Characterizing Carbon Fiber Electrodes for Their Application in Wireless Networks of Submillimetric Neural Interfaces

Abstract

Brain machine interfaces (BMIs) are a potential clinical solution for restoring functions lost to individuals suffering from neurological disorders such as spinal cord injury and amyotrophic lateral sclerosis. Enabling a user to control a BMI first requires sampling their brain activity, which is then algorithmically decoded into intended actions. While the conventional intracortical electrodes used for recording neural signals have enabled restoring several everyday human functions, their safety profile limits the clinical viability of BMIs. Implanting these electrodes is accompanied by a complex biological reaction that includes the loss of the neurons that these devices are intended to record. Furthermore, these electrodes often require percutaneous connections through the scalp, which are a persistent infection risk, are surgically limited to too few channels, and tether the user to decoding hardware. Clinically viable BMIs will require new electrode technologies that are wireless, more biocompatible, and can accommodate thousands of channels.

This dissertation presents three studies that aim to resolve this gap by further informing the application of highly biocompatible carbon fiber electrodes with subcellular-scale diameters (6.8-8.4 microns). In the first study, the recording site tips of explanted carbon fiber electrodes and neurons in their immediate vicinity were consistently localized in cortex for the first time. Quantifying the geometries of surrounding neuronal somata revealed that although they were stretched, the positions of the six neurons closest to the electrode were similar to fibers hypothetically implanted into contralateral control brain in simulations. Using these histological positions to simply model the electrophysiology that these neurons might produce suggested that only spikes originating in the nearest four or five neurons may be distinguishable when spike sorting the signals recorded by carbon fiber electrodes.

In the second study, carbon fiber electrodes’ recording capabilities were investigated further. Carbon fiber arrays were implanted into the motor cortex of nine anesthetized rats and electrophysiology was recorded at several depths, sampling from multiple cortical layers each experiment. Large spikes (at least 100 microvolts peak-peak) were reliably acquired in 167/197 recordings (85%) from depths estimated to be in layers 2-6. While spike clusters with the largest amplitudes and highest quantities were measured in estimated layer 5, those acquired in estimated layers 4 and 6 were also numerous and of high amplitude.

In the third study, the feasibility of implanting carbon fiber motes, which are composed of submillimetric chips with one carbon fiber each and are powered and communicate wirelessly, was demonstrated. More than 230 non-functional but mechanically equivalent analogs of proposed designs were assembled to validate a procedure that could easily and swiftly implant up to 25 devices simultaneously. This procedure was successful with 171/186 (92%) analogs inserting such that the chips rested on the cortical surface and only the fibers penetrated the brain 1 mm deep. After implantation, measurements of the change in their spatial arrangements were smaller than those of devices in the literature that were implanted inside the brain.

In summary, these results validate previous findings suggesting that chronically implanted carbon fiber electrodes retain surrounding neurons in motor cortex, deepen our understanding of carbon fiber electrodes’ capabilities in signal acquisition, and introduce a procedure for implanting wireless carbon fiber motes that avoids introducing unnecessary time and complexity to neurosurgery. Collectively, these studies further motivate the application of carbon fiber electrodes in future investigations in neuroscience and in clinical BMIs.