An individual neuron communicates using action potentials (“spikes”) to evoke synaptic potentials in a downstream neuron. Understanding how the brain performs complex functions, from sensory perception to decision making, requires analysis of this spiking activity over space and time. However, action potentials are brief (1 to 2 ms), small-amplitude events relative to voltage fluctuations in the extracellular space, making them difficult to record. Typically, recording electrodes must be inserted into the brain and placed in close proximity to neurons in order to acquire individual action potentials. Implantation of electrodes damages brain tissue and therefore limits recording to small, restricted areas, especially in humans. When information about neural activity over larger areas is required, for instance to determine the onset zone of a seizure or localize a certain brain function to an area of cortex, electrodes are placed on the surface of the brain (electrocorticography).

These surface electrodes record local field potentials (LFPs) that represent the summed input synaptic activity of hundreds to thousands of neurons. Although analysis of LFP recordings provides insight into neural processes, the relationship to neurons’ output in the form of spiking is not straightforward. The capacity to record both action potentials and LFPs without penetrating the brain surface would greatly increase the amount of information available in a single recording without introducing the potential complications of invasive devices.

In a recent issue of Nature Neuroscience, we demonstrated that a novel neural interface device, the NeuroGrid, can record action potentials of cortical neurons without penetrating the surface of the brain. This ability is due to several key design elements: (i) recording electrodes with similar size and spacing compared to neuronal cell bodies in cortex; (ii) use of a conducting polymer (PEDOT:PSS) instead of conventional metals as the interface between brain and electronics to increase the efficiency of neural signal transduction; (iii) advanced microfabrication techniques to generate a thin (4 μm) and highly conformable overall structure that can closely adhere to the complex curves of the cortex.

LFP and action potentials were recorded with the NeuroGrid both chronically in vivo in rats and intra-operatively from human patients undergoing epilepsy surgery. Action potentials in rat somatosensory cortex were stable in waveform appearance and distribution over recording electrodes for more than 1 week. Identification of action potentials generated by a putative single neurons (clustering) revealed characteristics suggestive of both excitatory and inhibitory neurons. Because the most superficial excitatory cortical neurons are located approximately 200 μm below the cortical surface (layer II), the NeuroGrid appears to have a detection radius of at least this distance.

This study establishes the NeuroGrid as an effective tool for large-scale recording of LFPs and action potentials from the cortical surface with minimal risk of tissue damage. Such a device could be used to record from human cortical areas characteristically too sensitive to risk implantation of penetrating electrodes, or within sulci/fissures that are inaccessible to conventional surface arrays. Data generated will facilitate understanding of the relationship between LFPs and spiking activity in different cortical regions, and could be used for control of brain–machine interfaces or therapeutic closed-loop stimulation in brain disease.

—Dion Khodagholy Araghy, PhD, and Jennifer N. Gelinas, MD, PhD

Read the paper “NeuroGrid: Recording action potentials from the surface of the brain” in Nature Neuroscience, published December 22, 2014.