Fabien Lotte has a PhD in computer sciences from the National Institute of Applied Sciences (INSA) in Rennes, France. As a Research Scientist at Inria Bordeaux Sud-Ouest in France, his interests include Brain-Computer Interfaces, human-computer interaction, pattern recognition and brain signal processing. During his scientific career, Fabien gained a lot of experience in BCI research and is involved in a few projects. We had the chance to talk with Fabien about his projects and work.
What projects are you currently working on? Can you give me a brief overview and tell me the main goals?
Fabien Lotte: “In our group we are working on 3 main areas related to BCI research:
- EEG signal processing, in order to develop algorithms to decode EEG signals robustly, despite their non-stationarity and noisy nature, and the scarcity of available data to calibrate BCI;
- User training, to understand how users learn to control a BCI (particularly mental imagery based BCI), and how to improve this learning, by providing users with appropriate feedback and training tasks;
- Neuroergonomics and neuroadaptive technologies/passive BCI, to monitor users’ mental states (e.g., workload, attention) during human-computer interaction, to assess the ergonomics quality of this interface, or to adapt dynamically this interface to the user.
Fabien Lotte: “Our funded projects now include BCI-LIFT, which is an Inria Project Lab (gathering various Inria research teams and research centers) aiming at making BCI more “plug-and-play”, by simultaneously improving EEG signal processing, BCI user training and Human-Computer Interaction methodologies. We are also exploring EEG-based rapid information search together with University of Ulster, UK, in a DGA-DSTL funded project.”
“Finally, our core focus now is BCI user training. We showed, both theoretically and practically, that current training approaches for BCI (feedback, training tasks) were suboptimal, and quite likely are major causes of BCI inefficiency and poor performances. We therefore started to study which kinds of BCI users can control a BCI and why, what are they learning when using a BCI and what kind of feedback and training to provide to improve this learning. We started this in ANR (French National Research Agency) project REBEL, and we are going to study that even deeper with a recently obtained ERC Starting Grant – project BrainConquest – that should start in the coming months. The idea is to formally understand and model how users learn BCI control skills and what are these skills, in order to be able to provide optimal training, both for healthy and motor-impaired users.”
What kinds of fields have you been studying with Brain-Computer Interfaces?
Fabien Lotte: “We mostly worked with mental imagery (MI)-based BCI (motor imagery, mental calculation, mental rotation of geometric figures, etc.). First, we mostly studied classifiers and spatial filters design for MI-based BCI, and their use for controlling video games and Virtual Reality applications. More recently, we studied BCI more from a cognitive science point of view, to understand, for instance, psychological predictors of BCI performance. We are also starting to use these BCI for stroke rehabilitation. Finally, since a few years ago, we also work on neuroergonomics, to monitor mental workload, attention or visual comfort from EEG during human-computer interaction tasks.”
Mental Imagery-based BCI user training (left hand motor imagery, mental subtraction, mental rotation). From C. Jeunet, B. N’Kaoua, S. Subramanian, M. Hachet, F. Lotte, “Predicting Mental Imagery-Based BCI Performance from Personality, Cognitive Profile and Neurophysiological Patterns”, PLOS ONE, vol. 10, no. 12, e0143962, 2015
What limitations and challenges does BCI face currently?
Fabien Lotte: “When used for communication and control, the main challenge is the usability of BCI, i.e., their efficacy (how reliably they can decode the user’s mental commands) and efficiency (how fast they can do so, and how fast they can be setup and used). To be used in practice, BCIs need to be much more reliable, need to be fast to setup and calibrate, and should work almost anytime anywhere. We are still far from that. For the moment, BCI do not compare positively with many alternative human-computer interfaces (e.g., for assistive technology or for gaming). Therefore, we should drastically improve the usability of BCI, and identify where and when they can be useful and better than alternatives.”
What solutions have to be developed to overcome these obstacles?
Fabien Lotte: “So far, BCI research has been a very empirical research field, with most findings based on trial and errors. This led to great progress, but now I think the field is mature enough that we can start building new theories and models of BCI. Actually, I believe this is required if we want to make further progress. We need to understand the sources of noise and non-stationarity of the EEG, or how BCI users learn to control a BCI. If we can then model that, then we could further improve the EEG signal processing and BCI user training.”
What are important applications, where BCI technology could be used?
Fabien Lotte: “So far, BCIs are promising for many applications areas, but it remains to be shown whether these promises will hold. BCI are notably promising for assistive technologies, motor and cognitive rehabilitation (e.g., after stroke), human-computer interaction and notably adaptive interfaces reacting to the user’s mental state (e.g., for education, air traffic control, driving and piloting), or for basic neurosciences. Most applications remain at the early stages though, and their value as compared to alternatives is thus still an open question.”
Multi-user motor imagery BCI gaming. From L. Bonnet, F. Lotte, A. Lécuyer, “Two Brains, One Game: Design and Evaluation of a Multi-User BCI Video Game Based on Motor Imagery“, IEEE Transactions on Computational Intelligence and AI in Games (IEEE T-CIAIG), vol. 5, num. 2, pp. 185-198, 2013
You started the OpenViBE platform, where engineers can contribute their software and hardware developments relating to BCIs. What are the major benefits of this platform for the BCI Community?
Fabien Lotte: “OpenViBE is a free and open-source real-time BCI platform (one of the first, if not the first such open source software) that enables anyone to use and design a BCI without writing a single line of code. OpenViBE comes with many already-made BCI tools (P300-speller, motor imagery, SSVEP, etc), which can be used right away. You can also create your own BCI very easily with OpenViBE, simply by visually assembling what we call “boxes”, each box being a brain signal processing or visualization module. It is like building something with Legos, but for BCI! As such, it enables many BCI enthusiasts, hobbyists or researchers, who are not engineers, to use, explore and design BCIs. And for the engineers and/or programmers, OpenViBE can be programmed in C++, and also offers scripting tools in Matlab or Python, so that many people could enjoy and benefit from it! Finally, OpenViBE supports multiple EEG acquisition devices (>40 and growing) and already offers many signal processing and machine learning tools and algorithms. There are many tutorials and documents online on openvibe.inria.fr if you want to try it!”
If you think of BCI in the future, how will it change? What will be the next steps?
Fabien Lotte: “As I mentioned above, I think (or I hope) BCI research will be become more formal and principled, leading to more theories and models. I think that this is an important next step. I also think BCI application areas will increase, and we will see BCI use for many new applications we have not thought of before. EEG caps will become cheaper, easier to wear and use, and more robust. In the longer term, I also hope we will get better sensors than EEG altogether!”
Fabien Lotte: “A trend that is worrying me a bit is the proliferation of pseudo-science products (notably consumer grade ones) or results claiming to be BCI, but that do not reflect rigorous science, or are not BCIs (but systems using muscle activity – EMG – recorded from EEG sensors), and which make extraordinary claims without scientific validation. Hopefully, when the BCI “hype” returns to reasonable levels, these will fade as well. Otherwise, this may generate unrealistic expectations and in turn distrust in BCI research, which would obviously not be good for the field. This thus means that next important steps for BCI also include ethical considerations!”
What’s the ultimate goal of Brain-Computer Interfaces?
Fabien Lotte: “Ideally, to enable anyone to interact effectively and efficiently with a machine using brain activity alone, at any time. This notably means restoring communication and independence for motor impaired users, but also offering new communication and interaction means to healthy users, widening the bandwidth between humans and machines.”
Which new user groups for BCIs do you expect to emerge?
Fabien Lotte: “On the medical side, I would expect several new users with various pathologies or impairments to benefit from BCI (and from Neurofeedback by the way), for rehabilitation and remediation notably. I also expect many more healthy people to start using BCI, e.g., artists for performances, teachers for education, geeks and hackers for fun and science, scientists to answer new questions in neuroscience and psychology, gamers for entertainment, designers and ergonomists for user experience evaluation, etc.”
Neuroergonomics experiment (mental workload monitoring during 3D navigation tasks). From J Frey, M Daniel, J Castet, M Hachet, F Lotte, “Framework for Electroencephalography-based Evaluation of User Experience”, ACM SIGCHI Conference on Human Factors in Computing Systems (ACM CHI), pp. 2283-2294, 2016
How are g.tec products integrated in this environment?
Fabien Lotte: “We heavily use the g.tec g.USBamp in our BCI experiments. We have two g.tec g.USBamps (16 channels each to make a 32 channel system), with active electrodes (g.LADYbird) for about 5 years now, and we conducted almost all our BCI experiments with it. I have to say that it is a very convenient and reliable system! Since it is easily transportable, we also use it for demos or for teaching BCI (and use it during labs). We are also using g.tec’s physiological sensors to measure Galvanic Skin Response (GSR), respiration and heart rate, when we work on mental state monitoring, to estimate, e.g., mental workload from EEG and/or physiological sensors. I am now considering purchasing a g.HIamp multichannel amplifier for our next experiments. But that depends on the upcoming grants.”
What’s the most exciting part of your daily work with BCI?
Fabien Lotte: “The most exciting part of the daily work with BCI is precisely that there is no daily routine with BCI research: it is new every day! It is such a recent, exciting and multidisciplinary field. We are always learning something new, meeting new people, doing new and various tasks, from various fields. I love it!”