Recent advances in wireless epicortical and intracortical neuronal recording systems

Abstract

An implantable brain-computer interface (BCI) has proven to be effective in the field of sensory and motor function restoration and in the treatment of neurological disorders. Using a BCI recording system, we can transform current methods of human interaction with machines and the environment, especially to help those with cognitive and mobility disabilities regain mobility and reintegrate into society. However, most reported work has focused on a simple aspect of the whole system, such as electrodes, circuits, or data transmission, and only a very small percentage of systems are wireless. A miniature, lightweight, wireless, implantable microsystem is key to realizing long-term, real-time, and stable monitoring on freely moving animals or humans in their natural conditions. Here, we summarize typical wireless recording systems, from recording electrodes, processing chips and controllers, wireless data transmission, and power supply to the system-level package for either epicortical electrocorticogram (ECoG) or intracortical local field potentials (LFPs)/spike acquisition, developed in recent years. Finally, we conclude with our vision of challenges in next-generation wireless neuronal recording systems for chronic and safe applications.

Full text

SCIENCE CHINA
Information Sciences
April 2022, Vol. 65 140401:1–140401:18
https://doi.org/10.1007/s11432-021-3373-1
c
Science China Press and Springer-Verlag GmbH Germany, part of Springer Nature 2022 info.scichina.com link.springer.com
.REVIEW .
Special Focus on Brain Machine Interfaces and Applications
Recent advances in wireless epicortical and
intracortical neuronal recording systems
Bowen JI1,2,3†, Zekai LIANG1,2†, Xichen YUAN2, Honglai XU4, Minghao WANG5,
Erwei YIN6,7, Zhejun GUO8, Longchun WANG8, Yuhao ZHOU1,2,3,
Huicheng FENG1,2, Honglong CHANG1,2* & Jingquan LIU8*
1Unmanned System Research Institute, Northwestern Polytechnical University, Xi’an 710072, China;
2Ministry of Education Key Laboratory of Micro and Nano Systems for Aerospace, School of Mechanical Engineering,
Northwestern Polytechnical University, Xi’an 710072, China;
3Collaborative Innovation Center of Northwestern Polytechnical University, Shanghai 201108,China;
4NEURACLE TECH (China)Co., Ltd., Changzhou 213100, China;
5College of Electronics and Information, Hangzhou Dianzi University, Hangzhou 310018, China;
6Defense Innovation Institute, Academy of Military Sciences (AMS), Beijing 100071, China;
7Tianjin Artificial Intelligence Innovation Center (TAIIC), Tianjin 300450, China;
8National Key Laboratory of Science and Technology on Micro/Nano Fabrication, Department of Micro/Nano Electronics,
Shanghai Jiao Tong University, Shanghai 200240, China
Received 9 August 2021/Revised 27 September 2021/Accepted 27 October 2021/Published online 10 March 2022
Abstract An implantable brain-computer interface (BCI) has proven to be effective in the field of sensory
and motor function restoration and in the treatment of neurological disorders. Using a BCI recording system,
we can transform current methods of human interaction with machines and the environment, especially to
help those with cognitive and mobility disabilities regain mobility and reintegrate into society. However,
most reported work has focused on a simple aspect of the whole system, such as electrodes, circuits, or data
transmission, and only a very small percentage of systems are wireless. A miniature, lightweight, wireless,
implantable microsystem is key to realizing long-term, real-time, and stable monitoring on freely moving
animals or humans in their natural conditions. Here, we summarize typical wireless recording systems, from
recording electrodes, processing chips and controllers, wireless data transmission, and power supply to the
system-level package for either epicortical electrocorticogram (ECoG) or intracortical local field potentials
(LFPs)/spike acquisition, developed in recent years. Finally, we conclude with our vision of challenges in
next-generation wireless neuronal recording systems for chronic and safe applications.
Keywords wireless implant, neuronal recording system, recording electrodes, processing chips, wireless
data transmission, power supply, system-level package
Citation Ji B W, Liang Z K, Yuan X C, et al. Recent advances in wireless epicortical and intracortical neuronal
recording systems. Sci China Inf Sci, 2022, 65(4): 140401, https://doi.org/10.1007/s11432-021- 3373-1
1 Introduction
The last decade has witnessed the rapid development of wireless implants for interaction with the brain.
Compared to electrical, optical, or chemical modulation, neural recording is much more challenging due to
highly complex procedures, including signal acquisition, amplification, filtering, digitization, transmission,
and decoding [1–5]. Small numbers of neurons can provide massive information through electrophysiolog-
ical recording technology and are promising for clinical applications over weeks, months, or even years.
High-quality neural signals from either the epicortical or intracortical area can be used to translate neural
activity to text [6,7], operate robotic prostheses or wheelchairs [8–11], control an exoskeleton [12], and
even restore body movement and sense of touch [13,14]. Until now, most implants are tethered to com-
mercial neural recording data acquisition systems (16 bits per sample at 30 kS/s per channel) with dozens
* Corresponding author (email: changhl@nwpu.edu.cn, jqliu@sj tu.edu.cn)
†Authors Ji B W and Liang Z K have the same contribution to this work.

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:2
to one thousand channels, such as Plexon, Blackrock, NeuroNexus, Tucker-Davis Technologies, and Neu-
ralynx. A wired table-top acquisition system is reliable and powerful, with no need to worry about weak
signal transmission and processing, power consumption, or strict packaging of implants. However, wire
connections to the system limit the range of free activity, increase the risk of infection, and constrain
subjects from recording anytime and anywhere.
Wireless neuronal recording technology is one of the most important future trends because of the
significant advantage of portable recording at any point in time and space [15]. This technology involves
two aspects: wireless data transmission outward to terminal equipment and wireless power transmission
into the neuronal recording system. For wireless data transmission, either epicortical or intracortical
neural signals contain a large amount of data, especially in terms of continuous recording time and high
channel count. Various wireless solutions have been used in animal studies in recent years, including
radio frequency (RF) [16], ZigBee [17], Bluetooth [18], and WiFi [19], owing to the rapid development of
wireless communication. For wireless power transmission, systems can either directly power the wireless
neural interface or recharge batteries through an inductive link [20–23]. However, research of wireless
neuronal recording is still in its infancy due to engineering challenges, and it therefore needs to balance
low power consumption, high bandwidth, and real-time transmission. In addition, a fully implanted
device might interfere with the wireless transmission of neural signals or affect its own lifetime in an
intracranial environment. Consequently, the demand for developing high-performance wireless recording
implants in free-behavior animals is high.
System-level integration can provide convenient operation and low-noise performance thanks to com-
pact size, short interconnections, and wireless transmission. A miniature, lightweight, wireless, and
implantable microsystem is key to realizing long-term, real-time, and stable monitoring on untethered ani-
mals in their natural conditions, including rodents [21,24], sheep [25], and non-human primates [18,26,27].
In 2021, a battery-powered wireless intracortical recording system was tested non-stop in two patients
with paralysis over a 24-h period at home by BrainGate [28], which can transmit and recognize sig-
nals generated by individual neurons and support full high-bandwidth signal transmission. BrainGate
provides the first wireless brain-computer interface (BCI) transmitter available to humans. The typical
configuration of a wireless neuronal recording system consists of five modules, as illustrated in Figure 1.
(1) Implantable electrodes for neural signal acquisition; (2) processing chips and a controller for signal
amplification, filtering, multiplexing, digitization, and controlling the parameter setting; (3) wireless data
transmission to a computer or other portable equipment for further data analysis; (4) power supply by
wireless charging or an inductive coil for system power consumption; and (5) packaging for ultracompact
size and long service life. To be ready for actual use, the trade-offs between channel count, power con-
sumption, packaging size, compatibility, and lifetime should be carefully considered. In view of the above
factors, wireless neuronal recording systems are rapidly developing toward miniaturization [24,29,30], in-
tegration [23,31,32], and intelligence [33–35], benefiting from micro-electromechanical systems (MEMS),
integrated circuits (IC), wireless transmission and powering, and machine learning algorithms. Wireless
neuronal recording systems can be not only applied in healthcare and interaction for people with severe
motor impairment but also has future potential in entertainment and military scenarios.
This paper reviews the latest progress in wireless epicortical and intracortical neuronal recording sys-
tems in five closely linked subsections, including recording electrodes, processing chips and controllers,
wireless data transmission, power supply, and packaging. Challenges facing these systems are also sum-
marized to create a direct path for future long-term and reliable recording systems according to existing
technical shortcomings.
2 Recording electrodes
Three types of electrical potentials can be recorded through intracortical neural interfaces: epicortical elec-
trocorticogram (ECoG), intracortical local field potentials (LFPs), and action potentials (spikes) [36,37].
ECoGs are recorded from the epidural or subdural area closer to the cortical surface. Thus, implantation
of ECoG electrodes only requires minimally invasive surgery, without intrusion into brain tissue [38].
As a comparison, penetrating electrodes can be used to record LFP and spikes with richer information
from neurons but with higher invasive trauma. Examples include Utah arrays, Michigan probes, and
microwires. An LFP is synchronized neuronal activity recorded from small populations of neurons by
their extracellular potentials and has higher spatial resolution than ECoGs [39]. Spikes represent a se-

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:3
RX
TX
ASIC
MCU
Wireless data
transmission
Processing chip
& controller
Power supply
Wireless charging
Package
Skin
Skull
Dura
Cortex
White
matter
ECoG
Recording electrode
Data analysis
LFP/spikes
ķĸ Ĺ
ĺ
Ļ
Recording system
Integration
Military
Healthcare
Entertainment
Interaction
Intelligence
Miniaturization
Applications
Figure 1 (Color online) Configuration, trends, and applications of wireless epicortical and intracortical neuronal recording sys-
tems.
quence of short electrical pulses in the membrane potential of individual neurons, or a small neuronal
group, compared to LFP [40]. The different properties of the latest recording electrode technologies are
compared in Table 1[32,41–49].
2.1 ECoG electrode
ECoG electrodes have been used to locate the epileptic focus and decode movements, vision, and speech.
From the size of the electrode site, ECoG electrodes fall into two basic categories: macro and micro.
Traditional macro-ECoG electrodes, with site diameters of 1–4 mm and a pitch of ∼1 cm, are commercially
available from Neuropace, PMT Corp., Medtronic, WISE Srl, among others. Softer and thinner silicone-
based macro-ECoG grids have been developed for better conformability to curved surfaces compared with
clinical macro-ECoGs [41] (Figure 2(a)).
Current trends of miniaturization have pushed spatial resolution to less than 1 mm, namely to micro-
ECoG electrodes. Latest work has mainly focused on stretchability for conformal contact to the brain
surface [50,51], transparency for optical imaging or resistance to artifacts [52,53], multiplexing for high
density [54,55], and multifunction for customizable intracranial applications [56,57]. Similar to silicone-
based macro-ECoGs, an ultrasoft and stretchable micro-ECoG grid with 50 µm×50 µm microelectrode
sites has also become a hot topic [58] (Figure 2(b)).
2.2 Utah/Utah-like arrays
In 1989, a typical penetrating electrode, namely the Utah electrode array (UEA), was proposed by Richard
Normann at the University of Utah [59]. With clinical approval by the Food and Drug Administration
(FDA) and an implant lifetime of more than seven years, it is the most widely used, gold standard
recording electrode for decoding minds and controlling external devices. A traditional UEA is composited
of 100 three-dimensional needles in a 10×10 configuration, 1.5-mm in length, with a 400 µm pitch and
4 mm×4 mm overall size. However, LFP/spikes can only be received from needle tips with set dimensions
and parameters. In 2020, Shandhi and Negi [44] proposed a multisite UEA based on the traditional Utah

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:4
Table 1 Comparison of different properties of novel recording electrode technologies
Ref.
(Year) Type Name Channels Feature sizes Materials Implantation
location
Chronic
stability
[41]
(2021)
Macro-ECoG
electrode
Soft
Grid 32
2.3 mm diameter;
5–10 mm pitch;
4 cm×10 cm coverage
PDMS substrate;
Cr/Au interconnects;
Pt nanoparticles
on electrode sites
Above the
somatosensory
and motor
cortex of minipig
2 weeks
[42]
(2020)
Micro-ECoG
electrode
Opto-E-
Dura 16
Ring shaped electro de
with an inner diameter
of 200 µm and an outer
diameter of 400 µm
PDMS substrate;
gold nanowire
(Au NW)tracks;
platinum (Pt)
particle electrodes
Above the dorsal
cortex of mouse 6 weeks
[43]
(2021)
Micro-ECoG
electrode –1152
(128ch×9)
50 µm×50 µm electrode
size; 295 µm pitch;
14 mm×7 mm coverage
Parylene-C substrate;
gold signal tracks
Above the
somatosensory
cortex of
macaque monkey
Acute
[44]
(2020)
Utah
array UMEA
75–900
(3/9 sites
per shaft)
One at the tip of shaft
and two others about
300 µm and 450 µm
below; 1.5 mm long
shaft with 400 µm pitch
Silicon substrate; TiW
and Pt as metal layer;
Paylene-C encapsulation
In vitro –
[45]
(2021)
Utah-like
array Argo 65536
(256×256)
5 to 25 µm metal
wires of diameters;
10 mm diameter
microwire array
Platinum-iridium
alloy in
microwire cores;
Parylene-C insulation
0.7–1 mm into
the rat cortex
and anditory
cortex of sheep
Acute
[46]
(2020)
Utah-like
array CHIME 100–1000
Electrodes with
22–25 µm OD
and 1–7 µm ID;
inter-wire spacing
of around 100 µm
Glass-ensheathed
gold microwires
Olfactory bulb
of mouse Acute
[47]
(2018)
Michigan
probe(stiff ) Neuropixels 960
12 µm×12 µm
electrode size;
20 µm pitch;
shank 70 µm×20 µm
cross-section with
10 mm length
TiN as electro de sites;
silicon substrate
Visual cortex,
hippocampus,
thalamus,
motor cortex,
striatum of mouse
Up to
60 days
[48]
(2021)
Michigan
probe(stiff )
Neuropixels
2.0
5120
(1280ch
per shank)
12 µm×12 µm
electrode size;
15 µm pitch;
shank 70 µm×20 µm
cross-section
with 10 mm length
TiN as electro de sites;
silicon substrate
Cortex,
hippocampus,
thalamus of mouse
Up to
309 days
[32]
(2019)
Michigan
probe
(flexible)
Neuralink’s
threads
3072
(32ch per
thread)
Probes with 32
electrode contacts
spaced by 50/75 µm
Surface treatments:
PEDOT/IrOx;
substrate, encapsulation
and dielectric of probe:
polyimide; gold thin
film trace
Unknown
implantation
location of rat
–
[49]
(2019)
Michigan
probe
(flexible)
Neurotassels 128–1024
3µm×1.5 µm
cross-section;
fibers total
diameter of ∼100 mm
Au microelectrodes
encapsulated by
polyimide; platinum
electrodeposition
Medial prefrontal
cortex of mouse
Up to
6 weeks
array, with nine sites per shaft, 900 active sites in total, and recording/stimulation capability from
different cortex layers (Figure 2(c)). In 2021, Sahasrabuddhe et al. [45] proposed a Utah-like electrode
array composed of 65536 channels of parallel platinum-iridium microwires bonded to a complementary

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:5
Utah/Utah-like array Michigan probeECoG electrode
(a)
(b) (d)
(c)
(f)
(e)
Figure 2 (Color online) Typical epicortical and intracortical recording electro des. (a) Macro-ECoG electrode array with a
diameter of 2.3 mm and large coverage of 4 cm×10 cm [41] Copyright 2021 John Wiley and Sons. (b) Micro-ECoG electrode array
with a size of 50 µm×50 µm and pitch of 200 µm. Scale bar: 1 mm [58] Copyright 2018 John Wiley and Sons. (c) Untraditional
Utah array with nine sites per shaft and 900 microelectrode sites in total [44] Copyright 2020 IEEE. (d) Argo: Utah-like array
with 65536 channels for high density recording [45] Copyright 2021 IOP Publishing. (e) Neuropixels: rigid silicon probe with 960
recording sites on a 70 µm×20 µm shank [47] Copyright 2018 Elsevier. (f) Neuralink’s threads: flexible polyimide probes with
3072 electrode sites p er array distributed across 96 threads (CC BY-ND 4.0) [32] Copyright 2019 Elon Musk, Neuralink.
metal-oxide-semiconductor (CMOS) voltage amplifier array (Figure 2(d)). This is the largest neural
recording capability to date.
2.3 Michigan probe
Another typical penetrating electrode is the Michigan probe, which is based on a rigid silicon substrate to
acquire LFP/spikes from individual neurons along the length of shanks rather than at the end (like a Utah
array). Recent work has mainly focused on high density [60,61], flexibility [49,62], and multifunction [63–
66]. Harris et al. [47] proposed multiplexed silicon probes based on CMOS technology named Neuropixels,
with a 130-nm wire width, 10-mm shank length, and 960 recording sites on a 70 µm×20 µm shank
(Figure 2(e)). They reported Neuropixels 2.0 with 5120 recording sites distributed over four shanks in
2021 [48]. To improve the mechanical match between probe and tissue, flexible probes have gradually
become a research hotspot. Musk and his company Neuralink [32] proposed high-density flexible probes
with 3072 electrode sites per array distributed across 96 threads, 4-6-µm-thick and approximately 20-mm
length, which can be inserted by a neurosurgical robot with micron precision (Figure 2(f )). Additionally,
multifunctional probes have drawn a lot of attention, enabling researchers to measure electrophysiological
or chemical signals and synchronously influence subject behaviors. Cai et al. [65] proposed a silicon probe
array with platinum nanoparticles and reduced graphene oxide nanocomposites(Pt/rGO) modification
for simultaneous real-time monitoring of dopamine (DA) concentration and neural spike firings under
deep brain electrical stimulation.
3 Processing chips, controllers, and basic circuits
Signal processing chips, controllers, and other basic circuits are essential components for the construction
of a complete wireless epicortical or intracortical neuronal recording system. They are also the key to
guaranteeing the high fidelity of neural signals and subsequent real-time transmission.
First, from the perspective of the signal-flow sequence, signal processing chips include analog signal
amplifiers, filters, multiplexers, and analog-to-digital converters (ADC) [18]. The electrical signal recorded
by an electrode is usually at a level from µV to mV. The weak signal makes it difficult to carry out
subsequent processing and is easily drowned in noise. Therefore, an amplifier is required to amplify
the signal to a higher voltage range that is easier to process. The function of the filter is to filter out
high-frequency noise and the power frequency signal and to improve the signal-to-noise ratio. In fact,

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:6
the signal of each channel needs to be amplified and filtered. Therefore, a large number of channels in
the system will lead to redundancy of the subsequent signal processing circuit and an increase in energy
consumption and equipment volume.
To address this issue, a multiplexer is employed to enable multiple signal recording channels to be
processed by one channel in a multiplexed method (e.g., 8:1, 16:1). At this point, the recorded signal is
analog, and the next steps of signal processing and wireless data transmission are carried out in the form
of digital signals; therefore, the signal should be converted to digital by an ADC. An analog signal is a
signal whose mathematical form is a continuous function in the time domain. Most signals in nature are
analog signals, such as EEG (electroence phalogram)/ECoG/spike signals and electromagnetic signals.
Digital signals are discrete signals in the time domain and in amplitude, which can easily carry out logic
or arithmetic operation through the processor and be stored in a medium. Analog processing methods
mainly include amplification, filtering, multiplexing, and signal conversion. The relevant processing
chips include power management, amplifier, filter, and signal conversion chips. As a comparison, digital
processing methods are mainly logic and arithmetic operations, with processing chips such as a central
processing unit (CPU), system on chip (SoC), microcontroller unit (MCU), and field-programmable gate
array (FPGA).
Second, as the brain of the whole system, the controller is responsible for configuring the parameters
of each module, collecting digital signals transmitted by the ADC, and controlling the radio circuit for
wireless data transmission. In addition, some systems may integrate temperature sensors, an inertial
measurement unit. The information from these sensors will be transmitted to a radio circuit or analyzed
by the controller.
Lastly, in order to keep the whole system functioning, basic circuits are also nonnegligible, such as the
power management unit (PMU), which provides a stable power supply to the system and ensures circuit
safety during charging, and the radio SoC, which conducts wireless transmission according to specific
protocols. The scheme of wireless transmission and power supply will be described in Sections 4 and 5.
Each of the units described above is generic and different wireless neural signal recording systems
have similar and corresponding functions. They could be divided into four schemes according to the
degree of system integration: (1) modular distributed [19,67–70], (2) stacked [24,26,71,72], (3) inte-
grated [32,73–79], and (4) ultracompact [80–84]. In addition, there are differences in miniaturization,
lightweight, and functional expansion between different schemes, which will be described in detail below.
3.1 Modular distributed scheme
This type of system usually uses a large number of commercial off-the-shelf (COTS) modules, which have
the advantage of quickly building a complete system with relatively low prices and little time. For exam-
ple, amplifier chips from the Intan company are used as front-end chips for signal amplification, filtering,
and analog-digital conversion [27,70,71,85]. ARM (Advanced RISC Machines) cortex architecture core
chips are used as controllers [18,19,22,68]. Commercial radio chips are used as wireless transmission
circuits, and commercial FPGAs are used as digital signal processing chips [19,22,67,71]. As shown
in Figure 3(a), the modular distributed scheme is typical [68]. The PennBMBI neural signal analyzer
(NSA) has both neuronal signal recording and analysis capabilities. NSA consists of an analog front-end,
microcontroller, power management module, commercial wireless transceiver module, and MicroSD card.
Its analog front-end has four channels, each containing two-stage amplifier circuits. The total gain of the
system can be set from 46 to 102 dB. NSA uses a 32-bit MCU (AT32UC3C1512C, Atmel company) that
integrates a 12-bit resolution ADC, multiplexer, sample hold (S/H) circuit, programmable amplifier, and
a 32-bit floating-point arithmetic digital signal processor (DSP) unit. NSA can process neural signals
online through DSP units. The volume of the system is 56 mm×36 mm×13 mm.
3.2 Stacked scheme
Building a modular distributed system with COTS components reduces the cycle and cost of system
development but simultaneously brings the problem of a large area of the printed circuit board (PCB),
making it difficult to miniaturize the system. To address this issue, some researchers have proposed a
method of dividing the PCB board into modules, connecting the modules with flexible flat cable, then
folding to reduce the area of PCB [22,71]. The 3D stack method is a representative method for reducing
system size [24,26,72,86]. Figure 3(b) shows a typical stacked system consisting of the front-end chip
(RHD2132, Intan Technologies), FPGA (Spartan-6, Xilinx company), radio chip (nRF24L01, Nordic

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:7
Modular distributed scheme Integrated scheme
Stacked scheme Ultracompact scheme
(a)
(b) (d)
(c)
in mm size
Figure 3 (Color online) Different composition schemes of a signal processing chip, controller, and basic circuit. (a) Modular
distributed scheme using a large number of COTS [68] Copyright 2015 IEEE; (b) stacked scheme in which units are connected
by flexible cables [71] Copyright 2017 IEEE; (c) typical custom application-specific integrated circuit scheme manufactured by a
CMOS process [73] Copyright 2014 IEEE; (d) ultracompact system with a volume of 2.4 mm3[82] Copyright 2016 Elsevier.
Semiconductor), PMU, and external memory [71]. All components are divided into three parts and
placed on a rigid-flex PCB. The three parts are connected by two flexible PCBs.
3.3 Integrated scheme
The stacked scheme reduces the size of the system to some extent, but it is not sufficient to realize
miniaturization and light weight, and modular systems often cannot meet the demand of a large number
of channels. In recent years, researchers have made great progress in miniaturization, high channel count,
and low power consumption using CMOS processing technology [32,73–79]. Table 2[16,23,71,72,74,77,
83,87,88] compares recent chips for implantable BCI using COMS technology. Although the latest CMOS
process level can reach 7 nm, the signal preprocessing chip is an analog application-specific integrated
circuit (ASIC) for BCI, which places more emphasis on the quality of signal processing, such as precision,
reliability, and stability. The analog ASIC has a higher operating current than the digital ASIC; thus,
the CMOS process for a 5-nm CPU cannot be used directly for preprocessing chips. However, a digital
chip that carries on the processing to the digital signals mainly focuses on the computation ability,
which needs more logic gates and arithmetic units, as well as a higher main frequency. Thus, the use
of more advanced processes is preferred whenever possible. In general, advanced processes can be of
great help in facing the challenges of high throughput, low power consumption, and miniaturization in
wireless neuronal recording systems. Elon Musk’s Neuralink company has developed a system consisting
of a custom ASIC, which integrates multiple ASICs on a PCB using a flip-chip integration to achieve
3072 channels for signal recording. As shown in Figure 3(c), this is part of an integrated system that
is manufactured using 65-nm CMOS technology (from STMicroelectronics) [73]. This chip integrates a
64-channel front-end, ADC, PMU, wireless transmission unit, bias circuitry, clock, and serial peripheral
interface (SPI). The total area of the chip is 2.4 mm×2.4 mm with a power consumption of 225 µW.
Such systems are usually customized for specific functions and integrate a variety of necessary functions
on a small chip. It greatly reduces the size and power consumption of the system and is more conducive
to implantation and wireless power supply.
3.4 Ultracompact scheme
Custom integrated systems manufactured using CMOS technology make it possible for a system to realize
a light weight and low power consumption. But some system components, such as antenna and electrode

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:8
Table 2 Comparison of different chips for implantable BCI using COMS technology
Ref. Process (nm) Area (mm2) Power supply (V) Power (µW/ch) CMRR (dB) PSRR(dB)
[16] 500 25.48 3 60 >60 >50
[87] 65 0.04 0.35/0.7 0.37 78 110
[71] 65 1.6 0.5 2.3 88 67
[72] 180 2.56 – 79 >83 –
[77] 350 1.1 1.8 <300 – –
[23] 130 12 1 11.7 – 41
[83] 130 12.19 1.2 12–98 – –
[88] 180 5 1.2 1.875 – –
[74] 350 0.115 3.3 Total 24.75 >70 >66
pads, tend to be larger in size. As shown in Figure 3(d), researchers have proposed a “Neuro dust”
scheme [80–84], a millimeter-scale system consisting of a pair of gold recording pads (0.2 mm×0.2 mm),
a custom single transistor, and a piezoelectric crystal [82]. The piezocrystal, which is impacted by
ultrasound pulses, converts the ultrasonic energy into electrical energy to supply power to the system.
Therefore, the neural dust does not need a built-in battery or a larger coil, and its total volume is only
2.4 mm3. In summary, the ultracompact scheme has fewer channels but can be used for high spatial
resolution and independent recording of multiple discrete points through multiple systems. It also has
greater advantages in miniaturization. In brief, it provides a new solution for building wireless neural
recording systems.
3.5 Some commercial products
Several commercial corporations have been working on wireless implants in recent years, mainly because of
the integration of high-performance chips, and are welcomed by a fast-growing number of users. Blackrock
Microsystems has launched the two latest wireless recording systems, the CerePlex Exilis and CerePlex
W. CerePlex Exilis supports 32/64/96 channels, 30 kSps sampling frequency, and 16-bit resolution during
signal acquisition. The input signal frequency range is 0.3 Hz to 7.5 kHz, the equivalent input noise is less
than 3 µVrms, the subject animals can freely move within a 1-m range, and the maximum continuous
operating time is 2.5 h while powered by a battery. In comparison, the CerePlex W has the same
sampling frequency, resolution, input signal frequency range, and equivalent input noise but uses the
Honey Badger ASIC Chip for longer continuous operation, up to 3.5 h, and a larger maximum free
movement range of 2 m. The CerePlex Exilis (9.87 g) is lighter than the CerePlex W (33.5 g). Compared
with current academic results, the main advantages of Blackrock’s products are their good recorded-signal
quality, higher sampling frequency, resolution, accuracy, and transmission rate. On the other hand, their
disadvantages are higher power consumption and shorter continuous operation time without a wireless
power supply.
The Plexon DataLogger wireless data recording system can support 32-channel signal acquisition with
a 40-kHz sampling rate and 16-bit sampling resolution. However, it does not support real-time neural
signal data transmission, which means the data needs to be downloaded after recording. The continuous
operation time is also short (up to 45 min).
Bio-signal technologies has also proposed a wireless electrophysiological recording system, Eros, which
can support 4/8 channels, 250-Hz sampling frequency, 24-bit sampling accuracy, 4–6 h battery life, and
a 10-m working distance. Compared with other products, it has fewer channels and a lower sampling
frequency.
4 Wireless data transmission
Compared with the wired system, a significant difficulty to cope with is the elimination of cables for
wireless communication, which includes uplink and downlink communications [89]. The uplink communi-
cation can transmit digitalized neural data to an external base station for status monitoring and further
analysis. Inevitably, acquired data from hundreds of recording electrodes can quickly come to hundreds
of megabits per second (Mbps), so uplink communication requires a higher data transmission rate. At
the same time, this leads to higher power dissipation. In comparison, downlink data communication
from the external station to the neuronal recording system results in much lower power dissipation, at

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:9
Table 3 Comparison of different wireless data transmission schemes
Ref. Number of
channels
Wireless data
transfer type Data rate Transmission
distance
Wireless transmission
power consumption Year
[92] 46 Infrared light
880–900 nm 4 Mbps – – 2012
[16,26] 100 FSK 3.2–3.8 GHz 24 Mbps >1 m 50 mW 2013
[93] 100 UWB 3.1–5 GHz 200 Mbps 1–2 m – 2014
[94] 64 UHF MICS band
402–405 MHz 450 kbps 2 m 19.8 mW 2015
[86] 64
2.4 GHz with
802.11 b/g/n
protocol
1.536 Mbps 2 m – 2016
[18] 128 2.4 GHz BLE 1.96 Mbps 2 m 11.8 mA (+4 dBm) 2019
[24] 8/36/72 RF 2.4 GHz 2 Mbps Max 50 m – 2020
both the transmitter and receiver sides, because it only needs to transmit configuration information and
control instructions, such as the active number of electrodes and the sampling rate of the ADC. The
ECoG signal frequency is less than 200 Hz for the human brain, the spikes are in the range of 0.1–7 kHz,
and the LFP is also smaller than 200 Hz [90]. According to the Nyquist-Shannon sampling theorem [91],
when a signal with frequency fis sampled, the sampling frequency must ideally be no less than 2fto
restore the sampled signal without distortion. Therefore, the sampling frequency of AD conversion for
epicortical ECoG signals (∼1 kHz) is much lower than that for intracortical spikes (∼20 kHz), which
results in different requirements for wireless data transmission and rate. In simple terms, ECoG signals
of 100 channels are recorded with an ADC for analog-to-digital conversion. The sampling frequency is
1 kSps and the resolution is 12-bit. The size of the serial data stream generated is 100 ch×1000 Sps/ch
×12 bit/S = 1.2 Mbps. Therefore, the wireless transmission rate should be at least greater than 1.2 Mbps.
This calculation method is also suitable for the transmission rates of spikes.
Typical wireless data transmission schemes from the last decade are compared by channel count,
wireless data transfer type, data rate, transmission distance, and wireless transmission power consumption
in Table 3[16,18,24,26,86,92–94].
According to the different physical forms, the wireless transmission mode adopted by a wireless neuronal
recording system can be divided into RF [26,67,68,73,85,94,95], infrared (IR) [26,92,96], and ultrasonic
communication [79]. RF communication has been widely used because of the little influence of occlusion
and the difficulty of technical implementation. With the development of RF communication technology,
it is easy for signal crosstalk to occur when using the same frequency band for communication. Therefore,
organizations and technical alliances such as IEEE made a division of the frequency band. Electronic
devices for short-range wireless communication usually work in the industrial scientific medical (ISM)
band. In addition, some standard communication protocols, such as Bluetooth [23,31,70], WiFi [19,86,97],
and ZigBee [17], have been gradually developed. These standard protocols set strict specifications at
the physical and link layers, so the cost and technical difficulty of implementation are low, and COTS
components can be easily used. In 2019, Zhou et al. [18] used 2.4-GHz low-power Bluetooth technology to
achieve wireless communication of the intracortical LFP recording system and base station in non-human
primates (Figure 4(a)). The commercial Bluetooth Radio SoC (nRF51822, Nordic Semiconductor) was
used for the customization of Bluetooth protocol to achieve a 2 Mbps modulation rate within a 2-m
distance.
However, because Bluetooth and WiFi are widely used in electronic devices, they are very prone to
signal interference when applied to implantable medical devices. Customized wireless communication
technology can effectively avoid this problem. In 2014, Mestais et al. [94] realized wireless data com-
munication between an epicortical ECoG recording system and PC using ultrahigh-frequency (UHF)
transmission technology in non-human primates (Figure 4(b)). The UHF link works in the medical im-
plant communication service (MICS) band using a custom protocol in the 400 MHz range. The low-level
communication protocol is realized by a radio chip (ZL70102, MicroSemi company) with an energy per
bit of less than 21 nJ and a maximum payload throughput of up to 500 kbps. Finally, a transmission
rate of up to 450 kbps is achieved by a custom platinum antenna at a maximum distance of 2 m.
The use of infrared communication can also reduce interference between different devices. In 2009,
Song et al. [96] adopted RF function multiplexing technology in a 16-channel broadband spike recording

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:10
RFRF & IR
UHFBLE
UHF antennae
(c) (d)
(a) (b)
Figure 4 (Color online) Different wireless data transmission methods. (a) Bluetooth low energy (BLE) communication with
radio SoC and a custom antenna [18] Copyright 2019 Springer Nature; (b) UHF radio frequency communication operating in the
MICS band [94] Copyright 2014 IEEE; (c) RF transmission function multiplexing and IR transmission [96] Copyright 2009 IEEE;
(d) RF for simultaneous power supply and data transmission [73] Copyright 2014 IEEE.
system in a nonhuman primate (Figure 4(c)). The 13.56-MHz RF signal is sent to a gold spiral on the
back of the epicranial section to supply power to the system. At the same time, the controller modulates
the signal to obtain clock signals and control instruction information. The controller converts the multi-
channel neural signal data into the driving current, drives the IR signal of a specific frequency, and
transmits data to the outside.
In addition, backscattering modulator technology can simultaneously carry out wireless communication
and supply power, avoiding the use of two antennas or coils, respectively, reducing the volume and power
consumption of the system. In 2014, Muller et al. [73] used backscattering modulator technology to
transmit serialized data at a rate of 1 Mbps. In this way, a single antenna can simultaneously power and
transmit data wirelessly to the system (Figure 4(d)).
In fact, modulation technology is the key to realizing wireless communication. To put it simply, the
function of modulation is to convert low-frequency baseband signals containing information into high-
frequency carrier signals. By modulating to different frequencies, interference of different signals can be
effectively avoided in communication. Modulation can be divided into analog modulation and digital
modulation according to different modulation signals. Compared with analog modulation, digital modu-
lation has stronger anti-interference capability and confidentiality and is convenient for device integration
and signal processing by computer. Digital modulation can be divided into amplitude shift keying (ASK),
phase shift keying (PSK), and frequency shift keying (FSK) according to the different high-frequency car-
rier parameters of digital signals. ASK is more widely used due to its technical difficulty and low energy
consumption. For binary ASK modulation technology, the digital signal “1” sends a carrier signal with a
larger amplitude, and “0” sends a carrier signal with a smaller amplitude. When the digital signal is “0”,
no carrier signal is sent so as to further reduce energy consumption, which is the simplified ASK mod-
ulation technology, and is known as on-off keying (OOK) modulation technology [98]. Therefore, OOK
modulation technology is very suitable for the wireless communication signal modulation of low power
electronic devices. In 2014, Yin et al. [93] adopted OOK modulation technology to achieve a maximum
transmission rate of 200 Mbps with low power consumption and a DC/RF conversion efficiency of 40%.
5 Power supply
Wired systems are generally powered by cables without energy shortage, but the power supply is a signif-
icant challenge for continuous monitoring wireless systems. A large number of electrodes, amplification,

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:11
Table 4 Comparison of different wireless power transmission schemes
Ref. Power supply mode Power consumption Running time Year
[26]200 mAh battery with
2 MHz wireless charging 90.6 mW 7 h 2013
[93]1.2 Ahr 3.6 V
disposable battery
17 mA(low-RF) or
27 mA(high-RF) 48 h 2014
[94]13.5 MHz
inductive link
75 mW (32 ch@1 kHz,
w/o charging)
350 mW w/ charging
Unlimited 2015
[73] 300 MHz RF Power 2.3 µW/ch total:225 µW Unlimited 2015
[85]Rechargeable 3.7 V battery
with ultrasonic charging 4.22–15.4 mA – 2016
[22]
1.6 Wh battery with
265 kHz magnetic
resonance coupling
200 mW(w/o charging)
400 mW(w/ charging) 8 h 2018
[18] 500 mAh Lithium-ion battery 172 mW 11.3 h 2019
[99]30 mAh 3.7 V
rechargeable battery 28.6 mW 2.5 h 2021
digitalization, and wireless transmission lead to great energy needs [89]. There are several methods to
power up wireless systems. Power consumption and running times of typical wireless power transmission
schemes are compared in power supply mode in Table 4[18,22,26,73,85,93,94,99].
5.1 Battery
The traditional solution is to use built-in batteries to power the system, as in pacemakers and cochlear
implants [93,98–100]. Batteries have been used in medical applications since 1973 [101]. Figure 5(a)
demonstrates a battery-powered solution. In 2014, Yin et al. [93] used a disposable AA Li-ion battery
(1.2 Ah, Saft Groupe S.A.) to power a system. The system integrates three customized low-power ASICs
so the battery can sustain power for more than 48 h.
5.2 Wireless charging
A battery has a limited life and needs to be replaced when the power runs out. Therefore, such a system
cannot be fully implanted or requires battery replacement by surgery, which often causes unnecessary
damage to biological tissue. To avoid the biological impact of battery replacement, some researchers have
proposed charging batteries wirelessly [22,26,72,85,102]. Figure 5(b) shows a typical wireless charging
scheme for a built-in battery. In 2018, Matsushita et al. [22] proposed a wireless neuronal recording
system, named W-HERBS, that uses a lithium-ion polymer rechargeable battery as the power supply.
Meanwhile, the magnetic resonance coupling method is used to charge the battery, avoiding the need for
battery replacement. During wireless charging, the external transmitter unit uses the 265-kHz magnetic
frequency to carry out magnetic coupling with the receiver unit. It can transmit 400 mW of electrical
energy, meeting the energy consumption of the system (200 mW) and battery charging (200 mW) at the
same time.
5.3 Wireless power supply
The battery power scheme solves the power supply problem for wireless systems, but batteries often oc-
cupy a large proportion of the volume and weight, which hinder the miniaturization and light weight of a
system. Meanwhile, a battery risks electrolyte leakage, which may cause serious harm to organisms dur-
ing implantation. Fortunately, a wireless power supply is a good solution to the above issues. According
to different forms of energy transfer, wireless power supply schemes can be divided into two types: elec-
tromagnetic energy transfer [67,73,92,94–96,103] and other energy transfer forms, such as infrared [104]
and ultrasound [82]. Figure 5(c) illustrates a wireless charging scheme based on inductive coupling, which
was proposed by Mestais et al. [94] in 2014. The system obtains 100 mW of power from the external
antenna through the 13.56-MHz induction link. It is sufficient to supply the system with an overall energy
consumption of 75 mW.

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:12
Wireless RF power supply Ultrasonic power supply
System
Battery Wireless charging
System
System
(d)
(c)
(b)(a)
Figure 5 (Color online) Different power supply modes. (a) Powered directly by disposable or rechargeable battery [93] Copy-
right 2014 Elsevier; (b) powered by rechargeable battery with wireless charging (CC BY) [22] Copyright 2018 Matsushita et al.;
(c) wireless power supply by RF [94] Copyright 2014 IEEE; (d) wireless power supply by ultrasound [82] Copyright 2016 Elsevier.
The influence of a wireless power supply on signal recording cannot be ignored. On the one hand,
the use of near-field magnetic coupling or an RF wireless power supply with RF wireless communication
will cause interference of communication signals and affect the quality of transmitted signals. On the
other hand, electromagnetic induction can also induce current on the metal electrodes, which can affect
the quality of the original neural signals to a certain extent. For the first issue, some researchers have
tried to use different working frequency bands. For example, the power supply frequency of magnetic
resonance coupling is set to 265 kHz, and the RF wireless communication frequency is set to 2.4 GHz [22].
Another work sets the high-frequency wireless power supply frequency to 13.56 MHz, the RF wireless
communication frequency to 402–405 MHz as the MICS band, and a filter is added to filter out the
electromagnetic wave of the corresponding frequency [94]. For the second issue, not only the signals
of wireless power supply but also the signals of wireless communication and environmental noise will
affect the original neural signals. Therefore, researchers usually filter out interfering signals during signal
pretreatment or reduce the influence on signal detection by adding metal shells [16]. In addition, other
forms of energy conversion, such as photovoltaic or ultrasonic wireless power supply, may be applied to
solve this problem.
Meanwhile, compared with near-field (NF) coupling, far-field (FF) radiation, or RF transmission,
ultrasound (US) is a superior method for powering a recording system for deeper implantation. US
utilizes sound waves as the carrier of energy. Figure 5(d) shows a charging scheme of the neural dust
mentioned above [82]. The external transducer sends US pulses to a piezoelectric crystal, which converts
the mechanical energy of the US pulse into electrical energy to supply power to the system.
In general, a battery can provide more power than a wireless power supply because high-power wireless
charging must consider heat generation and the electromagnetic intensity range acceptable to the human
body. Therefore, the power supply scheme selected needs to be comprehensively considered among
system power consumption, channel number, signal quality, wireless transmission scheme, and total
system volume.
6 Package
The above sections have summarized approaches to wireless neuronal recording systems from electrodes,
chips, data transmission to the power supply. At this point, the system is fully operational, but the
packaging of the system is nonnegligible to guarantee its life-lasting stability in vivo. There are two main
considerations in packaging. First, the system should meet strict requirements of water vapor tract ratio
(WVTR) and oxygen tract ratio (OTR) to avoid a short circuit via a good water and gas barrier [105–109].

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:13
Second, the packaging housing should employ biocompatible materials to prevent rejection by biological
tissue [110–113]. In 2012, Schuettler et al. [92] proposed a wireless epicortical neuronal recording sys-
tem named BrainCon using alumina ceramic as the hermetic package material to protect the core chips
and circuits from water-induced corrosion, as shown in Figure 6(a). Meanwhile, polydimethylsiloxane
(PDMS) and Parylene-C (optionally) were coated on the outside packaging. The package not only passed
the gross leak test, but fine leak tests on a mass-spectrometer based leakage tester (SmartTest HLT570,
Pfeiffer Vacuum, Asslar, Germany) also proved good hermeticity. Practically, the package works well in
animal experiments lasting for about one year. In 2014, Mestais et al. [94] reported another wireless epi-
cortical neuronal recording system with 64 channels named WIMAGINE, which used dedicated titanium
packaging with customized hermetic feedthroughs to guarantee all electronics were in hermetic housing,
as pictured in Figure 6(b). The hermetic feedthroughs are based on a ruby insulator, with additional
gold brazing for hermeticity. All components were tested individually by helium leakage testing with
10−9bar·cm3·s−1. Additionally, thermally conductive silicone pads were laid between the top electronic
PCB and titanium housing to avoid biological tissue damage by system heat. Moreover, the PCB stacking
is optimized using conductive glue and heat sinks to further weaken the thermal impact. Finally, the
overall size of WIMAGINE is 50 mm in diameter and 12.54 mm in height with a 10 cm2HF Antenna
and was implanted in nonhuman primates for ECoG recording, which lasted for 26 weeks with good
biocompatibility.
When infrared signals are used for communication, the transparency of the system housing should be
enhanced to reduce signal loss. When applying RF technology for communication or power supply, it
is also necessary to avoid energy loss caused by metal housing and the impact of induced current on
organism safety. In 2013, Yin et al. proposed a wireless intracortical neuronal recording system linked to
a Utah array with sealing and packaging by laser welding of two titanium metal cases (56 mm×42 mm
×9 mm, 30.6 g) [16,26], as shown in Figure 6(c). They assembled a single-crystal sapphire window with
a diameter of 29.2 mm by copper welding, achieving 93% IR transparency at a 5-mm distance from the
window and <50% RF loss for the radio transmitter. In 2019, Musk and his company Neuralink [32]
released an intracortical neuronal recording system linked to flexible probes and packaged in titanium
cases, as pictured in Figure 6(d). It was coated with Parylene-C as a moisture barrier to prevent fluid
inflow and prolong its lifetime. Two configurations of the recording system have been proposed, with
1536 and 3072 channels, and their corresponding volumes and weights are 24.5 mm×20 mm×1.65 mm
and 11 g and 23 mm×18.5 mm×2 mm and 15 g, respectively. They upgraded this system to LINK V0.9
in August 2020, a coin-size wireless charging version with 1024 recording channels, all-day battery life,
and a size of 23 mm×8 mm [114].
7 Conclusion and challenges
This review summarizes the latest development in wireless implantable neuronal acquisition systems,
including high-density neural recording electrodes, dedicated low-power ASICs, wireless data transmission
modes, wireless power supply modes, and highly reliable hermetic packages. The above units in the BCI
are important factors that determine communication efficiency and human-computer interaction ability.
The system-level wireless neuronal acquisition will surely become one of the most important development
trends in the future and will get rid of wired constraints and enhance independence and mobility for
human use.
The human brain has more than 80 billion neurons; however, current systems are far from sufficient to
handle such a large number. The development of wireless implantable neuronal recording systems still
faces many challenges as described below.
(1) High throughput. At the World Artificial Intelligence Conference 2021, Tiger H. Tao and
colleagues from Shanghai Microsystems Research Institute, China, released their latest work on high-
throughput, minimally-invasive, flexible probes with 2640 channels on a single device. Compared with
the increase of the number of electrodes, the greater challenges are how to firmly connect the electrodes
to the circuit when the number of electrodes reaches more than 10000, how to carry out high-fidelity
amplification, filtering, and analog-to-digital conversion of neural signals with high channel count, and
how to process high-throughput data algorithmically. In addition, consequent problems with wireless
transmission, power supply, and heat also need to be synthetically solved.
(2) Wireless transmission. For a wired recording system, the Argo with 65536 channels produces

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:14
NeuralinkBrown University
WIMAGINEBrainCon
(a)
(c)
(b)
(d)
hermetic package
RF coil
cortical electrode array
cable
ASIC
threads
Titanium
enclosure
Digital
USB -C
connector
5 mm
Figure 6 (Color online) Representative system-level packages for neuronal recording. (a) BrainCon: wireless epicortical neuronal
recording system packaged by alumina ceramic and coated with PDMS and Parylene-C [92] Copyright 2012 IEEE; (b) WIMAGINE:
wireless epicortical neuronal recording system packaged by hermetic titanium housing [94] Copyright 2014 IEEE; (c) wireless
intracortical neuronal recording system packaged by titanium cases (lid removed) and coated with Parylene-C [16] Copyright 2013
IEEE; (d) wireless intracortical neuronal recording system packaged by titanium casing with single-crystal sapphire window (CC
BY-ND 4.0) [32] Copyright 2019 Elon Musk, Neuralink.
a readout data rate of up to 26 Gbps, as previously introduced. In comparison, the wireless recording
system (LINK V0.9) by Neuralink can only support the wireless transmission of compressed data at a spike
firing by Bluetooth. Therefore, wireless transmission of high-throughput data remains a ma jor challenge
since commonly used RF technologies are often limited by power consumption and heat dissipation.
(3) Power consumption and supply. Power budgets often limit the number of recording channels
because signal processing and wireless data transmission consume large amounts of power. At the same
time, the heat from electromagnetic induction also limits the power of wireless energy transmission.
Taking Neuralink’s 3072 channel wired recording system as an example, its system power consumption
is 750 mW, and its latest wireless 1024 channel LINK V0.9 can work all day with a battery life of 24 h
(wireless charging) and with much lower power consumption at the expense of the complete waveform
of spikes. As for the power supply, the 128-channel W-HERBS system can reach a maximum charging
power of 400 mW, as mentioned earlier. However, for wireless recording systems with more than 10000
channels, the wireless charging power is far from sufficient under the premise of biosecurity.
(4) Heat accumulation. Due to full implantation in the intracranial space, a packaged system
needs to consider heat-induced temperature rise, which may lead to thermal damage to the brain tissue.
According to the ISO 14708-1:2000 E standard, the temperature rise should be constrained to 2◦C on
the external surface of the implanted device. Therefore, heating limits the wireless transmission rate and
power of the system. As previously mentioned, thermally conductive silicone pads, conductive glue, heat
sinks, and titanium housing are used in the WIMAGINE system for optimized thermal management.
Nevertheless, heat dissipation technology needs further development for wireless recording systems with
high-throughput wireless transmission and high-power supply.
(5) Minimization. Components such as electrode pads, processing and computing circuits, wireless
power supply coils, wireless data transmission antennas, built-in batteries, and heat dissipation units limit
the overall size of the system to be smaller. As mentioned earlier, the overall size of Neuralink’s LINK
V0.9 has been limited to 23 mm×8 mm as a highly integrated solution. In the future, the miniaturization
of a wireless recording system with more than 10000 channels needs further development in regard to
structure optimization and highly integrated system components.
(6) Lifetime. Until now, the longest implantation lifetime of a wired Utah array recording system
ranged from six (BrainGate project) to nine years (Nicho Natsopoulos, University of Chicago) for non-

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:15
human primates. Meanwhile, these systems have a lifespan of up to seven years for humans. However, the
lifetime of existing wireless neuronal recording systems is affected by many factors, and their reliability
for chronic implantation life has been insufficiently verified.
Overall, the wireless implantable neuronal recording system is an emerging research field with strong
comprehensiveness and interdisciplinary integration. It involves micro-nano manufacturing, microelec-
tronics, communication, energy, biomedicine, brain science, and artificial intelligence. With the devel-
opment of the above technologies, wireless implantable BCI systems will become more miniaturized,
integrated, and intelligent for broader applications in humans in the next decade.
Acknowledgements This work was supported by China Postdoctoral Science Foundation (Grant Nos. 2020TQ0246, 2021M692-
638), Shanghai Sailing Program (Grant No. 21YF1451000), Fundamental Research Funds for the Central Universities (Grant No.
31020200QD013), and Natural Science Foundation of Chongqing (Grant No. cstc2021jcyj-msxmX0825).
References
1 Homer M L, Nurmikko A V, Donoghue J P, et al. Sensors and decoding for intracortical brain computer interfaces. Annu
Rev Biomed Eng, 2013, 15: 383–405
2 Brandman D M, Cash S S, Hochberg L R. Human intracortical recording and neural decoding for brain-computer interfaces.
IEEE Trans Neural Syst Rehabil Eng, 2017, 25: 1687–1696
3 Szostak K M, Grand L, Constandinou T G. Neural interfaces for intracortical recording: requirements, fabrication methods,
and characteristics. Front Neurosci, 2017, 11: 665
4 Miller K J, Hermes D, Staff N P. The current state of electrocorticography-based brain-computer interfaces. NeuroSurg
Focus, 2020, 49: 2
5 Sharma K, Sharma R. Design considerations for effective neural signal sensing and amplification: a review. Biomed Phys
Eng Express, 2019, 5: 042001
6 Vansteensel M J, Pels E G M, Bleichner M G, et al. Fully implanted brain-computer interface in a locked-in patient with
ALS. N Engl J Med, 2016, 375: 2060–2066
7 Moses D A, Metzger S L, Liu J R, et al. Neuroprosthesis for decoding speech in a paralyzed person with anarthria. N Engl
J Med, 2021, 385: 217–227
8 Hochberg L R, Bacher D, Jarosiewicz B, et al. Reach and grasp by people with tetraplegia using a neurally controlled rob otic
arm. Nature, 2012, 485: 372–375
9 Flesher S N, Downey J E, Weiss J M, et al. A brain-computer interface that evokes tactile sensations improves robotic arm
control. Science, 2021, 372: 831–836
10 Rajangam S, Tseng P H, Yin A, et al. Wireless cortical brain-machine interface for whole-body navigation in primates. Sci
Rep-Uk, 2016, 6: 1–13
11 Libedinsky C, So R, Xu Z M, et al. Independent mobility achieved through a wireless brain-machine interface. PLoS ONE,
2016, 11: 0165773
12 Benabid A L, Costecalde T, Eliseyev A, et al. An exoskeleton controlled by an epidural wireless brain-machine interface in
a tetraplegic patient: a pro of-of-concept demonstration. Lancet Neurol, 2019, 18: 1112–1122
13 Bouton C E, Shaikhouni A, Annetta N V, et al. Restoring cortical control of functional movement in a human with
quadriplegia. Nature, 2016, 533: 247–250
14 Ganzer P D, Colachis S C, Schwemmer M A, et al. Restoring the sense of touch using a sensorimotor demultiplexing neural
interface. Cell, 2020, 181: 763–773
15 Maharbiz M M, Muller R, Alon E, et al. Reliable next-generation cortical interfaces for chronic brain-machine interfaces
and neuroscience. Proc IEEE, 2017, 105: 73–82
16 Yin M, Borton D A, Aceros J, et al. A 100-channel hermetically sealed implantable device for chronic wireless neurosensing
applications. IEEE Trans Biomed Circ Syst, 2013, 7: 115–128
17 Young C P, Liang S F, Chang D W, et al. A portable wireless online closed-loop seizure controller in freely moving rats.
IEEE Trans Instrum Meas, 2011, 60: 513–521
18 Zhou A, Santacruz S R, Johnson B C, et al. A wireless and artefact-free 128-channel neuromodulation device for closed-loop
stimulation and recording in non-human primates. Nat Biomed Eng, 2019, 3: 15–26
19 Fernandez-Leon J A, Parajuli A, Franklin R, et al. A wireless transmission neural interface system for unconstrained non-
human primates. J Neural Eng, 2015, 12: 056005
20 Wentz C T, Bernstein J G, Monahan P, et al. A wirelessly powered and controlled device for optical neural control of
freely-behaving animals. J Neural Eng, 2011, 8: 046021
21 Chang C W, Chiou J C. A wireless and batteryless microsystem with implantable grid electrode/3-dimensional probe array
for ECoG and extracellular neural recording in rats. Sensors, 2013, 13: 4624–4639
22 Matsushita K, Hirata M, Suzuki T, et al. A fully implantable wireless ECoG 128-channel recording device for human
brain-machine interfaces: W-HERBS. Front Neurosci, 2018, 12: 511
23 Lee B, Jia Y, Mirb ozorgi S A, et al. An inductively-powered wireless neural recording and stimulation system for freely-
behaving animals. IEEE Trans Biomed Circ Syst, 2019, 13: 413–424
24 Keramatzadeh K, Kiakojouri A, Nahvi M S, et al. Wireless, miniaturized, semi-implantable electrocorticography microsystem
validated in vivo. Sci Rep-Uk, 2020, 10: 1–13
25 Sauter-Starace F, Ratel D, Cretallaz C, et al. Long-term sheep implantation of WIMAGINE, a wireless 64-channel electro-
corticogram recorder. Front Neurosci, 2019, 13: 847
26 Borton D A, Yin M, Aceros J, et al. An implantable wireless neural interface for recording cortical circuit dynamics in
moving primates. J Neural Eng, 2013, 10: 026010
27 Schwarz D A, Lebedev M A, Hanson T L, et al. Chronic, wireless recordings of large-scale brain activity in freely moving
rhesus monkeys. Nat Methods, 2014, 11: 670–676
28 Simeral J D, Hosman T, Saab J, et al. Home use of a percutaneous wireless intracortical brain-computer interface by
individuals with tetraplegia. IEEE Trans Biomed Eng, 2021, 68: 2313–2325
29 Seo D, Carmena J M, Rabaey J M, et al. Model validation of untethered, ultrasonic neural dust motes for cortical recording.
J Neurosci Methods, 2015, 244: 114–122

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:16
30 Park S Y, Kyounghwan N, Voroslakos M, et al. A miniaturized 256-channel neural recording interface with area-
efficient hybrid integration of flexible probes and CMOS integrated circuits. IEEE Tran Bio-Med Eng, 2021. doi:
10.1109/TBME.2021.3093542
31 Liu X L, Zhang M L, Xiong T, et al. A fully integrated wireless compressed sensing neural signal acquisition system for
chronic recording and brain machine interface. IEEE Trans Biomed Circ Syst, 2016, 10: 874–883
32 Musk E. An integrated brain-machine interface platform with thousands of channels. J Med Int Res, 2019, 21: 16194
33 Willett F R, Avansino D T, Hochberg L R, et al. High-performance brain-to-text communication via handwriting. Nature,
2021, 593: 249–254
34 Silversmith D B, Abiri R, Hardy N F, et al. Plug-and-play control of a brain-computer interface through neural map
stabilization. Nat Biotechnol, 2021, 39: 326–335
35 Makin J G, Moses D A, Chang E F. Machine translation of cortical activity to text with an encoder-decoder framework.
Nat Neurosci, 2020, 23: 575–582
36 Sung C, Jeon W, Nam K S, et al. Multimaterial and multifunctional neural interfaces: from surface-type and implantable
electrodes to fiber-based devices. J Mater Chem B, 2020, 8: 6624–6666
37 Zhou Y H, Ji B W, Wang M H, et al. Implantable thin film devices as brain-computer interfaces: recent advances in design
and fabrication approaches. Coatings, 2021, 11: 204
38 Ha S, Akinin A, Park J, et al. Silicon-integrated high-density electrocortical interfaces. Proc IEEE, 2017, 105: 11–33
39 Xing D J, Yeh C I, Shapley R M. Spatial spread of the local field potential and its laminar variation in visual cortex. J
Neurosci, 2009, 29: 11540–11549
40 Buzs´aki G, Anastassiou C A, Koch C. The origin of extracellular fields and currents-EEG, ECoG, LFP and spikes. Nat Rev
Neurosci, 2012, 13: 407–420
41 Fallegger F, Schiavone G, Pirondini E, et al. MRI-compatible and conformal electro corticography grids for translational
research. Adv Sci, 2021, 8: 2003761
42 Renz A F, Lee J, Tybrandt K, et al. Opto-E-Dura: a soft, stretchable ECoG array for multimodal, multiscale neuroscience.
Adv Healthc Mater, 2020, 9: 2000814
43 Kaiju T, Inoue M, Hirata M, et al. High-density mapping of primate digit representations with a 1152-channel µECoG array.
J Neural Eng, 2021, 18: 036025
44 Shandhi M M H, Negi S. Fabrication of out-of-plane high channel density microelectrode neural array with 3D recording and
stimulation capabilities. J Microelectromech Syst, 2020, 29: 522–531
45 Sahasrabuddhe K, Khan A A, Singh A P, et al. The Argo: a high channel count recording system for neural recording in
vivo. J Neural Eng, 2020, 18: 015002
46 Kollo M, Racz R, Hanna M E, et al. CHIME: CMOS-hosted in vivo microelectrodes for massively scalable neuronal recordings.
Front Neurosci, 2020, 14: 834
47 Steinmetz N A, Koch C, Harris K D, et al. Challenges and opportunities for large-scale electrophysiology with Neuropixels
probes. Curr Opin NeuroBiol, 2018, 50: 92–100
48 Steinmetz N A, Aydin C, Lebedeva A, et al. Neuropixels 2.0: a miniaturized high-density probe for stable, long-term brain
recordings. Science, 2021, 372: eabf4588
49 Guan S, Wang J, Gu X W, et al. Elastocapillary self-assembled neurotassels for stable neural activity recordings. Sci Adv,
2019, 5: 2842
50 Ji B W, Ge C F, Guo Z J, et al. Flexible and stretchable opto-electric neural interface for low-noise electro corticogram
recordings and neuromo dulation in vivo. Biosens Bioelectron, 2020, 153: 112009
51 Dong R H, Wang L L, Hang C, et al. Printed stretchable liquid metal electrode arrays for in vivo neural recording. Small,
2021, 17: 2006612
52 Seo J W, Kim K, Seo K W, et al. Artifact-free 2D mapping of neural activity in vivo through transparent gold nanonetwork
array. Adv Funct Mater, 2020, 30: 2000896
53 Qiang Y, Artoni P, Seo K J, et al. Transparent arrays of bilayer-nanomesh microelectrodes for simultaneous electrophysiology
and two-photon imaging in the brain. Sci Adv, 2018, 4: 0626
54 Viventi J, Kim D H, Vigeland L, et al. Flexible, foldable, actively multiplexed, high-density electrode array for mapping
brain activity in vivo. Nat Neurosci, 2011, 14: 1599–1605
55 Schaefer N, Garcia-Cortadella R, Mart´ınez-Aguilar J, et al. Multiplexed neural sensor array of graphene solution-gated
field-effect transistors. 2D Mater, 2020, 7: 025046
56 Shi Z F, Zheng F M, Zhou Z T, et al. Silk-enabled conformal multifunctional bioelectronics for investigation of spatiotemporal
epileptiform activities and multimo dal neural encoding/decoding. Adv Sci, 2019, 6: 1801617
57 Ji B W, Guo Z J, Wang M H, et al. Flexible polyimide-based hybrid opto-electric neural interface with 16 channels of
micro-LEDs and electro des. Microsyst Nanoeng, 2018, 4: 1–11
58 Tybrandt K, Khodagholy D, Dielacher B, et al. High-density stretchable electrode grids for chronic neural recording. Adv
Mater, 2018, 30: 1706520
59 Campbell P K, Jones K E, Huber R J, et al. A silicon-based, three-dimensional neural interface: manufacturing processes
for an intracortical electro de array. IEEE Trans Biomed Eng, 1991, 38: 758–768
60 Shobe J L, Claar L D, Parhami S, et al. Brain activity mapping at multiple scales with silicon microprobes containing 1024
electrodes. J NeuroPhysiol, 2015, 114: 2043–2052
61 Jun J J, Steinmetz N A, Siegle J H, et al. Fully integrated silicon probes for high-density recording of neural activity. Nature,
2017, 551: 232–236
62 Wei X L, Luan L, Zhao Z T, et al. Nanofabricated ultraflexible electrode arrays for high-density intracortical recording. Adv
Sci, 2018, 5: 1700625
63 Shin H, Son Y, Chae U, et al. Multifunctional multi-shank neural probe for investigating and modulating long-range neural
circuits in vivo. Nat Commun, 2019, 10: 1–11
64 Liu C B, Zhao Y, Cai X, et al. A wireless, implantable optoelectrochemical probe for optogenetic stimulation and dopamine
detection. Microsyst Nanoeng, 2020, 6: 1–12
65 Xiao G H, Song Y L, Zhang Y, et al. Microelectro de arrays modified with nanocomposites for monitoring dopamine and
spike firings under deep brain stimulation in rat models of Parkinson’s disease. ACS Sens, 2019, 4: 1992–2000
66 Wang M H, Gu X W, Ji B W, et al. Three-dimensional drivable optrode array for high-resolution neural stimulations and
recordings in multiple brain regions. Biosens Bioelectron, 2019, 131: 9–16
67 Rizk M, Bossetti C A, Jochum T A, et al. A fully implantable 96-channel neural data acquisition system. J Neural Eng,

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:17
2009, 6: 026002
68 Liu X L, Zhang M L, Subei B, et al. The PennBMBI: design of a general purpose wireless brain-machine-brain interface
system. IEEE Trans Biomed Circ Syst, 2015, 9: 248–258
69 Bentler C, Stieglitz T. Building wireless implantable neural interfaces within weeks for neuroscientists. In: Proceedings of
the 39th Engineering in Medicine and Biology Society (EMBC), 2017. 1078–1081
70 Kanchwala M A, McCallum G A, Durand D M. A miniature wireless neural recording system for chronic implantation in
freely moving animals. In: Proceedings of IEEE Biomedical Circuits and Systems Conference (BioCAS), 2018
71 Gagnon-Turcotte G, Gagnon L L, Bilodeau G, et al. Wireless brain computer interfaces enabling synchronized optogenetics
and electrophysiology. In: Proceedings of IEEE International Symposium on Circuits and Systems (ISCAS), 2017
72 Shon A, Chu J U, Jung J, et al. An implantable wireless neural interface system for simultaneous recording and stimulation
of peripheral nerve with a single cuff electrode. Sensors, 2018, 18: 1
73 Muller R, Le H P, Li W, et al. A minimally invasive 64-channel wireless µECoG implant. IEEE J Solid-State Circ, 2015,
50: 344–359
74 Liu X L, Zhu H J, Zhang M L, et al. A fully integrated wireless sensor-brain interface system to restore finger sensation.
In: Proceedings of IEEE International Symposium on Circuits and Systems (ISCAS), 2017
75 Laiwalla F, Lee J, Lee A H, et al. A distributed wireless network of implantable sub-mm cortical microstimulators for
brain-computer interfaces. In: Pro ceedings of the 41st Annual International Conference of IEEE Engineering in Medicine
and Biology Society, 2019. 6876–6879
76 Lopez C M, Putzeys J, Raducanu B C, et al. A neural probe with up to 966 electrodes and up to 384 configurable channels
in 0.13 µm SOI CMOS. IEEE Trans Biomed Circ Syst, 2017, 11: 510–522
77 Zhang X, Pei W H, Huang B J, et al. A low-noise fully-differential CMOS preamplifier for neural recording applications. Sci
China Inf Sci, 2012, 55: 441–452
78 Chang S I, Park S Y, Yoon E. Minimally-invasive neural interface for distributed wireless electrocorticogram recording
systems. Sensors, 2018, 18: 263
79 Liu S Y, Moncion C, Zhang J W, et al. Fully passive flexible wireless neural recorder for the acquisition of neuropotentials
from a rat model. ACS Sens, 2019, 4: 3175–3185
80 Yeon P, Bakir M S, Ghovanloo M. Towards a 1.1 mm 2 free-floating wireless implantable neural recording SoC. In: Proceedings
of IEEE Custom Integrated Circuits Conference (CICC), 2018
81 Kim C, Park J, Ha S, et al. A 3 mm×3 mm fully integrated wireless power receiver and neural interface system-on-chip.
IEEE Trans Biomed Circ Syst, 2019, 13: 1736–1746
82 Seo D, Neely R M, Shen K, et al. Wireless recording in the p eripheral nervous system with ultrasonic neural dust. Neuron,
2016, 91: 529–539
83 Lee J, Laiwalla F, Jeong J, et al. Wireless power and data link for ensembles of sub-mm scale implantable sensors near 1
GHz. In: Proceedings of IEEE Biomedical Circuits and Systems Conference (BioCAS), 2018
84 Ghanbari M M, Piech D K, Shen K, et al. 17.5 A 0.8 mm 3 ultrasonic implantable wireless neural recording system with
linear AM backscattering. In: Proceedings of IEEE International Solid-State Circuits Conference (ISSCC), 2019. 284–286
85 Su Y, Routhu S, Moon K, et al. A wireless 32-channel implantable bidirectional brain machine interface. Sensors, 2016, 16:
1582
86 Yoshimoto S, Araki T, Uemura T, et al. Implantable wireless 64-channel system with flexible ECoG electrode and optoge-
netics probe. In: Pro ceedings of IEEE Biomedical Circuits and Systems Conference (BioCAS), 2016. 476–479
87 Lyu L, Ye D, Shi C J R. A 340 nW/channel 110 dB PSRR neural recording analog front-end using replica-biasing LNA,
level-shifter assisted PGA, and averaged LFP servo loop in 65 nm CMOS. IEEE Trans Biomed Circ Syst, 2020, 14: 811–824
88 Jang J W, Kim Y R, Lee C E, et al. A 32ch low power neural recording system with continuously monitoring for ECoG
Signal detection. J Integr Circ Syst, 2021, 7: 2
89 T¨ure K, Dehollain C, Maloberti F. Wireless Power Transfer and Data Communication for Intracranial Neural Recording
Applications. Berlin: Springer, 2020
90 Thakor N V. Translating the brain-machine interface. Sci Transl Med, 2013, 5: 210ps17
91 Shannon C E. Communication in the presence of noise. Proc IEEE, 1998, 86: 447–457
92 Schuettler M, Kohler F, Ordonez J S, et al. Hermetic electronic packaging of an implantable brain-machine-interface with
transcutaneous optical data communication. In: Pro ceedings of International Conference on Information Theoretic Security,
2012. 3886–3889
93 Yin M, Borton D A, Komar J, et al. Wireless neurosensor for full-spectrum electrophysiology recordings during free behavior.
Neuron, 2014, 84: 1170–1182
94 Mestais C S, Charvet G, Sauter-Starace F, et al. WIMAGINE: wireless 64-channel ECoG recording implant for long term
clinical applications. IEEE Trans Neural Syst Rehabil Eng, 2015, 23: 10–21
95 Deshmukh A, Brown L, Barbe M F, et al. Fully implantable neural recording and stimulation interfaces: peripheral nerve
interface applications. J Neurosci Method, 2020, 333: 108562
96 Song Y K, Borton D A, Park S, et al. Active microelectronic neurosensor arrays for implantable brain communication
interfaces. IEEE Trans Neural Syst Rehabil Eng, 2009, 17: 339–345
97 Xu J, Nguyen A T, Zhao W F, et al. A low-noise, wireless, frequency-shaping neural recorder. IEEE J Emerg Sel Top Circ
Syst, 2018, 8: 187–200
98 Sharma D K, Mishra A, Saxena R. Analog & digital modulation techniques: an overview. Int J Comput Sci Commun
Technol, 2010, 3: 2007
99 Idogawa S, Yamashita K, Sanda R, et al. A lightweight, wireless Bluetooth-low-energy neuronal recording system for mice.
Sens Actuat B-Chem, 2021, 331: 129423
100 Jia Y, Khan W, Lee B, et al. Wireless opto-electro neural interface for experiments with small freely behaving animals. J
Neural Eng, 2018, 15: 046032
101 Antonioli G, Baggioni F, Consiglio F, et al. Stinulatore cardiaco impiantabile con nuova battaria a stato solido al litio.
Minerva Med, 1973, 64: 2298–2305
102 Zaeimbashi M, Nasrollahpour M, Khalifa A, et al. Ultra-compact dual-band smart NEMS magnetoelectric antennas for
simultaneous wireless energy harvesting and magnetic field sensing. Nat Commun, 2021, 12: 1–11
103 Lee B, Koripalli M K, Jia Y, et al. An implantable peripheral nerve recording and stimulation system for experiments on
freely moving animal subjects. Sci Rep-Uk, 2018, 8: 1–12
104 Moon E, Barrow M, Lim J, et al. Bridging the “last millimeter” gap of brain-machine interfaces via near-infrared wireless

Ji B W, et al. Sci China Inf Sci April 2022 Vol. 65 140401:18
power transfer and data communications. ACS Photonics, 2021, 8: 1430–1438
105 Xie X, Rieth L, Williams L, et al. Long-term reliability of Al2O3and Parylene C bilayer encapsulated Utah electrode array
based neural interfaces for chronic implantation. J Neural Eng, 2014, 11: 026016
106 Fang H, Zhao J N, Yu K J, et al. Ultrathin, transferred layers of thermally grown silicon dioxide as biofluid barriers for
biointegrated flexible electronic systems. Pro c Natl Acad Sci USA, 2016, 113: 11682–11687
107 Shen K, Maharbiz M M. Ceramic packages for acoustically coupled neural implants. In: Proceedings of the 9th International
IEEE/EMBS Conference on Neural Engineering (NER), 2019. 847–850
108 Yao J L, Qiang W J, Wei H, et al. Ultrathin and robust micro-nano composite coating for implantable pressure sensor
encapsulation. ACS Omega, 2020, 5: 23129–23139
109 Bettinger C J, Ecker M, Kozai T D Y, et al. Recent advances in neural interfaces-materials chemistry to clinical translation.
MRS Bull, 2020, 45: 655–668
110 Sharma A, Rieth L, Tathireddy P, et al. Evaluation of the packaging and encapsulation reliability in fully integrated, fully
wireless 100 channel Utah Slant electro de array (USEA): implications for long term functionality. Sens Actuat A-Phys, 2012,
188: 167–172
111 Hwang G T, Im D, Lee S E, et al. In vivo silicon-based flexible radio frequency integrated circuits monolithically encapsulated
with biocompatible liquid crystal polymers. ACS Nano, 2013, 7: 4545–4553
112 Kiourti A, Lee C W L, Chae J, et al. A wireless fully passive neural recording device for unobtrusive neurop otential
monitoring. IEEE Trans Biomed Eng, 2016, 63: 131–137
113 Neely R M, Piech D K, Santacruz S R, et al. Recent advances in neural dust: towards a neural interface platform. Curr
Opin NeuroBiol, 2018, 50: 64–71
114 Luan H W, Zhang Y H. Programmable stimulation and actuation in flexible and stretchable electronics. Adv Intell Syst,
2021, 3: 2000228