COPHy 2024: Is artificial intelligence ready to replace physicians?

This spring, ophthalmologists from around the globe will gather at the 15th annual Congress on Controversies in Ophthalmology (COPHy). Leaders from the field will debate clinical best practices and industry trends in front of an audience of their peers. Two retina specialists from Italy spoke with us about their debate topic and defended their positions: arguing in the affirmative, Giuseppe Querques, MD, PhD, associate professor, University Vita-Salute, IRCCS Ospedale San Raffaele, Milan; and arguing in the negative, Paolo Lanzetta, chairman of the Department of Ophthalmology, University of Udine, and Director, Istituto Europeo di Microchirurgia Oculare, Udine.

Editor’s note: The below transcript has been lightly edited for clarity.

Hattie Hayes: Hi, my name is Hattie Hayes and I'm the editor of Ophthalmology Times Europe. Today we're speaking about the Congress on Controversies in Ophthalmology, which takes place this year in Athens, Greece. Joining me today I have Professor Guiseppe Querques and Professor Paolo Lanzetta. The professors will be debating the topic "Is artificial intelligence ready to replace physicians?"

Why don't you each go ahead and tell me about the position that you are arguing in this debate?

Giuseppe Querques, MD, PhD: I will be supporting the possibility that artificial intelligence will replace physicians and, and particularly ophthalmologists, in the near future. And I must admit that, since I'm supporting this hypothesis, this is a kind of easy win to me. This is mainly because, as you know, artificial intelligence is not the future, but it's the present. We are already using a lot of artificial intelligence in our clinical practice. Many of my colleagues in different countries, including my country, in Italy, are already using the clinical support from artificial intelligence when they are managing patients with different eye diseases, and particularly retinal diseases.

I think that this is particularly useful in areas in which there's a lack of the right number of physicians for managing patients. This is not exactly the case for my country. But as I said before, even in my country, we are using a lot of support from artificial intelligence. Probably at this time, I cannot say that artificial intelligence is replacing physicians, at least in my country and many other countries, but there are areas all over the world that are largely getting benefits from artificial intelligence. And this is a big, big benefit, not only for patients, but also for physicians.

Paolo Lanzetta, MD: First of all, let me start by saying that artificial intelligence in medicine is here to stay, to paraphrase the famous song by Neil Young – "Hey, hey, my, my," as at least those of my age may remember. So the topic I've been assigned in the debate at COPHy is related to artificial intelligence. And the question is whether artificial intelligence will replace physicians, not only in ophthalmology, but in medicine in general. So before answering these questions some days ago, while preparing my talk, some days ago, I went [on the] internet, namely, in PubMed. And the number of publications related to artificial intelligence during the last few years has really grown exponentially. The number I found was about 23,500 publications on artificial intelligence in medicine. We are now very familiar with this term artificial intelligence, which includes a number of different terminologies.

But this term is really dated, because it was coined in 1955 by a Stanford professor, whose name was John McCarthy. He defined artificial intelligence as "the science and engineering of making intelligent machines." So we are dealing with some entity which has some intelligence. Is this like humans or not? We don't know yet. We have seen a lot of development. But what I can say is that after many years, many advancements in the field of artificial intelligence, these technologies are still created, activated and controlled by humans. So if we move specifically to the retina field, we are also witness to an explosion of artificial intelligence-based systems, and those systems are dedicated to screening, diagnosis and patient monitoring. Most systems work automatically, as we would expect, and as we would like, but they're still guided by physicians who have a pivotal role in the management of patients.

Obviously, many of us may be concerned about artificial intelligence taking over jobs in medicine, mostly because the artificial intelligence algorithms are showing themselves to be better than humans. And this is the experience in different fields, such as radiology for example. In my opinion, however, this is unlikely to happen in the next future. First of all, machines will never substitute the physician-patient relationship. This is very important. In other words: machines will not replace empathy. And we know how much empathy is important in our relationship with the patients. Think about patients who develop some macular conditions, who may need some repeated injections with anti-VEGF. Empathy is really part of the relationship with the patients. Just to conclude, I believe that artificial intelligence and related technology will definitely assist physicians in being more accurate, more efficient, by reducing errors and facilitating their daily activities. But artificial intelligence per se will not fully replace physicians. It will be more like collaboration in between different professionals for better managing and treating our patients.

HH: I'd love to know what you think the biggest challenges are right now that physicians who use AI are facing when using that technology, and where are the areas where it presents the biggest opportunities?

PL: Obviously, there are dangers, controversial aspects and different implications of AI use in ophthalmology and other medical fields. In a recent publication from our group in the journal Graefe's [Graefe's Archive for Clinical and Experimental Ophthalmology], we did identify six main issues and we try to analyse and discuss those issues. These include biases and clinical safety, cybersecurity, health data and AI algorithm ownership, the "black box problem," very important, medical liability and the risk of widening inequality in healthcare, especially during these days. We see how unequal access to health care may be in different areas or different countries.

So, what are the most important [questions]? To me, one of the most important questions is related to the black box problem. We are becoming more familiar with time, but many of us are not really–do not have a specific knowledge of these black box issues. So, this concept refers to the opacity of the AI systems, in which I mean that it is difficult to understand, predict or systematically influence the way in which an input is transformed to an output, which is what we are taking out of our AI system. Therefore, we may be less confident in using machines whose workings we do not fully understand. And therefore, there is a proposed alternative, which is opposite to "black box" is called "white box" AI, which is more transparent in its decision making process. However, we have to say that these models are usually less efficient in terms of predictive capacity, which is something very solid we are looking for, if we want to be assisted by AI in managing, treating and diagnosing our patients.

So there are other models which mimic the human clinical reasoning and they are more likely to be adopted in clinical practice because they very much resemble the way we use our knowledge and our reasoning and our brain. They are the so-called "two stage machine learning systems," that first learn to identify relevant retinal lesions such as, in the retina field, maybe exudates, vascular lesions, hemorrhages. And then they base the subsequent predictions on these type of detections. And we have some experience on these, together with the department of mathematics, computer science and physics in our university in Udine, and we have recently developed a so-called explainable artificial intelligence model. In this case, for the detection of geographic atrophy, by using color retinal photographs, a very simple method. So to me, opportunities in the use of artificial intelligence are quite evident and obvious. And ultimately, these technologies will help physicians to be more efficient, less prone to error and mistakes, and probably more expert on our patients' needs.

GQ: While the biggest challenge, I think that maybe a couple of challenges are related to artificial intelligence, one is somehow cost and lack of standardisation. Provided that we are nowadays able to use or to apply the different systems to different devices that are needed for applying artificial intelligence in our clinical practice. But still, there's some concern about this, or about costs and about standardisation. The other important point is related to legal issues. And this is for sure, something our administrators need to work a lot. Because as I said, before, we are ready, patients are ready. But we need proper regulation for the proper use of artificial intelligence in clinical practice.

HH: I know that the [Congress on Controversies in Ophthalmology] is unique in that it has this debate setup. Why do you think that debates or dialogues like the ones we're seeing at this meeting are so important in eye care?

GQ: This is extremely important, because we will never get an agreement even when we are doing the same topics as ophthalmologists. And this is a, you know, you can imagine, a hot topic that also is considered as a risk for physicians. So how important is debating artificial intelligence, or debating in-person? Debate is important, because it helps to clarify issues that otherwise cannot be clarified. I think that this formula is the perfect one, for speakers and for attendees.

PL: Well, let me say that the COPHy meeting formula is one of the best ones you can meet by attending professional meetings. I've been in these meetings for many years now. And every time, I do come back home with a better knowledge in my field, in ophthalmology and specifically in the retina field. This Congress on Controversies in Ophthalmology is, first of all, an international event. It has brought, in its 15 years of age, about 13,000 participants with a huge number of faculty members, I think about 900 faculty members. So there's a lot of ideas and knowledge in this meeting. At the end of the full presentation, you really gain some knowledge, you develop some new ideas. So it's very, very helpful. So I think this meeting, the COPHy meeting, is really an event that you cannot miss, if you want to have the latest update on hot topics in ophthalmology.