Elsevier

Microelectronics Journal

Volume 95, January 2020, 104654
Microelectronics Journal

A low-power digitally programmable neural recording amplifier with selectable multiunit activity (MUA) and local field potential (LFP) recording modes

https://doi.org/10.1016/j.mejo.2019.104654Get rights and content

Abstract

This paper presents the design of a low-power implantable neural recording amplifier in a 0.13 μm CMOS technology. The structure consists of a cascade of 4-stages, including instrumentation amplifier (IA), buffer, high pass filter (HPF) and programmable gain amplifier (PGA). The proposed neural amplifier can select between multiunit activity (MUA) and local field potential (LFP) recording modes. A programmable gain is available in the MUA recording mode. Eight gain states can be selected from 58.4 dB to 79.5 dB with 3-bit digital control signals. The digital control signals can be made by a digital signal processing (DSP) unit. In MUA recording mode, the bandwidth is measured from 480 Hz to 8.7 kHz. In LFP recording mode, a constant midband gain of 58.4 dB is achieved in a passband of 0.36 Hz–330 Hz. A noise efficiency factor (NEF) of 1.47 is obtained for the amplifier. Power dissipation is decreased to 1.9 μW per recording channel, by biasing the transistors in the subthreshold region. The proposed neural amplifier can be used for electrocorticography (ECoG) recording in wireless implantable devices.

Introduction

Neural signal recording is used to study the local neural activities in many applications like visual neuroscience and brain research [[1], [2], [3], [4]]. Neural recording techniques are necessary to pinpoint the origin and localizing seizure activities such as Epilepsy, Parkinson's disease, and etc [5]. These techniques are also widely used in cognitive science for human cognition and cortical mapping. The information can then be applied to brain machine interface (BMI) technologies for brain control of external devices [6]. BMIs record brain signals and decode an intended response to control the movement of an external device.

Neural recording is a method of measuring the electrophysiological activities of neurons. Neurons are the basic functional units in the brain and communicate with each other using electrical signals. The electric current flow through the neuron creates a voltage potential on the nerve cell. These electrical currents generate magnetic fields in the brain. Neuronal activation and cerebral blood flow are also coupled, since blood flow increases in the active area of the brain. Therefore, brain activities can be captured in various ways, ranging from imaging techniques by recording magnetic fields in magnetoencephalography (MEG), and detecting changes associated with blood flow in functional magnetic resonance imaging (fMRI) [7,8], to the recording of electrical signals by electrodes in electroencephalography (EEG) and electrocorticography (ECoG). EEG is an electrophysiological monitoring method to record brain activities from outside the skull. It is typically noninvasive, since the electrodes are placed along the scalp. ECoG is another type of electrophysiological monitoring that uses electrodes placed directly on the exposed surface of the brain to record electrical activity from the cerebral cortex. Invasive microelectrodes are used in ECoG, and a craniotomy (a surgical incision into the skull) is required to implant the microelectrodes. Neural recording can be combined with other techniques, e.g., microelectrode recording combined with fMRI technique has been reported in Refs. [7,8].

The recording microelectrodes, placed near the generating cell, isolate and record the neural signals. Highly integrated multichannel recording devices with large channel counts have been developed utilizing multisite neuronal electrodes [9], such as Michigan probe [10] and Utah array [11]. Recently, wireless neural systems have gained popularity for ECoG from the surface of the cerebral cortex, due to the capability of compressing and lowering the transmitted data rate [12]. A multi-channel neural recording system, shown in Fig. 1, consists of several sections. After flowing through the neural amplifiers, the recorded neural signals are collected by an analog multiplexer (MUX), then the signals are sampled and digitized to be transmitted to the digital signal processing (DSP) unit for further processing [13]. The modules placed before DSP may require digital control signals. These digital signals can be provided by the DSP. Every recording channel requires its own neural amplifier. The design and specifications of neural amplifiers have a key role in these systems.

Neural signal, generally known as extracellular signal, includes multiunit activity (MUA) and local field potential (LFP) [[2], [3], [4]]. MUA represents the average spiking activity of small neural populations around the electrode, within a sphere of approximately one-third of a millimeter radius with the microelectrode at its center. It can include dendritic spikes and activity from interneurons [8]. LFPs, predominantly reflect synaptic events, including synchronized afferent or local spiking activity. Spatial summation of LFPs occurs within a couple of millimeters from the electrode tip [8]. LFPs are generated by the electrical activity of a larger population of neurons and can be used to monitor the brain state during sleep, behavioral activation, and etc [1]. Although MUA and LFP research are mostly based on animal experiments, ECoG is often recorded and analyzed in human clinical experiments; real-time estimations of finger movements and arm trajectories can also be produced [14]. Due to their small signal amplitudes, MUAs are easily interfered by internal flicker noise and external noise sources. Therefore, a high-gain and low-noise analog recording front-end amplifier is required to process the weak neural signals acquired by electrode interfaces [15]. In order to achieve implantable biomedical systems for long-term ECoG signal monitoring and processing, both power consumption and chip area of the circuits should be minimized [16]. Accordingly, various types of automatic biopotential detectors are used to locate the targeted MUA waveforms, and process only the relevant contents instead of the entire signals [13,17]. Significant bandwidth reduction can be achieved with such a scheme, since it enables fitting more transmitted channels over a given wireless link. However, these benefits come at the cost of increased complexity.

In the signals recorded by extracellular microelectrodes, MUAs occupy the 0.1–10 kHz frequency band, with amplitudes typically lower than 500 μV. The LFPs occupy the lower frequencies, below 100 Hz, with amplitudes below 5 mV. An effective approach to reduce the required bandwidth and power consumption for implantable and wireless devices is separating the LFP and MUA signal bands at the recording front-end. It also provides separated signals for the next processing stages, like analog to digital converter (ADC) and DSP [8,[17], [18], [19]]. However, the system dynamic range requirement is an issue. Since the MUAs have ten times lower magnitudes than the LFPs, they can be amplified only to one-tenth of the full output swing. To use the full output dynamic range, MUA and LFP signals need to be amplified with different gains. A higher amplification gain is required for MUA compared to LFP signal. A programmable gain amplifier (PGA) can solve the dynamic range issue. Logarithmic PGAs can also be used to compress the high dynamic range of the input signal [[37], [38]].

The main goal of this paper is to develop a low-power digitally programmable neural amplifier for implantable wireless microelectrode recording applications. However, it can be used in other electrophysiologic signal processing applications. The structure is intended to be capable of separating the LFP and MUA signals by selecting between two frequency bands. Moreover, different gains should be applied depending on the selected recording mode. Due to different amplitudes in LFP and MUA recording modes, the structure should be capable of programming the gain and selecting the bandwidth, simultaneously. Low power consumption, low noise, and small area per channel are other requirements for such a structure.

This paper is organized as follows: In section 2, the structure of the proposed neural amplifier is presented. The four stages of the structure are discussed in detail in subsections. The simulation results for the proposed neural amplifier and comparison with recent similar works are presented in section 3, and section 4 concludes the paper.

Section snippets

Neural amplifier design

The proposed neural amplifier in this paper consists of a cascade of 4-stages, including instrumentation amplifier (IA), buffer, high-pass filter (HPF) and programmable gain amplifier (PGA), as shown in Fig. 2. The proposed structure operates in two different modes, i.e., MUA and LFP recording modes. In the state of φ = 0 for the switches S1−5, MUA signals are recorded; and for φ = 1, LFP signals are recorded. In LFP recording mode, the proposed neural amplifier operates with a constant gain,

Simulation results

The proposed programmable neural amplifier in this paper is designed and simulated using a 130 nm CMOS technology. According to subsection 2.4, PGA is digitally programmed into eight gain states by changing the states of the switches S6−8. PGA gain is programmed in the range of 9.4 dB–30.5 dB. The total gain of the structure during MUA recording is programmed in the range of 58.4 dB–79.5 dB. Table 1 shows the PGA and total gains of the proposed structure for eight states of the switches.

The

Conclusion

The paper presents the structure, operation, and simulation results for a low-power programmable neural recording amplifier. It has two selectable recording modes, i.e., MUA and LFP recording modes. The low and high cutoff frequencies of the proposed neural amplifier in LFP and MUA recording modes are 0.36 Hz–480 Hz, and 330 Hz - 8.7 kHz, respectively. By biasing the transistors in the subthreshold region, power dissipation is decreased to 1.9 μW. The proposed neural recording amplifier is

References (38)

  • N.K. Logothetis

    The neural basis of the blood-oxygen-level-dependent functional magnetic resonance imaging signal

    Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci.

    (2002)
  • K.D. Wise et al.

    Wireless implantable microsystems: high-density electronic interfaces to the nervous system

    Proc. IEEE

    (2004)
  • Q. Bai et al.

    A high-yield microassembly structure for three-dimensional microelectrode arrays

    IEEE (Inst. Electr. Electron. Eng.) Trans. Biomed. Eng.

    (2000)
  • R. Muller et al.

    A minimally invasive 64-channel wireless uecog implant

    IEEE J. Solid State Circuits

    (2015)
  • A.M. Sodagar et al.

    A fully-integrated mixed-signal neural processing module for implantable multi-channel cortical recording

  • T. Yanagisawa et al.

    Electrocorticographic control of a prosthetic arm in paralyzed patients

    Ann. Neurol.

    (2012)
  • K.A. Ng et al.

    Implantable neurotechnologies: a review of integrated circuit neural amplifiers

    Med. Biol. Eng. Comput.

    (2016)
  • C. Wu et al.

    A 16-channel cmos chopper-stabilized analog front-end ecog acquisition circuit for a closed-loop epileptic seizure control system

    IEEE Trans. Biomed. Circuits Sys.

    (2018)
  • Y. Perelman et al.

    An integrated system for multichannel neuronal recording with spike/lfp separation, integrated a/d conversion and threshold detection

    IEEE (Inst. Electr. Electron. Eng.) Trans. Biomed. Eng.

    (2007)
  • Cited by (4)

    • A front-end amplifier with tunable bandwidth and high value pseudo resistor for neural recording implants

      2022, Microelectronics Journal
      Citation Excerpt :

      In this structure, the low cut-off frequency depends on the transconductance of the amplifier and no pseudo-resistor has been used which increases the power consumption and noise of the circuit. In [11], four stages are used to design the front-end amplifier where the cut-off frequency is tuned by changing the capacitors of the stages. It also uses the pseudo-resistors in the AP mode to generate a proportional low cut-off frequency, but due to the use of pseudo-resistors in the subsequently blocks, the linearity in this system is reduced because the voltage variations across the pseudo-resistors will be increased.

    • A high-sensitivity, low-noise dual-band RF energy harvesting and managing system for wireless bio-potential acquisition

      2021, Microelectronics Journal
      Citation Excerpt :

      The wireless bio-potential acquisition system consists of a bio-signal instrumentation amplifier (Bio-IA), an ADC, and a wireless transmitter. Generally, the supply voltage of the Bio-IA is between 1 V–1.8 V [1–5] to avoid the amplifier output saturation [6] and reduce power consumption. Similarly, 1 V or lower supply voltage is also used in ADC and transceiver to reduce power consumption and improve the signal-to-noise ratio.

    View full text