CMOS-Based High-Density Silicon Microprobe Arrays for Electronic Depth Control in Intracortical Neural Recording–Characterization and Application

This paper reports on the characterization and intracortical recording performance of high-density complementary-metal–oxide–semiconductor (CMOS)-based silicon microprobe arrays. They comprise multiplexing units integrated on the probe shafts being part of the signal transmission path. Their electrical characterization reveals a negligible contribution on the electrode impedances of $139 \pm 11\ \hbox{k}\Omega$ and $1.2 \pm 0.1\ \hbox{M}\Omega$ and on the crosstalks of 0.12% and 0.98% for iridium oxide $( \hbox{IrO}_{x})$ and platinum (Pt) electrodes, respectively. The power consumption of the single-shaft probe was found to be 57.5 $\mu\hbox{W}$ during electrode selection. The noise voltage of the switches was determined to be 5.6 $ \hbox{nV}/\surd\hbox{Hz}$; it does not measurably affect the probe performance. The recording selectivity of the electrode array is demonstrated by electrical potential measurements in saline solution while injecting a stimulating current using an external probe. In-vivo recordings in anesthetized rats using all 188 electrodes with a pitch of 40.7 $ \mu\hbox{m}$ are presented and analyzed in terms of single neural activity and signal-to-noise ratio. The concept of electronic depth control is proven by performing mechanical translation of the probe shaft while electronically switching to adjacent electrodes to compensate the mechanical shift. $\hfill$[2012-0027]

I. Introduction

High-Density intracortical recording of extracellular potentials is a method to investigate the complex communication among the overwhelming number of neurons in the brain [1]–[3]. Neuronal densities are about 50 000 , depending on the brain region [4]. However, most of the neurons show only sparse spontaneous activity, reducing the number of neurons recordable simultaneously [5]. The recordable extracellular action potential decreases rapidly as a function of the distance from its source, i.e., the neuronal soma and its axon and dendrites [6], [7]. As a consequence, the positioning of the recording electrodes close to these structures is a prerequisite to distinguish single-unit activity (SUA) from background neuronal activity and thermal electrode noise with sufficient signal-to-noise ratio (SNR) [6]–[9]. Experiments [1]–[3], [6]–[9] and simulation [6] have indicated that a distance of 50 or less is a safe choice for this purpose. At distances over 50 , individual spikes can still be recognized. However, the reliability of spike separation is decreased significantly [7]. with a separation distance of electrodes on a neural probe of about 50 , one can therefore be reasonably confident to be able to detect the potential SUA of any neuronal structure within a distance of 50 from the probe. Depending on the orientation of a neuron and its axon and dendrites with respect to the probe, signals may be recorded on a single electrode or a group of neighboring electrodes [7], [10], [11].