About 15% of the world's population lives with some form of disability. The invasive brain-machine interfaces (iBMIs) interpret brain signals into corresponding functions by advanced signal processing methods, which can be a promising solution to help disabled people. Since the brain has multiple cortical regions contributing to a single function, it is essential to have a better understanding of the relationship among regions to develop better iBMIs.

We are interested in modeling the functional relationship between the medial prefrontal cortex (mPFC) and primary motor cortex (M1), which are both actively involved in motor control. However, how the information is conveyed in spike trains between two regions has not been fully revealed by computational models. We investigate the co-acti...[ Read more ]

About 15% of the world's population lives with some form of disability. The invasive brain-machine interfaces (iBMIs) interpret brain signals into corresponding functions by advanced signal processing methods, which can be a promising solution to help disabled people. Since the brain has multiple cortical regions contributing to a single function, it is essential to have a better understanding of the relationship among regions to develop better iBMIs.

We are interested in modeling the functional relationship between the medial prefrontal cortex (mPFC) and primary motor cortex (M1), which are both actively involved in motor control. However, how the information is conveyed in spike trains between two regions has not been fully revealed by computational models. We investigate the co-activation between mPFC and M1 by utilizing point process models with different nonlinear capacities. Sprague Dawley (SD) rats with microelectrode implanted in both areas were trained to learn a new behavior task. Neural spike data was recorded during the learning procedure. The general linear model, the second-order general Laguerre Volterra model, and the staged point-process model are implemented to predict spike trains in M1 neurons from spike train input of mPFC neurons.

We find that M1 neural spike trains can be well predicted from mPFC neural spikes, which indicates a highly correlated functional relationship between mPFC and M1 during task learning. By comparing the performance across models, we find that models with higher nonlinear capacity significantly perform better than linear and 2^nd order models, which shows the relationship between mPFC and M1 is highly likely to be nonlinear. Finally, we find that the models perform the best when the subjects become well trained with the new task comparing with the start and middle stage of learning. We conclude that the correlation between mPFC and M1 evolves more during task learning.