Tomislav Milekovic (Brown University, USA)

Abstract:
Brain-computer interfaces (BCIs) hold great potential for restoring movement to people with paralysis by providing paralyzed individuals with the ability to control an external effector, such as a computer cursor or a robotic limb using their neural activity. The most powerful BCIs are based on recordings of action potentials of ensembles of individual neurons recorded by intracortically implanted microelectrode arrays (MEAs). BCI control could be further increased by conditioning of neuronal activity, as demonstrated by recent experiments with non-human primates. However, recordings of action potentials from MEAs are unstable across periods of days to months. Signals from individual neurons can be lost and the number of recorded neurons can diminish over time, making exploitation of conditioned neuronal activity for improved BCI control difficult. In addition to action potentials, intracortically implanted MEAs simultaneously record local field potentials (LFPs). Recent studies showed that LFPs recorded in the primary motor cortex of people with paralysis modulate with attempted movements, thus making BCI control using LFPs feasible.
We investigated the stability, reliability and precision of LFP-based BCIs for people with paralysis. In one study, an individual with tetraplegia used an LFP-based text-entry communication BCI to communicate with family members, type messages and write emails for over 3 months without any change in performance. During this time, decoding algorithm used to enable BCI control has remained fixed. In another study, two individuals with paralysis controlled a computer cursor in one dimension by modulating high gamma frequency LFPs recorded by a single MEA electrode. Our results indicate that LFPs can be used for BCI control, either as a supplementary signal to action potentials, or as a reliable, stable and sustainable alternative.