Analog frontend for multichannel neuronal recording system with spike and LFP separation

Abstract

A 0.35microm CMOS integrated circuit for multi-channel neuronal recording with twelve true-differential channels, band separation and digital offset calibration is presented. The measured signal is separated into a low-frequency local field potential and high-frequency spike data. Digitally programmable gains of up to 60 and 80 dB for the local field potential and spike bands are provided. DC offsets are compensated on both bands by means of digitally programmable DACs. Spike band is limited by a second order low-pass filter with digitally programmable cutoff frequency. The IC has been fabricated and tested. 3microV input referred noise on the spike data band was measured.

An Integrated System for Multichannel Neuronal Recording With Spike/LFP Separation, Integrated A/D Conversion and Threshold Detection

“…Several approaches for DC offset stabilization have been reported: Off-chip elements are sometimes employed at input stages [8], [25]. Several fully integrated approaches were also demonstrated: The signal can be capacitively coupled to the amplifier using the polarization capacitance of the electrode, shunt either by a weak-inversion MOS transistor [21] or a reverse-biased diode [26], both delivering a large small-signal impedance to form a low frequency pole at the input.…”

“…For example, it is possible to detect the presence of neuronal spikes as demonstrated in [6] and communicate only active portions of recorded signals, which may lead to an order of magnitude reduction in the required data rate [8]. Another order of magnitude reduction can be achieved if the neuronal spikes are sorted on the chip and mere notifications of spike events are transmitted to the host.…”

Low?Power Architectures for Spike Sorting

“…The main disadvantage of this arrangement is the need to connect a cable to the subject, restricting its movement. Instead, the Neuroprocessor chip performs front-end data processing [13] and data reduction by means of spike detection and sorting to enable bidirectional wireless communications that replace cables and allow free movement of the patient or test subject. While it is feasible to transfer some raw signal recordings over the wireless channel, combining data collected by a large number of electrodes is prohibitive [12], and additional data reduction must be carried out by the Neuroprocessor.…”

“…When a large number of signals are to be handled, typical transmission resources are insufficient [12]. We investigate Neuroprocessors, which are front-end integrated circuits for processing neuronal recordings, with spike sorting capabilities [13]. In this paper we focus on spike sorting algorithms and architectures that trade off some classification accuracy in return for significant savings in power.…”

Design and Development of Smart Brain-Machine-Brain Interface (SBMIBI) for Deep Brain Stimulation and Other Biomedical Applications

“…This module is designed to record the local field potential (LFP) of the neural activity. LFP signals typically are continuous, low frequency signals with frequencies less that 250Hz, and the amplitude up to 5mV [102,103]. In our design, the CPU's built-in ADC is used to convert the analog brain LFP signal To pick the commensurate sampling frequency, we need make a tradeoff between the quality of data recording and the required memory storage space in the electrode.…”