At CeBIT 2012 University of Warsaw presented a wireless brain-computer interface (BCI) system, called BCI Appliance. The device is a tablet-sized box with just one button, running entirely on Open Source software. Neurogadget.com could ask some questions via email about the BCI Appliance to the leader of the project. Welcome to an exclusive interview with Piotr J. Durka, professor of the University of Warsaw, Department of Physics.
Neurogadget: What commands the BCI Appliance is able to recognize? Does it “understand” thoughts like the letters of the alphabet or command such as push, pull, etc?
Professor Piotr Durka: The BCI Appliance gives you a dynamically adjusted set of choices. This video (raw, unedited) should explain your question:
What is the ultimate goal of the BCI Appliance?
To bring the BCI from lab to bedside. In the current stage, it is a successful proof of concept, intended to:
- Show that BCI can be implemented as a simple box with just one button, instead of a complicated computer-based system. The whole system fits inside the tablet-sized box, attached to the arm of a chair, battery powered and wirelessly connected to the EEG amplifier, robot, and optional computer that displays the brain waves (just for presentation).
- Implement a novel idea of stimulus rendering for the fastest BCI paradigm. That’s more technical, you have all the details on the openbci.pl wiki. Briefly: SSVEP (Steady State Visual Evoked Potential) is the fastest and simplest way to make a BCI, relying on concentrating users attention on flashing symbols. If you see right arrow flashing 17 Hz and left arrow flashing 19 Hz, and think of (concentrate on) the right arrow, there will be a trace of 17 Hz in your brain waves. The problem is that the frequencies around 10-20 Hz, which produce the strongest response, are also tiring (like a broken fluorescent lamp) and in some cases can induce a photoepileptic attack. Higher frequencies, say above 30-35 Hz, are cool and safe, but they are almost impossible to render reliably on a computer screen – due to limited refresh rates of displays and non-realtime nature of modern operating systems. We made it as flexible as rendering on computer screen (dynamically changeable menus) and as stable as hardware generators (LEDs flashing under LCD).
- Cut down the costs of BCI. We developped a complete Open Source framework for BCI, including EEG display for diagnostics using Svarog (signal viewer, analyzer and recorder on GPL). Also, BCI Appliance runs GNU/Linux that makes it possible to use lightweight and custom (Just Enough) OS, saving on hardware, battery life etc.
Where are you now on your development timeline?
Mission Accomplished :-) Development of the Appliance beyond a proof of concept is no more “scientific enough” to be pursued at the University, it would require forming a spinoff (which is quite possible in close future) and some dedicated grant (which is much less certain).
Is it going to be a standalone product or a prototype as part of a bigger scientific project, namely the BRAIN ?
The device was developed by more or less the same team that worked with BRAIN (BCIs with Rapid Automated Interfaces for Nonexperts), but it was not a part of this project. From the idea to the implementation it was just Department of Physics, University of Warsaw, Poland.
Products such as the BCI Appliance may be the only option for communication for people with neurodegenerative diseases. Have you tested your device with patients suffering ALS syndrome?
This is one of the theoretical goals of the whole BCI research, but we are hard science department, and basically do not work with patients.
The BCI Appliance is claimed to be the fastest BCI that has ever been presented at CeBIT. Is the installation of the headset fast, or the user adaptation time is very short?
I meant this year, “ever” should come at least with “probably” :) Headset (actually a headwrap) installation is indeed fast, this is due to a novel technology of water-based electrodes from our partners, Dutch company TMSI. User adaptation is rather system’s adaptation to the user’s brain waves. In this case called calibration, it takes few minutes – about a standard. The most interesting is the speed of operation. We have a choice (one in eight) per each decision, each decision taken from 4- or 2-seconds epoch. The actual bitrate would also depend on the error rates and would greatly differ between users. So we did not give any “expected performance”. We hoped for professional IT journalists to come to our booth and check it :-) A person was able to write with 4-second epochs per choice, and high-frequency SSVEP.
When we first got an email informing us about the presence of the BCI Appliance at CeBIT, the mail was sent from the Polish Ministry of Economy. Is it a governmental project?
Not directly, you may call this a University project. The government so far has been quite reluctant financing our research.
Do you see the potential in non-invasive brain-computer interfaces to become mainstream in the near future?
That’s a difficult, at best emerging, market. The speed and reliability of BCI is still much lower than other communication channels based on muscles or even gaze tracking, also sensors (EEG electrodes) are still quite far from “just putting a headcap on”. So a really convenient non-muscular communication remains a challenge, and it’s hard to make reliable predictions.
Fortunately, scientists not always have to work according to strict business plans. If this challenge will be pursued, major problems may be solved one day, just like so many things that seemed impossible some years ago. If you think of the hell on Earth experienced by people in locked-in state, you feel that this research is important. If you think about moving objects and communicating with nothing but your thoughts, you feel it’s cool and fascinating.