Item

Abstract

On-line classification of slow cortical potentials (SCP) of the EEG allows the characterization of two to four different brain states which can be used for computer based selection of letters or words. 7 completely paralyzed, artificially respirated and fed lockedin patients suffering from amyotrophic lateral sclerosis were trained to produce negative and positive SCP at the Vertex-electrode. After successful training selection of letters or words was possible with a maximum speed of one letter or word per minute. Patients whose training was initiated late at a completely locked-in-state were only able to answer yes-no questions with their SCP. New classification algorithms, in particular support vector machines, allow better classifications. First attempts to build a 150 channel MEG/EEG brain-computer-interface (BCI) for the locked-in and otherwise paralyzed patients are reported. Critical for successful application, however, is the selection of simple behavioral paradigms allowing patients with cognitive disorders and higher animals to produce classifiable brain states. The operant regulation with feedback seems to be an easy to apply and psychologically most adapted paradigm for BCI.