artificial intelligence (AI) has taken the ophthalmology field by storm, he said. There are diagnostic models where a human being isn’t even in the loop, which is only happening in ophthalmology—no other field is so advanced with AI, according to Lee.

With this massive change, ophthalmologists have to decide how they will react to these changes: deny and ignore, resist the change, or shape the change.

When a new technology is introduced, people experience the Gartner Hype Cycle. After the new technology first appears, there is a “peak of inflated expectations”—the idea that this technology will do everything and replace people. This is followed by the “trough of disillusionment”—the realization that the technology is not at that point to meet all the expectations of it. From there is the ascending “slope of enlightenment”—understanding the uses for this technology based on early adopters. Finally, there is the “plateau of productivity”—the beginning of mainstream adoption.

AI has been going through this cycle. The innovation trigger happened around 2015 when the first deep learning models were developed, Lee explained. Currently, we’re on the descending limb of inflated expectations. He expects that by 2026, we will be on the slope of enlightenment.

There are 2 possible futures that he sees for AI in ophthalmology. The first is what he called the “Skynet” future in which he imagines a conveyor belt of patients, getting imaging with a robot and then an AI model decides what they need: If everything is OK, they can go home, or they are directed to a room for a procedure. If they are directed to a room, a robot combined with AI conducts the procedure or surgery.

In this scenario, there is no ophthalmologist.

In the second scenario, the future is “made of the same stuff as the present.” One reason for this is that there are significant limitations with AI models.

  • There is a real problem with the fragility of AI models, Lee said. AI “can make a completely catastrophic mistake…and there’s no way to figure out why.”
  • It can be hard to get AI to do what we want it to do. For humans, we can see an example of a tree and then 6 different types of trees and understand they are all trees. AI models can’t do that yet.
  • It’s impossible to truly know how and why AI models work. So far, we haven’t been able to understand why or how a model can do some miraculous task.

As a result of these, one future could be that there won’t be a lot of change because “we can’t overcome some of these things” and the practice of ophthalmology may not transform very much, Lee said.

In his ideal future, AI would be used as a clinical tool to increase efficiency, improve outcomes, and save sight. It would mean ophthalmologists spend less time typing in the electronic health record and more time talking to the patient. AI would allow clinicians to integrate information and overrule the AI model when necessary.

Lee reminded people what happened with GPS units. There were people so blindly following the directions, they were actually driving into a lake or off a bridge.

“When we work with these tools, we still need to be the person in charge to integrate the information,” he said. Making clinical decisions is still an art form that incorporates knowledge of the patient and their life.

Lee was followed by April Maa, MD, associate professor, Emory University School of Medicine, who discussed how telemedicine will remain relevant after the pandemic.

“Telemedicine is part of the good that came out of the COVID disruption,” Maa said. Telemedicine is no longer the “red-headed stepchild” of medicine. Now, we can use telemedicine platforms to iterate and improve because so many doctors have been using them during the pandemic. “Now, we can take the lessons learned and apply them to the future.”

However, it’s important to note that telemedicine is not only a video chat. For instance, it isn’t actually possible to tell if a patient’s glaucoma has gotten worse via a video chat. Instead, telemedicine will be relevant for the future because of the possibility of hybrid models. Maa described a scenario in which a patient comes into the office for testing or has a test done at home, then the clinician reviews the results and discusses them via video with the patient.

Telemedicine also isn’t going away because patients love it. There is high patient satisfaction with telemedicine, and they find it more convenient. Maa noted that this may lead to better compliance and vision for their patients. In addition, patients are now expecting a telemedicine option after the pandemic, and not offering it may be a detriment to a practice.

“I think telemedicine in the future of our eye care delivery is going to be monetarily and lifestyle beneficial for us [doctors],” Maa said. “It’s going to increase patient satisfaction and be a practice builder. And most importantly, it’s going to promote health care equity and keep everyone seeing well for as long as they can.”