Low-frequency local field potentials in primate motor cortex and their application to neural interfaces

Abstract

For patients with spinal cord injury and paralysis, there are currently very limited options for clinical therapy. Brain-machine interfaces (BMIs) are neuroprosthetic devices that are being developed to record from the motor cortex in such patients, bypass the spinal lesion, and use decoded signals to control an effector, such as a prosthetic limb. The ideal BMI would be durable, reliable, totally predictable, fully-implantable, and have generous battery life. Current, state-of-the-art BMIs are limited in all of these domains; partly because the typical signals used—neuronal action potentials, or ‘spikes’—are very susceptible to micro-movement of recording electrodes. Recording spikes from the same neurons over many months is therefore difficult, and decoder behaviour may be unpredictable from day-today. Spikes also need to be digitized at high frequencies (~104 Hz) and heavily processed. As a result, devices are energy-hungry and difficult to miniaturise. Low-frequency local field potentials (lf-LFPs; < 5 Hz) are an alternative cortical signal. They are more stable and can be captured and processed at much lower frequencies (~101 Hz). Here we investigate rhythmical lf-LFP activity, related to the firing of local cortical neurons, during isometric wrist movements in Rhesus macaques. Multichannel spike-related slow potentials (SRSPs) can be used to accurately decode the firing rates of individual motor cortical neurons, and subjects can control a BMI task using this synthetic signal, as if they were controlling the actual firing rate. Lf-LFP–based firing rate estimates are stable over time – even once actual spike recordings have been lost. Furthermore, the dynamics of lf-LFPs are distinctive enough, that an unsupervised approach can be used to train a decoder to extract movement-related features for use in biofeedback BMIs. Novel electrode designs may help us optimise the recording of these signals, and facilitate progress towards a new generation of robust, implantable BMIs for patients.