Selective Chronic Recording in the Peripheral Nervous System

Abstract

Reliable interfacing with the peripheral nervous system has been and still remains a difficult problem to solve. Yet the ability to obtain signals from peripheral nerves would have significant benefits such as detection of motor intent in patients with amputation. Similarly decoding signals from the autonomic nervous system would allow continuous monitoring of organ function. However, there are many problems that prevent reliable signal detection in chronic animals and human patients. One of the problems is that axons are arranged in tightly packed bundles surrounded by membranes that are difficult to penetrate. Therefore, access to the signals is challenging, and neural engineers have designed many types of electrodes to address this issue. In this chapter, we will review the various types of neural interfaces such as cuff electrodes, intra- and extrafascicular electrodes, as well as regeneration electrodes. We will focus in one particular type of electrode, the flat interface nerve electrode (FINE), which has been shown to be reliable. It has been implanted in human patients for several years and can provide safe nerve stimulation for sensory substitution and has not been shown to record signals useful for functional recovery. There are many issues specific to chronic recordings that will be discussed. One major problem is EMG contamination, and several approaches to deal with this high amplitude signal will be discussed. Another major issue is the signal-to-noise ratio. The design of ultralow-noise amplifiers particularly well suited for ENG recording will be discussed. Another important issue deals with the recovery of fascicular signals from mixed signals generated by multiple fascicles active simultaneously. Various algorithms capable of extracting and separating fascicular signals will be discussed. Finally, the combination of finite element modeling and computation neuroscience allows accurate models of nerve bundles and recording electrodes. These models can provide important information about the bandwidth required for accurate ENG detection and the effect of the axonal diameter on the recorded signals and can lead to the design of improved peripheral nerve interfaces.