Response of brain tissue to chronically implanted neural electrodes

Abstract

Chronically implanted recording electrode arrays linked to prosthetics have the potential to make positive impacts on patients suffering from full or partial paralysis. Such arrays are implanted into the patient's cortical tissue and record extracellular potentials from nearby neurons, allowing the information encoded by the neuronal discharges to control external devices. While such systems perform well during acute recordings, they often fail to function reliably in clinically relevant chronic settings. Available evidence suggests that a major failure mode of electrode arrays is the brain tissue reaction against these implants, making the biocompatibility of implanted electrodes a primary concern in device design. This review presents the biological components and time course of the acute and chronic tissue reaction in brain tissue, analyses the brain tissue response of current electrode systems, and comments on the various material science and bioactive strategies undertaken by electrode designers to enhance electrode performance.

Introduction

Reports that monkeys accurately and reproducibly controlled a robotic arm via chronically implanted cortical electrodes rekindled the hope of approximately 200,000 patients suffering from full or partial paralysis in the U.S. (Carmena et al., 2003). The implants, made out of dozens of wire electrodes, sampled extracellular potentials from portions of the monkeys’ cortex (Fig. 1). The potentials recorded from neurons adjacent to the electrodes were correlated to observed physical motion, eventually allowing the researchers to translate the neuronal activity directly into robotic arm movements. Because of the length of time required for training and the eventual clinical necessity of a long-term brain–machine interface, the implants were chronic, that is they were permanently implanted into the monkeys’ brains.

The purpose of these implanted devices is to record signals with high signal to noise ratios (SNR) from as many individual neurons, termed single units, as possible. Single units must be separated from the background electrical noise and from the overlapping signals, or multi-unit potentials, of other nearby neurons. Principle component analysis and other signal processing techniques can be used to separate single units from noise and multi-unit potentials (Williams et al., 1999b), with a target SNR for recordings around 5:1 (Maynard et al., 2000, Rousche et al., 2001).

A number of engineering groups are developing multi-channel recording electrode arrays for chronic applications. These designs span several different technologies, including microwires (Williams et al., 1999a, Kralik et al., 2001), polymers (Rousche et al., 2001), and different types of silicon micromachined implants (Drake et al., 1988, Campbell et al., 1991). The basic sensor and instrumentation requirements of such neural implant systems are well established, and in general, the present designs perform as intended in short term studies. However, many designs perform inconsistently in chronic applications, a problem that poses a significant barrier to the clinical development of this promising technology.

For example, Nicolelis et al. (2003) reported a 40% drop in the number of functional electrodes between 1 and 18 months. Only 7 of 11 electrode shafts in Rousche and Normann's implanted array recorded signals at implantation, and the number of electrodes recording signals dropped to 4 of 11 after 5 months (Rousche and Normann, 1998). In a study by Williams et al. (1999a), eight cats were implanted with microwire electrode arrays and electrical activity was recorded over time. Three of the electrode arrays failed within the first 15 weeks, presumably due to the tissue reaction and loosening of the skull cap used to keep the electrode in place. The other electrode arrays remained active until the cats succumbed to unrelated medical complications between 15 and 25 weeks post-implantation. Liu et al. (1999) found a large amount of variation in the stability of neural recordings between different electrode shafts on the same array and between arrays implanted in different cats. Most neuronal recordings in the study either grew in stability until day 80 post-implantation, after which the recording remained stable, or degenerated from a high stability to nearly no signals around 60 days post-implantation.

These studies are reminders of the different requirements of a chronic brain implant suitable for use in basic research and the reliability threshold needed for clinically relevant neuroprosthetics. For clinical applications, implanted microelectrode arrays intended as control and communication interfaces need to record unit activity for time periods of the order of decades (Nicolelis and Ribeiro, 2002) and must have a substantially high performance reliability.

This review highlights what is known of the tissue reaction to implantable electrodes, commonly referred to as the “biocompatibility” of the implant. In addition, we survey some of the potential strategies used to improve the biocompatibility of chronically implanted cortical electrodes. The purpose of the review is not to provide an exhaustive examination into any one research thrust, but rather to present a broad view of the issues surrounding this biocompatibility problem so that computational neuroscientists, biomedical engineers, materials scientists, and neurobiologists can come together with a common understanding of the issues and advance the field.