Learning the contributions of the motor, premotor, and posterior parietal cortices for hand trajectory reconstruction in a brain machine interface

摘要

The ability to record, in real-time, the activity of hundreds of cortical neurons gives the ability to selectively study the function of clusters of cortical neurons in Brain Machine Interface (BMI) experiments. We have demonstrated using a recursive multilayer perceptron (RMLP) that using the appropriate signal processing theory in a well-chosen parsimonious model, we can develop constructs that agree with basic physiological modeling of neural control. By looking through the trained model, we have found interesting relationships between the neuronal firing and the movement. The RMLP allows us to continuously study the relationship between neural activity and behavior without the active interference of the experimenter. The findings presented in this study offer an opportunity for the neuroscience community to compare the cortical interactions as constructed by the RMLP to what is known about motor neurophysiology.