Site visit in Espoo on 12 May 2003

The coordination unit made a site visit at the Research Centre for Computational Science and Engineering at the Helsinki University of Technology on 12 May 2003. Academy Professor Mikko Sams together with his research group presented their project for developing an On-line Adaptive Brain-Computer Interface.

The project aims at using physiological signals (EEG and MEG) of the brain for communication and operation by associating the signals to single commands. The result will be a new and noninvasive brain-computer interface (BCI), which can be used for, e.g., producing text, controlling electronic devices or controlling prostheses (artificial arms, etc.)

The project studies the neurocognitive basis of BCIs concentrating on the following fundamental issues.

• The role of the human user in BCI - how qualitatively new functions can be created in which brain activity is directly used for new tasks. Here the role of different learning strategies and the feedback will obviously become crucial.
• How the brain interface iskept constantly tuned to its user - as users gain experience they change their behaviour and brain signal patterns.
• Brain activity of paralysed patients, especially in somatomotor cortical areas, when they attempt to perform movements.

A BCI functions as follows. Brain activation signals are measured and pre-processed. Some features from the signals are extracted and classified and the classification results are then used for controlling a device.

In a first MEG experiment, 8 persons where asked to lift their left and right index fingers (a real motor task). The BCI tries to recognize which action was taken. For feature extraction, a subset of the total of 312 MEG channels is used. The most promising features were defined from the MEG recordings using time-frequency representations. Most prominent changes during the movement occur in 10-Hz and 20-Hz activity. Activity in these frequencies is diminished during and before the movement. After movement, there is a strong "rebound" in activity, which is clearly stronger over the hemisphere contralateral to the movement. This rebound provides one good feature for signal classification. A radial basis function (RBF) classifier is used. The results are promising, leading to almost error-free classification of this particular action (deciding which index finger was moved), when several signals are averaged together. However, the challenge is errorless recognition of the features from single trials.

The results have so far been analysed off-line. The group is gradually moving to real-line measurements and classification. The classifier learns how to classify the signals and in addition the human part should learn (from feedback) how to produce ever better signals (i.e. signals that are easier to classify). A user interface is being built for real-time measurements.

Some of the results have been published recently in Tiedelehti (2/2003), and also presented on Finnish Television (in the OBS programme in December 2002, and coming up, in the PRISMA programme during 2003). The group has extensive collaboration within the Laboratory of Computational Engineering, Brain-Research Unit of the Low Temperature Laboratory of HUT and the Invalid Foundation.

More information
For more information, please contact Academy Professor Mikko Sams at the Helsinki University of Technology. See also the project's web page at http://www.lce.hut.fi/research/bci/ or the web page at this site.

Researchers involved in the project are

Helsinki University of Technology (HUT), Laboratory of Computational Engineering, Research Centre for Computational Science and Engineering:

• Academy Professor Mikko Sams, Academy Research Fellow Jukka Heikkonen, and PhD students Laura Laitinen, Janne Lehtonen and Tommi Nykopp
  

Greger Lindén
Programme Coordinator
Greger.Linden@cs.helsinki.fi