On-line interpretation of imagined temporal patterns from EEG: The next step towards neuronal control of motor and communication prostheses (NGT.6614)

This project aims to combine recent advances in brain imaging and signal processing to develop new tools for direct man-machine interaction (BMI, brainmachine interaction). Recent studies have shown that imagining the execution of a particular sensori-motor task gives rise to almost the same pattern of neuronal activity in the central nervous system as actual performance of the sensori-motor task. We will use this observation and new neuro-imaging methods to develop new methods that enable patients with impairments of the motor system to control devices and even to communicate (via a text display or speech synthesizer).
The idea to "read out" the brain by measuring EEG signals as a measure for intended actions and to use these signals for rehabilitation in patients with motor disorders is not new. The state of the art is that correct decoding of the EEG signal is possible to a very large extend. Although this may seem impressive, it is still not good enough for applications since the erroneous responses in a remaining 10% can lead to completely wrong actions.
The key idea of this proposal is to use temporal modulation of EEG by imagining temporal patterns, such as musical rhythm, imagining a task at a frequency of 1 Hz or a Morse-like code, to enhance the correct decoding rate. Traces of neural activity caused by mental imagery will be picked up by sensors (EEG) and will be classified in discrete categories. These categories will correspond to specific commands, characters or words. In a pilot study we already have been able to proof the feasibility of this method. In contrast with most other BMI work, our approach will use non-invasive measurement techniques, and discrete (symbolic) output. Furthermore, it will not rely on intensive (bio-feedback) training of the user. Instead, the system itself will be adaptive to recognize EEG patterns as representations of response categories on the basis of known examples of measurements of each category separately. To achieve this goal, sophisticated methods to extract the relevant information from the noisy single trial data (e.g. Independent Component Analysis) will be applied, as well as state-of-the-art classification methods.
This proposal deals mainly with the recording and data analysis of EEG signals. Using the results for clinical studies (e.g. functional electro-stimulation of muscles or nerves) would become feasible when the success rate of correct interpretation of EEG signals has been improved.

Utilization
There is a great demand to measure neuronal activity and to interpret brain signals to bypass damaged brain structures with the aim to assist patients during rehabilitation, or to assist patients with motor disorders.
There are several companies in The Netherlands, which focus on the analysis and interpretation of biomedical signals. Biosemi and Brain Research Company consider the proposed project as a logical next step to the existing techniques for neuroengineering. Their interest in the results of this proposal becomes evident from the major investments that these companies are willing to contribute.
The Sony research center in Paris, with various directions of research in music and language, also works on new data-analysis techniques (software and hardware) in the field of neuroscience. Sony is interested in these applications for the control of various appliances and recently founded a new research group for that aim in Tokyo.
The St. Maartenskliniek is one of the major rehabilitation institutes in The Netherlands, and is associated to the University Medical Center Nijmegen. This institute is eager to serve as a test-bed for new advanced technology for rehabilitation. Their interest in using EEG classification for symbolic manipulation can be understood from the goal to enable patients with severe paralyses to control devices, like a TV set or a wheel chair.