Past events

A scalable design for neuronal recordings using readout integrated circuits and cast microwire bundles

Andreas Schaefer, Professor of Neuroscience UCL

Abstract

In most areas of the mammalian brain, functionally coherent networks consist of thousands to millions of neurons. The distributed nature of neuronal coding within these networks and the reliance of these codes on relative spike timing mean that correlating network activity to complex patterns of behaviour requires that a substantial fraction of the neurons within the network be recorded simultaneously. In addition, short distance connections are very important in neuronal circuits, and unambiguous source attribution within these local assemblies requires that sampling be not only broad but also dense.

An ideal electrophysiological recording technique therefore should employ small electrodes in a highly scalable configuration, in order to record from many single neurons. Finally, to avoid the signal-to-noise problems often associated with small electrodes, each electrode should also be optimised for both low electrode-electrolyte interfacial impedance and low stray capacitance.

Here we present a novel approach that provides a solution to this challenge by combining bundles of insulated metal wires with arrays of highly sensitive amplifiers based on readout integrated circuits (ROICs) from high-speed infrared cameras. Glass-ensheathed metal wires with customisable metal core (2-15 um) and glass (10-40 um) diameters were produced by means of the Taylor-Ulitovsky method, and assembled in bundles with 100-100.000 individual wires using a custom-designed semi-automated process. Individual electrodes had a stray capacitance as low as 0.5 pF/mm, while electrode impedance (1 kHz) was reduced below 30 MΩ by electrochemical Iridium Oxide deposition. The readout of recorded currents was established by using a ROIC of a Xenics Cheetah 640-CL1700 camera, incorporating over 300,000 capacitive transimpedance amplifier circuits with <10fF feedback capacitance at a pixel pitch size of 20 um. Tight but reversible coupling between wire bundle and ROIC pixels was assured using a custom-designed parallel force application system. Noise levels were less than 1% with 14bit full range digitisation and 1.6 kHz full frame (640x512 pixel) readout. Our data suggests that the combination of bundles of insulated metal wires and high-speed ROICs provides a highly scalable approach to neuronal unit recordings.

Short biography

Andreas Schaefer, was appointed Professor of Neuroscience at University College London in 2013 and is a Program Leader at the National Institute for Medical Research, now the Francis Crick Institute, London. Previously, he was Professor of Neuroanatomy at the Institute for Anatomy and Cell Biology at the University of Heidelberg, group leader at the Max-Planck Institute for Medical Research, Heidelberg and a BBSRC David-Phillips fellow at UCL. Schaefer’s research aims to understand how complex behaviour emerges from the properties of molecules, cells and ensembles of cells, one of the key challenges in neuroscience.