The Human Intracerebral EEG Platform and the power of big data: Dr. Philippe Ryvlin

Reported by Dr. Maryam Nabavi-Nouri | Edited and produced by Nancy Volkers

---

Cite this article: Nouri M. The Human Intracerebral EEG Platform and the power of big data: Dr. Philippe Ryvlin. Epigraph 2024; 26(2):62-68.

---

Listen below or download the episode.

Find Sharp Waves episodes on Spotify, Apple Podcasts, Amazon Music, iHeart Radio, and Stitcher.

The Human Intracerebral EEG Platform is a cloud-based, collaborative environment that encourages centers to share data and conduct research with state-of-the-art methodologies and software. Dr. Maryam Nouri interviews Dr. Philippe Ryvlin about the potential of the platform to advance the understanding of human brain function.

Sharp Waves episodes are meant for informational purposes only, and not as clinical or medical advice.

Podcast Transcript

[00:00:00] Dr. Maryam Nouri: Thank you for joining one of our episodes of Sharp Waves. It’s delightful to have Professor Philippe Ryvlin join us from Lausanne.

 I was just going to start by asking you to introduce yourself. You’re very well known in the epilepsy field, but just for our audience and tell us a little bit about your expertise

[00:00:17] Dr. Philippe Ryvlin: Well, to summarize my career, I'm primarily a physician and neurologist, but very early on in my career, I started being interested in epilepsy, and that has been my main field of work of clinical work and research for the last 35 years.

I've been working primarily in France until 2015, where I decided to move to Lausanne, and the primary reason was that I wanted to be very close to an area that is extremely well known as a strong reputation in technology and neurotechnology, and this is the polytechnic school of Lausanne, and the polytechnic school of Lausanne worked very closely with the University Hospital of Lausanne. And actually those are the two institutions which originally created the Human Brain Project (HBP) a little bit more than 10 years ago.

[00:01:12] Dr. Maryam Nouri: Oh, wonderful. I think that's a great segue to just letting us know a little bit—can you walk us through the briefly about the Human Brain Project and how it evolved to include patients with epilepsy and different epilepsy projects?

[00:01:28] Dr. Philippe Ryvlin: HBP was one of the two flagships that Europe has decided to fund in 2013; it was supposed to be two $1 billion projects. In reality, it has been a little less, and it proved to be very challenging project and a reason why this framework has not been replicated by the European Commission.

Originally, it was aiming at creating a digital brain, to put it simply. I think eventually that's not what it produced, for a number of reasons. And there had been political turmoil early on. When I arrived in Lausanne and took over some roles in the clinical research in HBP, which is a very minor portion of the project, things were already completely reformatted into a project that I would say primarily focused on digital neuroscience. Let's put it that way. That is, developing tools and research based on those tools that are primarily digital and that include, for instance, the so-called virtual brain that has been developed by Viktor Jirsa from Marseille, which is a complex connectomic model. There's many others being developed and used in epilepsy at the moment, but this one is clearly one of the most sophisticated.

But it also includes developing a platform for helping the advances in the field of neurorobotics. There was also neuromorphic computing, which aims at creating chips for computers, the organization of which will resemble an assembly of neurons and a classical chip, you know, all those kinds of things.

But myself, I was involved in more trivial activities, which are really about data sharing. And it's interesting because eventually that proved, I think, more and more relevant to what is happening in the field of medicine, in the field of neurology. If we think of the role of AI [artificial intelligence], if we think of the role of large language models, definitely big data is an essential part of it.

And big data starts by organizing and facilitating data sharing. So that was our main objective in the Human Brain Project.

[00:04:04] Dr. Maryam Nouri: Right. Can you now tell us about how in addition to the virtual epilepsy brain, how the HIP [Human Intracerebral EEG Platform] could complement or add to the Virtual Epilepsy Project within the eBrains platform?

[00:04:29] Dr. Philippe Ryvlin: It’s truly complementary because the HIP, the Human Intracerebral EEG Platform, is really a platform for data sharing upon which you can use a number of tools like the virtual brain, which can be run on what we call sEEG data on the HIP. So it's different, but complementary. Importantly, the HIP was developed very late in HBP. So, when I arrived at the HBP in 2017 I took over the activities that were launched by others. But then when we got to the last stage, the last three years, I proposed to build the HIP.

The whole idea is first of all, to acknowledge that intracerebral EEG is a unique set of data on the human brain function, really a unique window on how the human brain is operating because we record directly from a neural assembly in a human being, and there's almost no other such situation.

Of course, we do it not for research. We do it for a clinical purpose, but this offers a unique opportunity, and that has been acknowledged for decades. There's nothing new about it. However, what happened is that most centers will perform sEEG in a limited number of patients every year. I will say an average of between 10 and 15 per center. There's one center in China that can perform up to 100 per year. And there's a few centers in Europe that perform 30, 35 per year. But the average is between 10 and 15. Very importantly, the regions of the brain that are being sampled are different between patients, dramatically different, and in each patient you sample actually 0.5 percent of the total cortical gray matter.

So you immediately understand that if you want to make progress in human brain function using sEEG you will need to collect data from many, many patients, and that cannot be organized at a center level. I will say even at a national level if you are ambitious.

And that's the premise of setting up the Human Intracerebral EEG Platform that is offering a solution that will encourage centers around the world to share more data. We have 150 sEEG centers in the world, approximately. That means possibly between 1,500 to 2, 000 patients undergoing sEEG every year. If we could get together half of this information, we will definitely make huge progress.

But it's not just a question of having the technology. The question of finding a way to motivate people and addressing all the issues that I will say are more and more complex when it comes to sensitive health data sharing and export.

So that's a big part of the concept of the HIP. It's not just having a platform, but it's making that platform a success.

[00:08:00] Dr. Maryam Nouri: How does it differ from other platforms? Because there are some that already exist, again, not at this magnitude, but I understand your vision for this platform is different. And what you are offering as part of it is different, including protocols, including software expertise.

[00:08:18] Dr. Philippe Ryvlin: It's different because most of what exists can be described as repositories, and repositories of public data are fine. It's a way to promote open science and advance in medical research, definitely. But what we offer is different indeed and we try to be more ambitious.

So first of all, the concept of the platform is a so-called cloud based trusted research environment. What does it mean? It means that if you participate in the project, you can upload your data on the platform and then you can work on your data and the data of others if they decide to share it with you. You have access to the state-of-the-art methodologies, software, pipelines to take the best out of your data.

So that's a very different concept than a repository where people will go and download data of interest. And that's a way to motivate people to participate. You know, you tell them, well, you don't actually need to get on your computer all the state-of-the-art tools to understand how to make them work, to ensure the appropriate upgrade of all the software, because we do that on the central platform. It will always be state of the art on the platform. So that will facilitate your work. That's number one.

Number two, of course, it's promoting data sharing. Promoting data sharing, but still protecting the value of each center’s data and we do that by having both private space and collaborative space. Private space means you're from a hospital in Detroit, and you want to upload your data, but you don't want to share them at the moment. Thank you very much. Well, you upload them in your private space, you can do whatever you want on your data, curation, post processing, analysis, with all those state-of-the-art tools that I mentioned, but you're the only one having access to the data.

Then you want to collaborate with other centers, but first of all, the platform offers communication tools, so that projects can be shared in the community. And then, depending on the type of project, you might invite 5, 10, 20, 50, centers to join the project. Then you have a collaborative space where those partners and only those partners have access to the data. Where they can share whatever they want to share. You do not necessarily have to share all the data you upload in your private space. Maybe only part of it. Whatever. And we believe that really this is the essence of the platform, promoting new collaborative projects that can be run on the platform.

There are a few other tools that are worthwhile mentioning, some clinical EEG software used for sEEG recording. And some of those are very widely used in Europe in particular, but not only in Europe. What we have organized is that you can upload your data on the platform from your clinical software using the copy-paste function that we're using every day when we want to archive our recording after curation. That means that makes uploading extremely easy.

[00:12:03] Dr. Maryam Nouri: The data that's being collected is mostly neurophysiology, correct?

[00:12:09] Dr. Philippe Ryvlin: Yes, it’s intracerebral EEG recordings, but for interpreting those recordings you also need to know exactly where the electrodes are placed within the patient brain. So you also need the associate neuroimaging. And part of the state-of-the-art tools that are implemented on the platform include pipelines to make very fine localization of your electrodes on the patient individual brain. So we do also have neuroimaging data.

 And the data format we use for sEEG data is so-called BIDS (brain imaging data structure), BIDS sEEG, and that format can also include some clinical information.

[00:12:50] Dr. Maryam Nouri: I mean, there are enormous benefits, like you have been alluding to, that can be gained from optimizing usage of big data platforms such as HIP, such as implementing of artificial intelligence and machine learning, precision medicine, and generalizability of the data.

So with that I wanted to know, what are the challenges both from a data acquisition perspective and end user perspective that you have come across?

[00:13:18] Dr. Philippe Ryvlin: It's very clear that the main challenge today in health data sharing is contracts between institutions. I don't think it is about the concept of sharing data for the physician who collected it, because I think we all understand we can only gain by sharing data if you can protect your own, I would say intellectual property, which is the offer with the HIP.

But the main problem is all the legal errors at our institution, at least in Europe. I don't know in the US, but in Europe, what has happened is that lawyers have taken more and more responsibility in hospital and with often limited knowledge of what exactly we're doing and also with limited resources. To give you an example—it's extremely frequent that one of the contracts I need for sharing data with another hospital will sit six months on the desk of our legal department. So you can imagine when, in our case, we're dealing with contracts with tens of different institutions. And the problem is the same in the other institutions.

So that's the main challenge. That's the main challenge. It's purely legal.

[00:14:40] Dr. Maryam Nouri: And I can attest to that working in the North American center that it becomes an issue of privacy, which is probably sitting at the heart of this, but these are anonymized data like you had mentioned.

[00:14:54] Dr. Philippe Ryvlin:  On the medical informatics platform, we only use anonymized data. On the human intracerebral EEG platform, it's supposed to be primarily synonymized data, but the lookup table that allows you to know who is who, of course, stays in the hospital. It's not put on the platform. So on the platform, you could argue that the data are anonymized, but that also means that we only accept on the HIP data that has been collected from patients who have to provide their informed consent.

[00:15:28] Dr. Maryam Nouri: So, this really bears importance in allowing the scientific community to understand epilepsy and its complex interplay, really what you're saying above and beyond your single patient that's in front of you.

So how far do you think we are from integrating this neurotechnology into the clinical workflow and whether there needs to be studies like the clinical trial that's being done in Marseille, whether those need to be implemented before they can be implemented into the workforce.

[00:16:01] Dr. Philippe Ryvlin: I think here it's very important to distinguish the two main types of research that can be done on the HIP. One type is epilepsy directed and aims at identifying new biomarkers. That can have immediate impact on the management of patient. I'll give you a few examples. The other is human brain function at large, which could in the future also have an impact on how we care for our patients, but it's really start by simply better understanding the brain through performing different type of tests while the patient is being implanted and recording how his brain responds, how the different regions of his brain respond to a language task, a memory task, a social cognition task, and so forth.

So that's really two different fields, both of which can massively benefit from big data and currently suffer from not having access to big data. So coming back to your question, I will concentrate on the first part, the epilepsy research. High-frequency oscillation (HFO). It's been about 15 years that our community is hoping that HFO will prove to be a more reliable biomarker of the epileptogenic zone to guide epilepsy surgery and of better surgical outcome than using the current tests. And many uncontrolled retrospective data suggest that it could be the case yet. There has been one randomized controlled trial performed in the Netherlands, and that was published last year in the Lancet Neurology, and that was negative. That trial faced a number of challenges, so its conclusion is not definitive, but it really showed that if we want to move forward in a field where there is a lot of need to still improve epilepsy surgery, we need to amass more data.

Another example, we discussed rapidly the virtual brain. So, the virtual brain can be run on HIP. At the moment, there is a clinical trial that does not use HIP, that is being run in France. Actually, it's over. They have been recruiting 400 patients, randomized into using or not using the virtual brain prior to epilepsy surgery decision.

And the goal is to demonstrate that if you perform that complex connectivity model you get some information that helps you better delineate the epileptogenic zone and offer more chance of postoperative seizure freedom. We don't know whether that will be the case. I'm hoping it will be, but we don't know yet.

But imagine that this is the case, or even if it's not, I think it will be interesting to continue in parallel being able to address this question. Or related questions, like is this model better than other connectomes that are currently available on data that will be shared on the HIP? So, you know, those are two, I would say, simple examples of existing biomarkers whereby we still really miss a clear-cut answer and guideline as to whether we can improve epilepsy surgery using those biomarkers.

[00:19:26] Dr. Maryam Nouri: And how much technical and scientific knowledge is required of end users to be able to engage with this platform?

[00:19:36] Dr. Philippe Ryvlin: Well, I think the platform is made to enable all who work in the field to perform some activity with the expertise, the knowledge and the tools they already master. Most of the data analytical software available on the platform are well known by many of the researchers in the field, whether we're talking about Brainstorm or about for the imaging 3D slicer, FreeSurfer.

So we have a number of other tools, but just to cite some of them and illustrate that people will engage in the project and see that they can use the same tool that they have on their computer. Maybe they will be keen at discovering other tools that they don't have and that they believe will be useful for them. And here, I assume they will benefit from some tutorial, but you don't have to be a very sophisticated data scientist to get your head on the platform.

[00:20:39] Dr. Maryam Nouri: It's essentially one part is uploading the data and the other part is understanding the results. And from what I understand, and I haven't tried this myself, but this is meant to benefit everyone with different data science backgrounds.

[00:20:56] Dr. Philippe Ryvlin: Yes, absolutely. And hopefully it will also bring the field to an organization of research that will more resemble classic clinical research like clinical trials or big cohort studies, where the essence of collaboration is that many, many stakeholders contribute to data sharing and eventually one biostatistician performed the analysis. But this person that performed the analysis will not necessarily have a leading role in the project. Because there's also a great value for all those who contribute to data as well as those who propose the ideas. And so it could be a way maybe to move apart from a situation that we often encounter in sEEG, where if you're not a sophisticated data scientist, you feel you cannot do anything and you give your data away.

[00:21:58] Dr. Maryam Nouri: I've been five years out of my training and I feel like this is completely a new world. A world of neurotechnology has taken the epilepsy field by storm.

And this has been an interest of mine. So I try to always seek it out, but I wonder how can an average epileptologist, average meaning a clinical epileptologist, become more proficient in this technology? And I would say a new language that's sort of a new addition to our clinical practice now. I wonder what your thoughts are on that.

[00:22:35] Dr. Philippe Ryvlin: No, I agree. This is a challenge. But on the other hand, a majority of sEEG centers are interested in performing sEEG-based research. We have done some surveys. And so, you know, when people get interested in the sEEG, they can hardly escape getting involved in research. So I think there is an incentive in most centers performing sEEG and again, to participate in a project on the HIP does not require that you master a lot of new technology, but what it requires is that you understand the principle of how it works and give a little of your resources and energy to participate to this joint effort.

I think that's the main challenge.

The most important thing when you start sharing sEEG data is first having an excellent curation. If it's a bad seizure, when seizures start, when seizures stop. It's about the localization of your electrode, being able to really capture the exact position in the different anatomical structures. And if it's about a cognitive test, you need to perform the cognitive test in your center, then you share the data.

But overall, I think 90 percent of the work has to be done by the clinical person in the sEEG center. And only 5 percent in the end that will lead to the final result.

[00:24:06] Dr. Maryam Nouri: I guess all of this is probably perfectly aligned with ILAE's new neurotechnology section. At the heart of it is that we bring the clinical side to all of these advances.

[00:24:20] Dr. Philippe Ryvlin: Yeah, I think we really need to be on board and we can praise the ILAE for having launched this initiative. It's very recent. We still have to see what it will deliver, but I think it's a very important one. The field is moving so fast in medicine in general with AI. If we're not in now, we'll be out forever.

[00:24:39] Dr. Maryam Nouri: That’s a very good quote to perhaps conclude our discussion today. You're right. If we're not in it now, we may always be left out. I really hope this serves as a sort of an impetus for discussion for everybody who's curious about the field to learn more about this big initiative.

We will include some of the websites and the links attached to this podcast so everybody can gain information. And I do understand that there's also some dedicated expertise and support. So if anybody needed to learn more about how to work around this platform that they can get in contact, correct?

[00:25:25] Dr. Philippe Ryvlin: Absolutely. And everything's free, also. I'm looking forward to having a number of colleagues that will get in touch with us to participate in this great project.