Francois Vialatte is a researcher at Brain Science Institute RIKEN. He works on deciphering brain signals for creating better Brain-Computer interfaces and tools that can diagnose Alzheimer’s at early stages. Learn more about his research.
On this interview BCI researcher Francois Vialatte answers some unpleasant questions surrounding BCI research, such as does BCI really has its future? Also we discuss current problems with publicity of science, potential hypnotic effects of BCI and possibility of entering altered state of mind while playing BCI games.
When and why did you become interested in BCI?
When I was a young Master’s degree student, I had an intuition that machine learning methods could be applied to extract information from brain signals. I had not read anything related to BCI, but was already developing my own theory. Interestingly, I discussed my idea with my roommate, who believed that my theories were some kind of fringe science, or science-fiction fantasy. Regarding dates, it would be 2002… Later on, I had the chance to join a lab where BCI is one of the main research topic.
There is an opinion that hype created by media around BCI is a big bubble which is about to blow, and BCI cannot really compete with its alternatives in any application. Anyone who invests in BCI will lose his money. What is your opinion about this?
There is a part of truth in this affirmation. The media tend to present a distorted picture of what BCI systems can do. This being said, there are certainly numerous applications where BCI will become profitable, with feasible industrial applications.
For example?
Clinical applications for paralyzed patients (especially with ALS) are the most obvious applications. Video games will most probably be a rich field to explore. I think also that neurometrics are viable applications. Somehow, what can be said about BCI is that the technology recently moved from the basic research to the applied research, so it is still far from being optimal.
But what about existing interfaces for disabled persons based on eye-tracking? Can they compete with BCI?
Nice question. Why should eye-tracking be opposed to BCI? Both technologies can be integrated (in this case we generally speak about BMI, or Brain Machine Interface). My opinion is that BCI will be integrated in the future technologies (such as combined eye-tracking/BCI systems)
But can eye-tracking become cheaper and more easy-to-use alternative to BCI?
If you really want to put the two technologies in competition, then there has to be a looser. The more “easy to use” is not the winner, but the one with the higher bit rate transfer. On this aspect, no one knows what the future will be. And there is no guarantee that BCI or eye-tracking would win or loose. This is why I would suggest industrial applications to be based on both (and also on other physiological measurements, such as ECG, etc… when possible).
What are main obstacles that prevent BCI to become widely used?
As I mentioned above, BCI recently moved from the basic research to the applied research. The last step, the development of industrial prototypes, is not completed (or only in the beginning). Some applications could already be developed, but I suppose that companies are waiting until an optimal level of research is reached. Several questions remain opened, many elements are still untested.
What is the coolest thing about BCI?
This technology is one of the most impressive application of the recent progresses in neuroscience. BCI is cool because it is nearly a science-fiction technology. This is probably the reason why it became so fashionable in the media. But to me, BCI is interesting because it let me dream of what could be done at the next step (when we will hopefully have access to more precise measurement technologies, or more accurate mathematical tools).
And something about science publicity… When I just started to using this science papers, I always wondered where is the source code section. Usually only results and description are published, for code, this is not common. What are you opinion?
I agree, it is not common to see it published, and it would be a good habit. This is a problem similar to data withholding in science. Data and code are intellectual properties, the scientist is not always the owner of these elements, and not always encouraged (or even allowed) to share them. Somehow, this is detrimental for developing countries (with less fundings, their scientists have access to less data or code than the developed countries). In other words, modern science is not international, and not public. Only its results are made public. A few journals have, exceptionally, policies asking the authors to submit their data and code together with the paper.
For check?… So the situation is that research groups just want to keep their secret to themselves?
Yes. If you keep your data and your code, then you have some kind of “intellectual trust”. This way you can keep control of your field, and avoid competition.
This not good (should be corrected)…
Yes, I agree, it should be corrected. But it needs some institutional will to move it towards the good direction: research is not directed by scientist, but by administrative policy makers, who are not totally aware of all the aspects of the problem.
Well if even Microsoft finally had to make moves towards opensource, I think for science we won’t wait for too long. But it will require some competitive pressure from independent researchers… otherwise there will be no need for institutions to change existing system.
Note that I understand that the problem is not the same in industry, but there is a hope. A company has to make money on short term, this is obvious, and if they share their code and their data (if they are the only one doing so), it is like shooting in their own foot. But for science, there is hope. If the new generation of young scientists are willing to make these changes, they can spread the idea. Furthermore, they will be the scientific leaders of tomorrow.
Let’s talk about your research in signal synchrony. It is not focused on BCI specifically but BCI improvement is one of its possible applications. But was BCI improvement the main idea from which this study started?
Well, the idea is somehow “bidirectional” improve BCI on one hand, get a better understanding of synchrony on the other hand. Synchrony should increase the reliability of BCI signal processing tasks.
Concerning other applications, I am also working on Alzheimer’s disease. In this context, we use measures of synchrony to investigate the early stages of the disease. For Alzheimer, the purpose is two-fold: obtain a better understanding of the disease, and possibly develop methods to help the diagnostic of the disease before its onset (a task also termed as early detection). Potentially, these measures have many applications. Generally speaking, if brain signal synchrony is a relevant feature for the brain, any technology involving the brain could exploit similar techniques (especially if higher cognition is involved).
So how is synchrony helpful for BCI?
Synchrony provides us with a higher level feature. Synchrony and other methods could be compared metaphorically to the use of a microscope vs. direct observation.
With synchrony more like observation – i.e. it helps us to see the forest for the trees.
Yes, synchrony gives a bigger picture. For my own research, I have to solve the problem of integrating the time, frequency, and space dimensions of EEG (space = electrodes) Fourier measures are electrode-specific, they do not indicate relationships between different brain areas. Other recent measures could be compared with synchrony, like for instance measures of information flow.
Ok, could you tell something about application of synchrony to SSVEP-based BCI?
My preferred BCI systems are the P300 and SSVEP based systems. The reason is that with this system, the subjects do not need lengthy and tiring periods of training. SSVEP produces a stable signal, with a few number of commands it can be a good choice for BCI.
But it is also quite annoying, for how long can you use SSVEP without getting your eyes tired?
It depends on two elements: what paradigm is used, and which frequencies are used. Low frequency is more tiring than high frequencies. For instance one of my subject consistently felt asleep with low frequency “hypnotic” stimuli.
Concerning the paradigm, it is much easier to concentrate on a game than on a screen with numbers. Several studies have shown that BCI systems with a friendly feedback, and especially in a game environment, are more efficient (including a study from our lab using a car driving game).
I heard that 7Hz-frequency stimulation can make person frightened to death. Is it what low-frequency SSVEP can do?
Frightened to death? This looks like an urban legend! Low frequency stimulation is extremely boring and if the fixation is too long, a little hypnotic (it sometimes gives me the feeling of having an empty mind), but I was never scared by these stimuli.
There is however another real danger, which is the risk of photogenic epilepsy. This risk is seldom mentioned in studies about SSVEP, despite being well known. The most famous case of epileptic seizures associated with repetitive flashes of lights was documented after a “pokemon” animation was aired on a Japanese channel. Radford and Bartholomew estimated that 12,000 children suffered from symptoms and/or illness due to this TV show (the symptoms included photogenic seizures, but also photogenic migraines).
Can it be that there was no epilepsy before, but after it was induced, symptoms arose and affected later life?
No, those who had symptoms had already a sensibility (the show did not create the disease), but they did not know. They did suffer from epileptic seizures though. My point is that subjects who undergo experiments with SSVEP-based BCI should be controlled for possible photo-sensibility. Without such control, there is a risk that the subject might have a seizure during the experiment. If they are prone to photosensitive epilepsy, SSVEP-BCI would be dangerous for them, if they are not, then this type of BCI is safe (photo sensitive epilepsy cannot be, as far as I know, induced in healthy subjects). Here is a link about photosensitive epilepsy.
Ok, so it is safe except some special conditions like epilepsy. Just searched Google if low-frequency can cause something bad in healthy persons, and it turned to be “Brown Note” urban legend - you were right.
Yes. Furthermore, concerning repetitive pattern stimuli like SSVEP, you should know that we are constantly exposed to similar stimuli. Computer screens flicker at about 60 Hz, neon lights also flicker, we are surrounded by such stimuli, and as far as I know no one even tried to measure the long term effect of these stimuli.
And, there are such devices sold around – “Mind Machines”: “A mind machine uses pulsing rhythmic sound and/or flashing light to alter the brainwave frequency of the user. Mind machines are said to induce deep states of relaxation, concentration and in some cases altered states of consciousness, that have been compared to those obtained from meditation and shamanic exploration.” Considering that SSVEP uses the same technique – i.e. rhythmic flashes – will we have shamanic experience each time we use it?
I am not aware of any serious publication about this kind of toys. If it did really induce an altered state of consciousness, it should be possible to measure a long term effect (which was not shown). Indeed, SSVEP can have some effect on cognitive functions.
You mentioned there was a hypnotic effect, but very little.
Yes, I did mention this in one of my publications. Synchrony is strongly altered during SSVEP stimulation. I do not rule out the possibility to induce some effect using SSVEP. For instance, one study showed an enhancement of working memory in elders after a 10 Hz stimulation. I did observe the hypnotic effect. But to do so, we need to gather several conditions:
- strong light
- long stimuli duration (> 30 sec)
- large stimuli
Effects might also depend on the stimulus frequencies and colors. I am not aware of any study in BCI using these parameters. So, in other words, you will not have a shamanic experience with an SSVEP-BCI. But the possible effects of SSVEP on higher cognitive function seem intriguing, and I would be curious to read any serious investigation about this topic (I might do one if I have time).
So mind machines might have a point despite no serious publication/effect observed yet.
Yes, my concern is to now what is the real effect (especially long term effect) of this kind of toy to know.
