Hattie Hayes: Hello, my name is Hattie Hayes and I’m the editor of Ophthalmology Times Europe. Today we’re speaking with Professor Ursula Schmidt-Erfurth about her keynote lecture at the Congress on Controversies in Ophthalmology. The title of that lecture is “How good are we really without artificial intelligence tools?” Professor, thank you so much for joining me today.
Ursula Schmidt-Erfurth, MD: Thank you for the invitation, and thank you for your interest.
HH: So first and foremost, what are the biggest challenges connected to artificial intelligence right now, and what are the greatest opportunities that AI provides?
USE: So the greatest challenge is that it is very innovative. It has just developed to be used in the clinic over the last few months, actually, because AI is not just an algorithm that you come up with, that you design in your lab. AI, for the clinical use, has to go through a rigorous validation and regulatory process, which is very demanding. There is so far in Europe, only one MDR-approved device or AI tool, which is the fluid monitor and the GA monitor, that means for using AI in diagnosing, monitoring and screening for wet AMD and dry AMD. These are also the opportunities because we know that wet AMD is already treated. It is the most frequent therapeutic intervention in all medicine. We all treat millions of people. However, we know that the output is not so convincing, not as high. The benefit is not as high in the clinical routine as it is in the clinical studies. The clinical studies have reading centers and other sites and specific protocols that help to make the right decision. In the clinic, there’s nothing like this: doctors are overworked, have to go through all of these 50 shade-of-grey videos back and forth. And this is why we have to empower doctors to treat wet AMD by mouseclick, to reduce workflow and improve the quality by introducing standards, and this is what a fluid monitor does. In geographic atrophy, the situation is even more difficult, or more promising, because GA biomarkers are not seen on an OCT. So easily, the changes are subclinical: we talk about photoreceptor alteration, which is reflected in the EZ [Ellipsoid Zone] layer. So we are dependent on AI to extract this biomarker from the OCT. The FDA has just decided that EZ is the most important primary outcome biomarker for treating GA. So in the clinic, we can use this now by using a tool like the GA monitor completely certified, validated, and it gives you insight into the area of photoreceptor loss and the area of RPE loss at one glance, which is otherwise not possible.
HH: Do you think that ophthalmology as an industry is uniquely suited to benefit from all that AI is capable of? Why or why not?
USE: I think that ophthalmology is one of the most promising targets for many reasons, because doctors are already using digital imaging in large amounts and in huge volumes per image. So millions and millions of pixels, in millions of millions of patients’ retinal images. And this is the perfect playground for precision AI. Also we have a lot of patients and retina is unforgiving. So we are fighting to maintain to preserve visual acuity. And this is why we have to identify the right biomarkers that should trigger treatment and we have to identify them in the most precise, that means quantified, way. OCT is the most frequent diagnostic examination in all medicine. OCT offers this evidence, and AI can extract it for us doctors so that we can use it to the benefit of our patients.
HH: I know that the Congress on Controversies in Ophthalmology has a unique setup, with this debate structure. Why are dialogues like those explored at this meeting so vital to eye care?
USE: I think communication is always key. Debate is always key, and you want to collect all of the evidence. So, the purpose of such debates is information flow, because information volume increases in modern medicine on an hourly basis, almost within seconds. So, information is the base for decision making and for having an opinion. The other point is exchanging experience, because everybody is working a lot, but from many different perspectives, and with different focuses. Exchange of expertise between human experts [is vital]; expertise makes them experts, and shared expertise would increase the level of expertise. So this is important. And of course, there are always pros and cons that help to shape the perfect indication. There is no such thing as, “If you have a hammer, everything looks like a nail.” You have to have an idea of what the hammer and the nails should be used for. And it is with this identification of the best indication that we will obtain the best results.
HH: Wonderful, thank you so much for your time and thank you so much for sharing your expertise.