Identifying functional connectivity of motor neuronal ensembles improves the performance of population decoders

Estimating the response properties of cortical neurons is an essential step to decode movement intentions in cortically-controlled brain machine interface applications. Among these properties is the variable degree of interaction between neurons while subjects carry out similar motor tasks. In this paper, we use a dynamic model of motor encoding, previously shown to fit experimental data from primary and supplementary motor areas in nonhuman primates, to demonstrate the utility of identifying interaction patterns in improving decoding performance. Neuronal interaction is quantified by estimating the functional connectivity among neurons in a cooperative network that are driven by heterogeneously-tuned neurons in an input noncooperative network. A reward-based functional plasticity is induced in the model during repeated execution of a center-out reach task and the connectivity is continuously estimated to track changes in the interaction patterns. Results demonstrate that the ability to track cortical adaptation can contribute significantly to improvement in motor control of neuroprosthetic devices.