With advances in neuroscience and technology, brain-computer interfaces (BCIs) offer the potential to restore lost function to individuals with spinal cord injury or neurological disorders. Over the last two decades, numerous studies have demonstrated the utility of BCIs driven by populations of simultaneously-recorded primary motor cortex (M1) neurons. The conventional approach to BCI control involves decoding the pre-existing outputs and/or interactions of M1 ensembles before mapping neuron firing rates to BCI parameters. Earlier studies, however, demonstrated that primates can directly control a single degree-of-freedom output with individual M1 firing rates without conventional neural decoding. If individual M1 neuron firing rates can be voluntarily modulated to control an external device, can arbitrary combinations of M1 neurons likewise be voluntarily co-modulated for BCI control? If so, how are M1 neurons collectively or selectively modulated when arbitrarily mapped to a BCI? To address these questions, two Rhesus monkeys were operantly conditioned to control a one-dimensional BCI by modulating the firing rates of unique ensembles of one to four arbitrarily-selected M1 neurons. In general, BCI performance improved significantly over the course of ~20-30 minute sessions. Although several factors contributed to variability in BCI performance across unique ensembles, BCI performance was generally better with larger ensembles. Encoding redundancy associated with larger ensembles enabled the monkeys to achieve precise BCI control while using a wide range of ensemble co-modulation patterns. In particular, neuron firing rate patterns during BCI target acquisition were preferentially distributed within the task-redundant dimensions of the ensemble's joint firing rate space. Moreover, BCI control neurons were modulated more strongly and precisely than other M1 neurons recorded during the BCI task, indicating preferential modulation of BCI control neurons within single experimental sessions. Arbitrarily-selected M1 neurons can therefore be selectively co-modulated for rapid acquisition of BCI control, demonstrating the inherent flexibility of motor cortex that may also underlie natural motor learning and skill acquisition.
展开▼
