Dr. Sandor Beniczky: Hello. My name is Sandor Beniczky. I'm an MD-PhD. I'm working in Denmark. I'm a professor at Aarhus University, and I'm the head of the diagnostic department at the Danish Epilepsy Center in Dianalund.

Dr. Sandor Beniczky: Right. So, I trained as a neurologist, and I did my PhD and my medical school in Hungary at the University of Szeged. And then I completed the clinical neurophysiology fellowship, which is two and a half years full-time training in EEG and EMG and polysomnography, in Denmark. And then I also did the European certification in epileptology in 2008 and 2009, which was organized by the European Academy of Epilepsy.

Dr. Sandor Beniczky: Right. So, in the United States, it is a three-year residency program, but the prerequisite is one year of internal medicine. So altogether, it's four years, but one-year internal medicine plus three years of neurology. In Europe it is between four and five years of neurology training. And then, of course, there are numerous differences in the learning objectives and the specific curricula. But I think the most obvious is that it is a duration difference.

Dr. Sandor Beniczky: Clinical neurophysiology training is very, very different. So, in some parts of the world, it is completely absent, and in others it is part of neurology. So, during the residency they can learn EEG, EMG, and then perhaps they get the certification in that. Now in other countries, at the opposite end of the spectrum you have those countries where clinical neurophysiology is a specialty, not a subspecialty, but you can opt for that right after medical school. That's the case in some western European countries like the UK, Ireland, some of the Scandinavian countries, Portugal. And when it’s specialty training, that's usually four or five years.

Now six months of neurology is compulsory within the clinical neurophysiology when that's a main specialty. However, in most of the countries, this is a subspecialty. And then it's either a kind of a master’s in clinical neurophysiology, or formalized subspecialty training, as in Denmark.

Dr. Sandor Beniczky: Yeah, so this paper you are alluding to was authored by Jonathan Cole and Anita Kamondi and it was published in Clinical Neurophysiology.

Basically, they start by mapping what's out there, and as I've just told you it's a huge heterogeneity. So, they start by giving an overview about the state of the art, what's going on in the different countries, and then they come with recommendations.

And I must say, I really like these recommendations because they are very flexible. I mean, it's unrealistic to imagine that based on this guideline, all the countries will harmonize and do the same thing, but they allow for this heterogeneity. So, they issue recommendations for those countries where this is an independent specialty, and then they issue recommendations for those countries where this is a subspecialty. So, they are flexible enough to adapt to the local peculiarities. Nevertheless, within each category, they specify the minimum criteria and the guidelines, and they focus very much on the governance, how this should be done.

Then the competencies that must be mastered during the training, then they specify also the number of investigations, the numbers of tests that must be done, how many EMGs you have to do and supervise, how many EEGs you need to read and supervise and so on. They also have this very, again, flexible, modularly approach. And then they specify that there are some primary competencies. And, of course, these are the EEG, EMG, nerve conduction and evoked potentials. And then as complementary competencies, they list all the rest, like intracranial EEG and single fiber EMG and so on.

Right. So, first of all, let me say that I'm aware that many, many papers proved that it's only a moderate agreement when neurophysiologists look at EEGs, but that's not totally true because most of those papers addressed an artificial scenario. Experts were given 10 seconds of EEG with one discharge. And the question was, is it an epileptiform discharge or not? But this is an unrealistic scenario. What's a realistic scenario? You give the expert an EEG and then you want to know: Does this EEG contain epileptiform discharges, for example, or diffuse slowing. So, you look at the whole EEG.

So, if you just ask the experts about one discharge, indeed, the interrater agreement is moderate. But if you give the experts the whole EEG, then the interrater agreement is substantial. And for some items, like for example, the generalized epileptiform discharges, it's even almost perfect.

We've recently published a paper in JAMA Neurology where this was one of our findings. Well, the main goal in that paper was to validate an artificial intelligence model, but as a side product of that paper, we had the interrater agreement of experts when reviewing the whole EEG, and then the interrater agreement is much better.

So first I wanted to do a kind of myth busting with saying that we are not that bad. It's an artificial question when you ask just about 10 seconds.

Now, having said that, of course, there is room for improvement, and then we must increase the inter-rater agreement. How can we do this? Of course, we must speak a common language, so we need to come with a clear list of common data elements, a clear definition of what is what, and then, of course, it would help if we would use a computer software to do it, to label all these abnormalities that we spot on the EEG so that we avoid the variability which may be derived from using free text. So, if you just click on the items which you identify, then of course you use the same terminology, the same nomenclature.

SCORE was developed exactly for this purpose: To provide a common terminology and a common way of describing it. And then in the long run this is an excellent input also for developing AI models.

Dr. Sandor Beniczky: Well, I think that the future implementation will be much aided if algorithms and AI models will use this system. But the biggest obstacle is still that often SCORE runs as a standalone or an independent software package from the EEG reader. And often we encounter difficulties in implementing it. For example, the IT departments of a hospital do not want to allow an external software getting access to the EEGs and so on. So, I think using AI models with SCORE and integrating SCORE into the EEG reader would really help in spreading this out.

Dr. Sandor Beniczky: Right. Well, first of all, I have to emphasize that although the goal is the same, I mean, the size and the coverage of these initiatives are completely different. So, the ILAE Academy is really aiming at comprehensive education in epileptology and obviously EEG is part of it.

Now within the ILAE Academy, we have different tools to teach EEG. Perhaps the most important is the VIREPA, which is an online tutored six-month course. We have the basic course and then we have another course for neonatal and pediatric and then we have a third EEG course for the advanced EEG. Besides this, we have also self-paced interactive tools, and then very importantly the roadmap is consisting of a series of short lectures, and we are developing this further, but these address specific items in EEG reading and these are also available on the Epileptic Disorders homepage, and we are working in integrating it also into the ILAE Academy.

So again, these two points toward the same direction, and the roadmap is just a small part in the big ILAE Academy.

Dr. Sandor Beniczky: So, first of all the hardware. Together with Denmark's technical university, we've developed a very inexpensive EEG machine. So, it's using the computation power, for example, of tablets or smartphones. So, the goal was to give the hardware, the machines, to resource limited areas, to the low- and middle-income countries.

But of course, the hardware is not enough. I mean, you need to interpret it. And then there are several ways of solving this. One is to educate people. So that's why many of the ILAE's educational programs are online. So, education is one component. And then the other component is facilitating telemedicine. If you do not have experts locally, then perhaps you can get help via telemedicine from experts sitting in another place. And then the third component would be, and I really believe that will be the game changer, is developing artificial intelligence models for solving this.

Dr. Sandor Beniczky: Well, to do research, you really need to know epileptology in advance. And I think that the young trainees have a fantastic tool to learn epileptology and it's available online and it's the ILAE Academy. So, no excuse not knowing about epilepsy because now you have the proper tools to learn it and it's available anywhere you are in the world. If you have access to the internet, you have access to the ILAE Academy too.

And then the research, well, I always tell my students that if we want to do research, then you have to be curious about a problem and then come up with ideas how to solve it.

Dr. Sandor Beniczky: I can compare the current situation with what I've been in 20 years or 25 years ago, and I can see already a huge improvement. There are lots of available educational resources on the Internet. Obviously, there is still room for improvement, and the self-paced learning modules will again be a game changer.

And then again, artificial intelligence will also play a major role in education, so it may provide an adaptive tool for learning. So, I think we have made a huge progress, and I can expect that this will improve considerably in the near future.