Decoding Hand Trajectories from Micro-Electrocorticography in Human Patients

Abstract

A Kalman filter was used to decode hand trajectories from micro-electrocorticography recorded over motor cortex in human patients. In two cases, signals were recorded during stereotyped tasks, and the trajectories were decoded offline, with maximum correlation coefficients between actual and predicted trajectories of 0.51 (x-direction position) and 0.54 (y-direction position). In a third setting, a human patient with full neural control of a computer cursor acquired onscreen targets within 6.24 sec on average, with no algorithmic constraints on the output trajectory. These practical results illustrate the potential utility of signals recorded at the cortical surface with high spatial resolution, demonstrating that surface potentials contain relevant and sufficient information to drive sophisticated brain-computer interface systems.

Additional details