Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Neural mechanisms of operating an intracranial brain-computer interface for imagined speech decoding
Poster B8 in Poster Session B, Tuesday, October 24, 3:30 - 5:15 pm CEST, Espace Vieux-Port
Silvia Marchesotti1, Pierre Mégevand2,3, Laurent Spinelli2, Anne-Lise Giraud1; 1Speech and Language Group, University of Geneva, 2Service de Neurologie, Hôpitaux Universitaire de Genève, 3Human Neuron Lab, University of Geneva
Brain-computer interfaces (BCIs) for speech decoding can provide a new means of communication for patients that have lost the ability to communicate with the surroundings. Recent years have seen great advances in the field of speech-BCI, with systems able to decode attempted speech with impressive effectiveness. However, decoding motor representations cannot be exploited in disorders of language, such as aphasia, in which the damage concerns regions involved in speech production which are located upstream with respect to the motor representations. A BCI suitable for these disorders would rather require decoding representation of speech units produced through imagined, rather than attempted, speech. So far, only a handful of studies have attempted the decoding of imagined speech in real time, thus the knowledge on the neural mechanisms of operating a speech-imagery BCI remains scarce. In this study we have developed a BCI based on intracranial EEG recordings to investigate the neural correlates of operating a syllable-imagery BCI. Four patients, implanted with stereotactic electrodes for pre-surgical epilepsy evaluation, controlled a visual feedback presented on a computer screen by imagining pronouncing one of two syllables, chosen with different articulatory and phonetic features. Results show a high-interindividual variability in the decoding features, but overall similar decoding accuracy across participants (around 60%). Features appeared to be stable across days, as assessed in one patient that agreed to take part in a three-days training. Real-time BCI-control was rather limited and variable across participants, likely due to differences in the site of the implant. These results shed light on the neural mechanisms involved in operating a covert speech-BCI and will be applied in the future to the improvement of the current BCI systems.
Topic Areas: Language Production, Speech-Language Treatment