Motor trajectory decoding based on fMRI-based BCI — A simulation study

Abstract

Recent brain computer interface (BCI) studies using chronically implanted microelectrode array demonstrated that electro-physiological responses from primary motor cortex (M1) can be successfully used to control a robotic arm by reading subjects' intention to move their arm [1]. In order to avoid the invasiveness of electrophysiological recording, more non-invasive approaches such as EEG or fMRI was likewise proposed. However, most non-invasive BCI studies suffer from the fact that they classify brain differential activity states, rather than deciphering the actual neural responses underlying the target behavior. In this simulation study, in order to decode the brain activity states underlying the target behavior from the fMRI signals, we found the directional tuning properties, a basic functional property of neural activity in M1, at the voxel level for motor trajectory decoding, and we performed a simulation to demonstrate that it is feasible to control the robotic arm in real time based on multi-voxel patterns.