State and trajectory decoding of upper extremity movements from electrocorticogram

Electrocorticography has been widely explored as a long-term signal acquisition platform for brain-computer interface (BCI) control of upper extremity prostheses. However, a comprehensive study of elementary upper extremity movements and their relationship to electrocorticogram (ECoG) signals has yet to be performed. This study examines whether kinematic parameters of 6 elementary upper extremity movements can be decoded from ECoG signals in 3 subjects undergoing subdural electrode placement for epilepsy surgery evaluation. To this end, we propose a 2-stage decoding approach that consists of a state decoder to determine idle/move states, followed by a Kalman filter-based trajectory decoder. This proposed decoder successfully classified idle/move states with an average accuracy of 91%, and the correlation between decoded and measured trajectory averaged 0.70 for position and 0.68 for velocity. These performances represent an improvement over a simple regression-based approach.