A brain-computer interface based on micro-ecog arrays for recording and stimulation in rat somatosensory cortex

In this thesis, we studied the recording and stimulation capabilities of microelectrocorticography (µECoG)-based brain computer interfaces (BCI) in the hindlimb area of rat primary somatosensory cortex (SI). First, rats were implanted with 16-channel commercial platinum-iridium (Pt-Ir) or newly developed reduced graphene oxide (rGO) µECoG arrays. In recording sessions, unrestrained rats performed self-paced lever-press tasks, in which vibrotactile stimulation at two frequencies and three amplitudes was presented to the hindlimbs of the rats following the lever-press. Both lever-press and somatosensory information were predicted from µECoG recordings in the same brain area using a support vector machine (SVM) algorithm. rGO and Pt-Ir µECoG arrays showed similar decoding performances. While motor intention and the presence of sensory stimulus were decoded with high accuracy, vibrotactile frequency and amplitude classification performances were moderate. Additionally, some rats performed a yes/no vibrotactile detection task, in which right/left levers were pressed. By using several machine learning algorithms, right/left lever presses were decoded. Thus, both experiments showed that motor information can be decoded from rat SI hindlimb area. In a different study, freely moving rats were subject to cortical electrical stimulation at various parameters to measure motor threshold and discomfort levels by video analysis. Then, artificial sensation thresholds were measured psychophysically in a yes/no detection task by estimating psychometric curves as a function of stimulus parameters. Overall, this thesis shows that SI hindlimb area is suitable for BCI applications with µECoG electrodes. The newly devoleped rGO electrodes yielded good performance in chronic recording and stimulation experiments.