Representation of natural arm and robotic arm movement in the striatum of a marmoset engaged in a two choice task

Abstract

Modifications to the Body Schema refers to the idea that the brain's mapping can be extended to include an external device. Recent research has shown how Brain-Machine Interfaces (BMIs) can be thought of an extension of this concept. While most BMI work has focused on the cortex in relation to this idea, here we study the striatum to see how the actions of a robotic arm are represented in a deep brain structure. We show consistent neural firing patterns in the striatum depend on the movement direction of a robot even when the robot movement is preceded by distinct movements of the actual limb. Neural data during the robot movement was analyzed using a sliding time window (500ms with 400ms overlap) and was compared to the baseline firing when there was no natural reach or robot movement. We achieved an average accuracy of 88.4% over 4 sessions using unsupervised clustering techniques to classify the robot movement types.