Sensors 20 00904 v2
Uploaded by
generationSensors 20 00904 v2
Uploaded by
generationsensors
Review
Multi-Channel Neural Recording Implants: A Review
Fereidoon Hashemi Noshahr 1, * , Morteza Nabavi 1 and Mohamad Sawan 1,2,3
1 Polystim Neurotech. Lab., Department of Electrical Engineering, Polytechnique Montreal,
Montreal, QC H3T 1J4, Canada; [email protected] (M.N.); [email protected] (M.S.)
2 School of Engineering, Westlake University, Hangzhou 310024, China
3 Institute of Advanced Study, Westlake Institute for Advanced Study, Hangzhou 310024, China
* Correspondence: [email protected]
ÄÅÇÉÑÖÜ
Received: 30 December 2019; Accepted: 4 February 2020; Published: 7 February 2020 ÄÅÇÉÑÖ
Abstract: The recently growing progress in neuroscience research and relevant achievements, as well as
advancements in the fabrication process, have increased the demand for neural interfacing systems.
Brain–machine interfaces (BMIs) have been revealed to be a promising method for the diagnosis
and treatment of neurological disorders and the restoration of sensory and motor function. Neural
recording implants, as a part of BMI, are capable of capturing brain signals, and amplifying, digitizing,
and transferring them outside of the body with a transmitter. The main challenges of designing
such implants are minimizing power consumption and the silicon area. In this paper, multi-channel
neural recording implants are surveyed. After presenting various neural-signal features, we investigate
main available neural recording circuit and system architectures. The fundamental blocks of available
architectures, such as neural amplifiers, analog to digital converters (ADCs) and compression blocks,
are explored. We cover the various topologies of neural amplifiers, provide a comparison, and probe
their design challenges. To achieve a relatively high SNR at the output of the neural amplifier, noise
reduction techniques are discussed. Also, to transfer neural signals outside of the body, they are
digitized using data converters, then in most cases, the data compression is applied to mitigate power
consumption. We present the various dedicated ADC structures, as well as an overview of main data
compression methods.
Keywords: neural recording implant; brain–machine interface; analog front-end; neural amplier;
low-noise amplier; chopper stabilization technique; compressive sensing
1. Introduction
In the past decade, researchers have worked on the brain to understand its functions and monitor the
brain’s electrical signals to research, diagnose and treat its disorders, as well as to utilize these signals to
control artificial limbs. Brain–machine interfaces (BMIs) can serve people with different clinical disorders.
For example, researchers have implemented robotic limbs [1,2], speech synthesizers [3], and human
neuroprosthetic control of computer cursors [4–6], utilizing less than 300 electrodes [7]. In addition,
monitoring the bio-potential signals is a fundamental and vital part of a medical diagnostics system.
For this purpose, patients are generally connected to a massive bio-potential acquisition equipment.
However, this limits the patient’s daily routine on the one hand, and on the other hand requires the
long-term arduous monitoring of diagnostics [8]. One of the most promising solutions is to use neural
recording implants as a part of BMI systems, which are in high demand and are being developed and
improved as technology develops.
Inability to record from large numbers of neurons has limited the development of BMI.
Noninvasive methods are capable of recording millions of neurons through the skull, however this
signal is nonspecic and distorted [9,10]. Utilizing electrodes placed on the surface of the cortex,
Sensors 2020, 20, 904; doi:10.3390/s20030904 www.mdpi.com/journal/sensors
Sensors 2020, 20, 904 2 of 29
an invasive method, records proper signals. However, the disadvantage of this is not being able to
record deep in the brain and they average the activity of thousands of neurons [11]. Invasive techniques
have been utilized by some BMIs. This is because recording single action potentials from neurons
in distributed, functionally-linked ensembles are necessary for the most accurate readout of neural
activities [7]. Therefore, increasing the spatial resolution and the number of electrodes are essential for
developing BMI.
The implementation of a neural recording implant is multi-disciplinary, as it involves the various
scientic elds such as electronics, medical, materials, electrodes, and system integration. Increasing
the number of electrodes and, consequently, the number of channels (in the range of thousands),
creates new challenges for neural recording in the various elds mentioned. Microelectrode technology
is not appropriate for these large-scale recordings [12]. Recently, Neuralink Company has built arrays
of small and exible electrodes (3072 electrodes per array), which have enabled thousands of channel
recordings [13]. In the microelectronics eld, large-scale recordings create many challenges with
regards to decreasing the power consumption and chip area.
In the design of the neural recording implants, the two constraints, the power consumption and
chip area, should be addressed. Implantable circuits should consume very low power to avoid any
damage to the surrounding tissue due to generated heat. Additional challenges in the design of the
analog front-end (AFE) of the neural recording systems arise in advanced and scaled technologies.
The main reason is due to the short-channel effects of MOS transistors. These effects in the MOS
down-scaled technologies decrease the transconductance (gm ) of the transistor on one hand and on the
other hand increase the gate leakage current, the icker, and thermal noise power of an MOS transistor.
This creates challenges in the design of the high gain and low noise neural amplier, which will be
explained in this paper.
Although there are few reviews on neural recording microsystems and low-frequency low-noise
ampliers [14–16], a comprehensive review focused on large-scaled neural recording implants and
their challenges was not found in any prior-art publication. To help readers to get involved with the
state-of-the-art and challenges of this rapidly growing eld, led us to write this paper.
The remainder of this paper is organized as follows. Section 2 reviews the different types of
neural signals and their properties. Section 3 presents the essential neural recording architectures in
the literature. Section 4 surveys the neural ampliers. The neural ampliers are the most challenging
part of a neural recording implants. They must be compact, high gain, low power, and low noise
ampliers. To satisfy these constraints in the design of the neural ampliers, various topologies and
techniques are proposed which are presented in the subsections of Section 4. Section 5 covers the
ADCs that are suitable for neural implant applications based on their various architectures. Finally,
Section 6 discusses the data compression methods in neural recording systems and Section 7 presents
the conclusion.
2. Neural Signals
The electrical activities of the brain can be recorded through three different methods: (1) from
the scalp: its corresponding signal is electroencephalogram (EEG); (2) the surface of the brain, which
extracts an electrocorticography (ECoG) or intracranial electroencephalography (iEEG) signal; and
(3) within the brain, which captures the extracellular activities of neurons. The extracted signals from
these methods have a frequency range of a few mHz to 10 kHz and their amplitude is at the range of
20 µV to 10 mV [17].
In the rst method, surface electrodes can be used to un-invasively measure the biopotentials of
EEG on the scalp. In contrast, in the second method, the electrodes can be placed directly on the brain
surface invasively to record electrical activity from the cerebral cortex. The brain signals provided
with this method have a dramatically high signal-to-noise ratio (SNR) and are less sensitive to artifacts
than EEG. Furthermore, these signals have high spatial and temporal resolution. Capturing the signal
