Becerra-Fajardo et al.

Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Journal of NeuroEngineering

https://doi.org/10.1186/s12984-023-01295-5
and Rehabilitation

BRIEF REPORT Open Access

First-in-human demonstration of floating EMG

sensors and stimulators wirelessly powered
and operated by volume conduction
Laura Becerra-Fajardo1, Jesus Minguillon1,2,3, Marc Oliver Krob4, Camila Rodrigues5,6, Miguel González-Sánchez7,
Álvaro Megía-García8, Carolina Redondo Galán8, Francisco Gutiérrez Henares8, Albert Comerma1,
Antonio J. del-Ama9, Angel Gil-Agudo8,10, Francisco Grandas7, Andreas Schneider-Ickert4, Filipe Oliveira Barroso5,10
and Antoni Ivorra1,11*

Abstract
Background Recently we reported the design and evaluation of floating semi-implantable devices that receive
power from and bidirectionally communicate with an external system using coupling by volume conduction. The
approach, of which the semi-implantable devices are proof-of-concept prototypes, may overcome some limitations
presented by existing neuroprostheses, especially those related to implant size and deployment, as the implants
avoid bulky components and can be developed as threadlike devices. Here, it is reported the first-in-human acute
demonstration of these devices for electromyography (EMG) sensing and electrical stimulation.
Methods A proof-of-concept device, consisting of implantable thin-film electrodes and a nonimplantable miniature
electronic circuit connected to them, was deployed in the upper or lower limb of six healthy participants. Two
external electrodes were strapped around the limb and were connected to the external system which delivered high
frequency current bursts. Within these bursts, 13 commands were modulated to communicate with the implant.
Results Four devices were deployed in the biceps brachii and the gastrocnemius medialis muscles, and the external
system was able to power and communicate with them. Limitations regarding insertion and communication speed
are reported. Sensing and stimulation parameters were configured from the external system. In one participant,
electrical stimulation and EMG acquisition assays were performed, demonstrating the feasibility of the approach to
power and communicate with the floating device.
Conclusions This is the first-in-human demonstration of EMG sensors and electrical stimulators powered and
operated by volume conduction. These proof-of-concept devices can be miniaturized using current microelectronic
technologies, enabling fully implantable networked neuroprosthetics.
Keywords Wireless power transfer, Volume conduction, AIMDs, Bidirectional communications, Electromyography,
Sensor, Electrical stimulation, Neuroprostheses, Semi-implantable devices, Human validation.

*Correspondence:
Antoni Ivorra
[email protected]
Full list of author information is available at the end of the article

© The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use,
sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and
the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included
in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The
Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available
in this article, unless otherwise stated in a credit line to the data.
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 2 of 16

Background HMI) [6] and crosstalk, and allow the control of more
Neuroprosthetic technologies based on active implant- degrees-of-freedom through the use of multiple elec-
able medical devices (AIMDs) have the potential to be trodes [7].
an important tool for improving the quality of life of Within the EXTEND collaborative project, funded by
patients with neurological disorders and patients that the European Commission and in which the authors of
require external systems as prostheses and exoskeletons. this study participated, it has been defined the concept of
For example, intramuscular microstimulators could be Bidirectional Hyper-Connected Neural Systems (BHNS).
used to reduce pathological tremor in essential tremor It refers to systems consisting of minimally invasive com-
patients, as intramuscular electrical stimulation has been munication links between multiple nerves or muscles
shown to reduce tremor in this population of patients in the body and external devices which may be digitally
up to 24 h [1]. Moreover, intramuscular electromyogra- interconnected between them (Fig. 1a). This will pro-
phy (EMG) microsensors could be used to control pros- vide the means of a synthetic chain of action-reaction of
theses and exoskeletons [2–4], especially as it has been sensorimotor activity using musculoskeletal modelling,
demonstrated to be a reliable input to feed neuromus- aiding in applications as tremor management and HMI
culoskeletal models to estimate the user-intended joint control. The BHNS is envisioned as a dense network of
movements for human-machine interface (HMI) control wireless implantable devices that can act as stimulators
[5]. and EMG sensors, and that communicate in real time
Implantable neuroprostheses form a chronic inter- with the external systems. The external systems, in turn,
face with the nervous system that overcomes one of the process and analyze the neuromuscular activity and con-
most important limitations of EMG sensors and electri- trol the stimulation (e.g., for tremor management) and/or
cal stimulators based on surface electrodes: the lack of the action of machines (e.g., exoskeletons).
selectivity. In addition, implantable electrodes improve This BHNS paradigm requires fully implantable wire-
repeatability by avoiding the impact of skin imped- less devices that can be easily deployed by injection. Per-
ance changes and movements, and allow accessing deep cutaneous systems would not be adequate to implement
muscles. In the case of EMG sensing for HMI, implant- the BHNS concept because they require daily mainte-
able electrodes avoid the need to replace the electrodes nance and dressing, and may present sweat gland block-
at each use (which in turn requires recalibration of the age [7]. Most fully implantable devices are developed as a

Fig. 1 Bidirectional Hyper-Connected Neural System (BHNS) concept and proof-of-concept semi-implantable devices. (a) Schematic representation of
the envisioned BHNS. (b) Miniature electronic circuit for EMG sensing and stimulation and its comparison with a 1 cent euro coin for acute human dem-
onstration of the BHNS concept. (c) Complete proof-of-concept semi-implantable device showing the dedicated insertion needle, the thin-film intramus-
cular electrodes, the thin guiding filament that exits through the needle hub, and the polymer capsule that houses the miniature circuit (opened lid) [17]
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 3 of 16

central metallic case that houses the control unit, power presents the use of these proof-of-concept semi-implant-
circuit and pulse generator. The electronics are connected able devices in the limbs of healthy participants, a first-
to the electrodes through leads, requiring complex sur- in-human demonstration of devices wirelessly powered
gical implantation procedures. Moreover, the leads tend and operated by volume conduction.
to fail [8, 9]. As with the percutaneous systems, this
approach would not be adequate for the implementation Methods
of the BHNS concept. To avoid the leads and their asso- Electronic system
ciated complications, it has been proposed the develop- The electronic system consists of (1) an external system
ment of wireless networks of single-channel injectable that delivers the HF current bursts for WPT and bidirec-
devices that integrate the electrodes and the electronics. tional communications, and (2) wireless devices for EMG
Nevertheless, the injectable devices capable of intramus- sensing and electrical stimulation. This electronic system
cular stimulation that were developed in the past are stiff was extensively described in [17]. Yet some aspects are
and relatively large (diameter > 2 mm) due to the need of highlighted here to facilitate the reading of this first-in-
batteries — with their limited lifespan and large volume human demonstration.
[10] — or to the need of components required for the
wireless power transfer (WPT) method used to energize External system
them (e.g., coils in inductive coupling [11–13]). Recently, The selected hierarchical architecture for the BHNS pro-
WPT methods as inductive coupling, ultrasonic acous- poses the use of an external system that consists of one
tic coupling and capacitive coupling have obtained high top-level controller that communicates with the wireless
miniaturization levels at the expense of link efficiency, AIMDs through external low-level control units. These
penetration depth, or functionality. Reviews on these control units apply HF current bursts to the tissues to
methods can be found in [10, 14–16]. power and communicate with the wireless devices. The
In the last years we have proposed and demonstrated currents are applied using external textile electrodes
a WPT method that avoids the use of bulky compo- strapped around the limbs (Fig. 1a).
nents inside the implant, and that allows to implement The implemented control unit delivers HF bursts for
the implants as thin, flexible and elongated devices suit- powering. They include a single 30 ms “Power up” burst
able for implantation by means of injection [18, 19]. to power up the wireless implantable devices located
The implants are powered and controlled by applying between the two external electrodes, and power mainte-
— through textile electrodes — high frequency (HF) nance bursts with a repetition frequency (F) of 50 Hz and
current bursts that flow through the tissues by volume duration (B) of 1.6 ms to keep them energized. To per-
conduction [20]. The currents are picked-up by the form bidirectional communications, a custom communi-
implant’s electrodes (located at the ends of the elongated cation protocol stack was created. In the case of downlink
body, Fig. 1a) and are rectified for powering and bidirec- (i.e., information sent from the external system to the
tional communications. This allows the integration of wireless devices), the HF bursts are amplitude modulated
the electronics in an application-specific integrated cir- by the external system. A set of 13 different commands
cuit (ASIC). We have validated in humans that HF cur- can be used to (1) ask if an implant with a specific address
rent bursts complying with safety standards and applied is active (hereinafter referred to as “Ping”), (2) configure
through two textile electrodes strapped around a limb the sensing and stimulation parameters, (3) control their
can provide substantial power from pairs of implanted execution, and (4) uplink EMG samples. In the case of
electrodes, and are innocuous and imperceptible [21]. uplink (i.e., information sent from a wireless device to
We have recently reported the development and the external system), the external system delivers HF cur-
evaluation of semi-implantable devices based on WPT rents that are modulated – by means of load modulation
by volume conduction that are capable both of per- – by the wireless devices to generate an uplink reply. This
forming stimulation and EMG sensing [17]. These are is seen by the control unit of the external system as min-
proof-of-concept prototypes of the electronic architec- ute changes in current.
ture that will be integrated into an ASIC to obtain flex- The duration of the downlink and uplink bursts, as
ible threadlike devices as those demonstrated in vivo in well as the EMG sensing and stimulation parameters
[22], with stimulation, EMG sensing and bidirectional that can be set from the external system are reported in
communication capabilities. The external system and a [17]. The shortest uplink message is that replied after the
bidirectional communications protocol that allows wire- command “Ping”. During downlink, the external system
less communication between the external system and sends this command to a specific device identifier. The
the semi-implantable devices are also described in [17]. implants located between the two external electrodes
The devices were evaluated using an agar phantom and receive this information, and if their address matches that
in hindlimbs of anesthetized rabbits. This brief report sent by the external system, an 8-bit frame acknowledge
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 4 of 16

(ACK) is uploaded. In the case of the downlink “Get Con- consumption modes to guarantee energy efficiency. The
fig” command, which is used to know the current EMG current consumption measured during idle mode, pro-
acquisition or stimulation configuration of the device, the cessing and basic operations mode, and sensing mode
floating device replies with an uplink message encoded in are 180 µA, 185 µA, and 400 µA respectively [17]. With
3 bytes. The longest uplink message is received when the this current consumption, it is possible to power all the
external system requests an EMG sample. The addressed electronic components and the control unit’s peripherals
device replies with 4 bytes. Because of this, “Ping” is the required for the power mode defined. At first, the device
initial command used to test the performance of bidirec- is configured to power up and wait for instructions sent
tional communications. If the performance is adequate, from the external system (i.e., idle mode). The power up is
more complex commands (i.e., 3 and 4 bytes) are tested. done using the 30 ms “Power up” burst, and the device is
kept energized with the power maintenance bursts (burst
Wireless devices for EMG sensing and electrical stimulation frequency F of 50 Hz, and duration B of 1.6 ms) using a
As explained in [17], the proof-of-concept wireless device set of two 10 µF capacitors connected in parallel, which
for the acute human demonstration reported here con- provides a stable dc voltage for the control unit and the
sists of ultrathin intramuscular electrodes based on thin- rest of the electronics during and in-between HF bursts.
film technology. Their biocompatibility was previously To perform electrical stimulation or EMG sensing, the
demonstrated in cytotoxicity, sensitization and irritation device must be first configured from the external system.
tests according to ISO 10,993 [23]. The intramuscular In the case of electrical stimulation, the external system
electrodes are connected to a miniature electronic circuit can program the device to deliver biphasic or monopha-
to be fixed on the skin. Each device has only two double- sic pulses, and the direction of the pulses. The external
sided electrodes on a narrow polyimide filament (coined system sets the frequency of stimulation, the pulse width
Electrodes ‘A’ and ‘B’ in Fig. 1c) for powering, bidirec- of the stimulation pulse, and the interphase dwell in case
tional communications, electrical stimulation, and EMG a biphasic symmetric pulse is required. This is done by
sensing. This main filament has a width of 0.42 mm, modifying the frequency and duration of the applied HF
a length of 81.6 mm, and a thickness of 0.02 mm. Each bursts during the stimulation phase. Usually, the users set
electrode contact has a width of 0.265 mm and a length the stimulation frequency between 30 and 200 Hz, the
of 7.5 mm, and their edges are rounded to avoid sharp pulse width between 100 and 200 µs, interphase dwell of
corners that would lead to high current densities. The 30 µs, and cathodic-first pulse. The current limiters inside
distal contacts located in the top and bottom layers of the the electronic circuit fix a maximum stimulation ampli-
main filament are electrically connected in the miniature tude of 2 mA. In the case of EMG sensing, the external
electronic circuit to use them as a single distal electrode system must configure the type of EMG acquisition to
(Electrode ‘A’), with a total surface area of 3.8 mm2. This be done (raw, or 2 options of parametric acquisition), the
is also done with the two proximal contacts located in the sampling frequency (250, 500, 750 and 1000 Hz, similar
top and bottom layers of the filament, creating a proxi- to those reported in [24–27]), and the window size in
mal electrode of 3.8 mm2 (Electrode ‘B’). The impedance case the parametric option is set (15 different options
magnitude of the electrodes, as measured across two starting at 10 ms, up to 500 ms, similar to those reported
electrode pairs in 0.9% NaCl, is flat beyond 10 kHz, and in [1, 28–30]). These specific parameters (i.e., algorithm
lower than 190 Ω at 1 MHz [17]. The distance between for parametric acquisition, possible sampling frequen-
the electrodes’ centers is 30 mm. To facilitate the inser- cies and windows) can be changed by reprogramming
tion of the electrodes in the muscle, the distal end of the the control unit of the proof-of-concept device using the
electrodes’ filament has a U-shape structure that ends 6 programming pads accessible from the top layer of the
in a guiding filament (Fig. 1c). This guiding filament is circuit. When the EMG acquisition is started from the
inserted through the lumen of a 23 G hypodermic needle external system, the wireless device acquires a sample
(Sterican 4,665,600 by B. Braun Melsungen AG) having a and stores it in the random-access memory (RAM) of the
length of 60 mm and an outer diameter of 0.6 mm, and control unit (ultra-low power microcontroller, MKL03Z-
the main filament with the active electrodes runs exter- 32CAF4R by NXP Semiconductors N.V.). Because of the
nally to the needle. As the guiding filament is so thin and size of the microcontroller, this memory is limited to 2
the U-shape lies very close to the needle’s bevel, the bevel kB. When acquiring in parametric mode, the parametric
was smoothed with a laser (Picco Laser, by O.R. Laser- value is calculated, and the result obtained for a sampling
technologie, Germany) prior to inserting the guiding fila- window is saved in memory, improving the efficiency of
ment inside the needle and towards the needle’s hub. memory usage.
The miniature electronic circuit connected to the The analog front-end (AFE) designed is explained in
intramuscular electrodes (Fig. 1b) includes a microcon- detail in [17]. It has been configured to have a gain of 54
troller that has been programmed to run in three power dB, a signal-to-noise ratio (SNR) of 46 dB, and a common
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 5 of 16

mode rejection ratio (CMRR) of 88 dB. The circuit board by the external system. By doing this blanking [31], it is
includes a test point to access the output of the AFE for possible to identify saturation and recovery time win-
debugging. The AFE’s output is connected to the con- dows without misinterpreting this effect with high ampli-
trol unit’s analog-to-digital converter (ADC), which is tude EMG activity.
programmed to have a resolution of 10 bits. This resolu- Figure 2 shows the different sequences that can be
tion can be increased to 12 bits by reprogramming the obtained with the interaction of the external system and
device with the programming pads. During EMG sens- its low-level controller, and the wireless implants that are
ing, the control unit monitors the output of its downlink within the textile electrodes of this controller. During
demodulator to identify if a HF burst is being applied by downlink, the information is modulated in the HF cur-
the external system. If this is the case, the AFE’s output rents, and is demodulated by the implant (Fig. 2a). If an
is saturated by the HF bursts for a very short period of ACK reply is requested (e.g., with the “Ping” command),
time (1.6 ms, equivalent to the duration of the burst), and the external system delivers a burst, and the implant
then recovers from it in approximately 4 ms. When this does load modulation to uplink this information, which
saturation artifact happens, the control unit replaces the is seen by the control unit as minute changes in current
samples corresponding to this saturation with a constant consumption (Fig. 2b). For stimulation, the external sys-
value (e.g., five samples for a sampling frequency of 1 tem delivers HF bursts with the pulse width and repeti-
ksps), without affecting the sampling frequency defined tion frequency defined for stimulation, and the implant

Fig. 2 HF current bursts sequence. It shows the different sequences that can be obtained with the combination for the external system and its low-level
control unit, and the wireless implantable devices. See main text for detailed information
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 6 of 16

rectifies these volume-conducted HF bursts to cause The approximate site for the deployment was defined
local low frequency currents capable of stimulating excit- using anatomical cues. The criteria were (1) that the elec-
able tissues. Figure 2c shows a low-pass filtered biphasic trodes were aligned to the applied electric field, (2) that
stimulation waveform obtained in vitro using an oscil- at least 40 mm of the needle were completely inside the
loscope [17]. For EMG acquisition, the external system muscle to guarantee that both electrodes were in this tis-
delivers power maintenance bursts (50 Hz, 1.6 ms) which sue, and (3) if possible, to insert the tip of the needle –
causes a short saturation in the AFE of the implant, fol- which is few mm away from Electrode ‘A’ – close to the
lowed by a time window for EMG acquisition (Fig. 2d). motor point of the target muscle. Before inserting the
The example shows the output of the AFE (i.e., AFE test thin-film electrodes of the semi-implantable device, the
point) obtained in vitro using an oscilloscope when no area was cleaned with chlorhexidine. This procedure was
EMG signal is present across the intramuscular elec- guided by ultrasound to verify that the needle was cor-
trodes [17]. The recorded samples include the samples rectly inserted in the muscle belly, and that its orienta-
with saturation artifacts (samples to be blanked by soft- tion was as parallel as possible to the skin (to maximize
ware) and the valid samples obtained during the acquisi- the alignment of the electrodes with the applied electric
tion window. field delivered using the external textile electrodes). After
For this first-in-human demonstration, the proof-of- the needle was inserted in the target location, the min-
concept devices were double blister-packaged in clear iature electronic circuit (enclosed in a polymer capsule
Stericlin bags and heat sealed. The devices were sterilized for protection) was fixed to the skin of the participant
by Ethylene oxide gas sterilization (8%) at low tempera- using adhesive tape to avoid the accidental extraction of
tures (38–42 °C). the electrodes. Finally, the end of the guiding filament
adhered to the needle hub was cut, and the needle was
Participants gently extracted. Figure 3 shows proof-of-concept semi-
The assays were performed at the Movement Disorders implantable devices deployed in the biceps brachii of
Clinic of Gregorio Marañón Hospital (Madrid, Spain) participant 2, and in the gastrocnemius medialis of par-
for the upper limb and at the Biomechanics and Assis- ticipant 5.
tive Technology Unit of National Hospital for Paraplegics Once the intramuscular electrodes were successfully
(Toledo, Spain) for the lower limb. The procedures were inserted in the muscle belly, the participants remained
conducted in accordance with the Declaration of Hel- seated on a chair. Then, the textile electrodes were
sinki and approved by the local Ethics Committees (refer- strapped around the limb and were connected to the
ence numbers 18/2020 and 565 respectively), as well as external system. In the case of the upper limb assays, the
by the Spanish Agency of Medicines and Medical Devices participant was asked to keep the upper arm aligned with
(AEMPS) – records 858/20/EC and 856/20/EC respectively. the torso, to lay down the forearm on the armrest (thus
Three healthy volunteers participated in the upper limb forming a 120° angle in the elbow), and to keep the wrist
assays (23, 31, and 47 years old, all male), while other three in neutral position.
healthy volunteers participated in the lower limb assays (29, Several tests were performed to test the ability of the
32 and 37 years old; two males, one female). The volunteers external system to power and bidirectionally communi-
were recruited through a call for participation sent by email cate with the proof-of-concept device. To evaluate the
to colleagues and by another call posted in the facilities of bidirectional communications, it has been defined the
both hospitals, and were not paid for their participation in success rate: the number of messages correctly decoded
the study. Before starting the experimental procedure, the by the external system (testing the downlink and uplink
participants were provided with oral and written informa- sequence), divided by the number of messages that
tion regarding the study (including risks, contingency mea- should have been theoretically received. If any of the
sures, benefits, and data protection aspects) and signed an error detection mechanisms available in the external unit
informed consent form. is triggered (i.e., error in Manchester coding, parity bit,
frame length or command code), the message is marked
Experimental procedure as incorrect. In the case of a “Ping”, the success rate
At first, the target muscle was defined for each partici- would only evaluate the 8-bit frame ACK. In much more
pant. In the case of the upper limb, the target muscle complex commands and subsequent replies such as “Get
was the biceps brachii or the triceps brachii; while in the Config”, the message is understood as the total amount
lower limb it was the tibialis anterior, the gastrocnemius of information (3 bytes or 4 bytes) to be received by the
medialis or the gastrocnemius lateralis. A single semi- external system.
implantable floating EMG sensor and stimulator was At first, the external system was set at an initial peak
set up per participant. The electrodes were implanted amplitude of 30 V, and it was configured to deliver the
using the dedicated 23 G hypothermic needle (Fig. 1c). “Power up” burst and “Pings” requests. If the address
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 7 of 16

Fig. 3 Proof-of-concept semi-implantable devices deployed in the target muscles of two participants. (a) Device in biceps brachii of participant 2. (b)
Device in gastrocnemius medialis of participant 5

Fig. 4 Setup used during the study in upper limb. The participant was always sitting on a chair, with the upper arm aligned with the torso and the elbow
at an angle of 120°. The strapped external textile electrodes were connected to the external system. (a) For the force measurement during isometric con-
tractions of the biceps brachii, the wrist was strapped to the armrest of the chair, and to a force gauge. (b) Zoom of the region where the wireless circuit
was located, and approximate location of thin-film electrodes

sent during downlink matched the address defined in After it was confirmed that the semi-implantable
the semi-implantable device, this device sent an ACK device was properly powered by the external system,
frame to the external system. The peak amplitude was and that the bidirectional communications were success-
then increased until uplink waveforms with the ACK ful, electrical stimulation and EMG sensing assays were
frame were seen in the demodulator’s shunt resistor of performed. In the case of electrical stimulation assays,
the external system (Fig. 4a). If this bidirectional com- videos were recorded to capture the moment in which
munication trial was successful (i.e., success rate above a stimulation sequence was triggered from the external
90%), longer communication frames were tested to guar- system. In the case of EMG sensing for biceps brachii
antee that the external system could correctly configure assays, it was not possible to assess muscle contraction
the semi-implantable device for electrical stimulation using surface electrodes connected to a commercial
and EMG sensing, trigger these functions (i.e., stimulate, EMG amplifier as the amplifier saturated with the HF
start/stop sensing) and upload the samples of EMG activ- current bursts applied – as expected – but it could not
ity acquired by the floating device. return from the saturation fast enough to acquire EMG
signals in-between bursts. For this reason, a force gauge
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 8 of 16

was used to measure the tensile force exerted by the arm Table 1 Demographic information and bidirectional
when the biceps brachii was contracted with the hand communications success rate in each participant
closed. The force gauge was attached firmly to the partici- Participant Gender Age Target Communications
muscle success rate
pant’s arm – which was held in the armrest with an atrau-
1 Male 47 Biceps brachii No
matic band – and its base was fixed to a base of the chair, communication
forcing an isometric contraction (Fig. 4). The participant 2 Male 31 Biceps brachii Ping: 100%
was requested to perform different force profiles, such Get Config: 100%
as maximum sustained contraction, target forces (e.g., 3 Male 23 Biceps brachii Ping: 100%
3 kg, 5 kg, 10 kg), and sequences of hard and short con- 4 Female 37 Gastrocne- No
tractions. In the case of target forces, the obtained mea- mius medialis communication
surements were read aloud as a means of feedback for the 5 Male 29 Gastrocne- Ping: 40%
participant. During the contractions, EMG was acquired mius medialis
by the floating device, and the samples were uploaded to 6 Male 32 Gastrocne- Ping: 80%
mius medialis
the external system for offline analysis.
Each participant was constantly asked if he or she per-
ceived any type of sensation related to the HF current
bursts applied or to the floating device implanted, and
the region close to the textile electrodes was visually
monitored continuously. The surveillance also included
measurements of the impedance across the external tex-
tile electrodes to identify possible changes in the inter-
face between the textile electrodes and the skin. This was
done using an active differential probe (TA043 from Pico
Technology Ltd) to measure the voltage applied across
the external electrodes, and a current probe (TCP2020A
from Tektronix, Inc.) to measure the HF current flow-
ing through the tissues. Both probes were connected to a
battery-powered oscilloscope (TPS2014 from Tektronix,
Inc.).
After finalizing the assays, the external system was
turned off and the textile electrodes were unstrapped.
Fig. 5 Ultrasounds image obtained from participant 2 immediately after
Then, the intramuscular electrodes were extracted by the needle was inserted in the biceps brachii. The needle was injected
holding the guiding filament and the electrodes’ filament from the top right corner (not shown in the image), passed through sub-
and gently pulling out, and the insertion region on the cutaneous fat and the muscle fascia, and ended up in the target muscle.
skin was cleaned with chlorhexidine. The participants The tip of the needle is seen at the left of the image at a depth of approxi-
mately 1 cm
remained for 30 more minutes in the hospital for moni-
toring, and were monitored again 24 h later through a
phone call to ensure their well-being and to note possible the study was finally performed only in the biceps bra-
adverse effects. chii and the gastrocnemius medialis. In the cases in
which the implantation process failed, the broken thin-
Results film electrodes and the needle were gently extracted by
Implantation simultaneously pulling from the main filament, the guid-
The insertion mechanism was minimally invasive and did ing filament and the needle hub. Table 1 summarizes the
not require local anesthesia. In the case of the implan- final location of the proof-of-concept device in all the
tation procedure in the triceps brachii, due to limita- participants.
tions in the room, the position during implantation Additional file 1 shows a video of the implantation pro-
was uncomfortable for the physician. This hampered cedure in the biceps brachii of participant 2 and in the
the correct insertion of the electrode, which broke dur- gastrocnemius medialis of participant 5. Figure 5 shows
ing the implantation. In the case of the tibialis anterior the ultrasound image obtained immediately after the
and gastrocnemius lateralis muscles, the intramuscu- needle was placed in the biceps brachii of participant
lar electrodes broke during insertion as the dedicated 2. Only 3.4 cm of the 6 cm needle are seen in the image
needle experienced more stiffness due to the insertion (limited by the ultrasound’s transducer length), but the
angle required to maximize the alignment of the elec- resulting insertion length was approximately 5.5 cm. The
trodes with the applied electric field. For these reasons, participant had a subcutaneous fat layer of approximately
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 9 of 16

3 mm, and the needle – corresponding to the position of (Fig. 6a), who had the semi-implantable device deployed
the thin-film electrodes – had an angle of approximately in the biceps brachii, success rates of 100% were obtained
8º with respect to the skin. for bidirectional communications, both in short frames
After the implantation, the participants did not report as the “Ping” command, as well as in longer frames such
discomfort due to the intramuscular electrodes. The as “Get Config” when the external system was delivering
filament was thin enough so that the participants could bursts with a peak amplitude of 60 V and 1.42 A, and the
move the limb under study without any type of sensa- textile electrodes were separated 10 cm. It is worth not-
tion. For this same reason, the participants were asked to ing that these peak amplitudes for electric potential and
remain seated until the study was finished to avoid acci- current are scaled down due to the use of bursts that have
dental extraction. a duty cycle of 0.08 (power maintenance burst frequency
F: 50 Hz, and duration B: 1.6 ms). For another upper limb
Bidirectional communications participant, participant 3, success rates of 100% were
Figure 6 shows an example of the waveforms obtained obtained for pings to a device implanted in the biceps
from the external system when it requested a “Ping”, and brachii. In the case of participant 5 (Fig. 6b), whose float-
the floating device replied with an ACK frame. The wave- ing device was in the gastrocnemius medialis, the maxi-
forms were obtained in the upper arm (Fig. 6a) and the mum success rate for pings was only 40% (68 V, 2.28 A,
lower leg (Fig. 6b and c) of three different participants. 10 cm distance between centers of textile electrodes). Its
The external system was delivering HF bursts at a fre- orange waveform clearly shows a HF component in the
quency of 1.1 MHz (power maintenance burst frequency uplink waveform (i.e., output of demodulator of external
F: 50 Hz, and duration B: 1.6 ms). For any uplink, the system), which may have generated an incorrect digital
external system delivers a HF current burst to trigger signal in the analog comparator of the external system’s
the reply and give energy to the semi-implantable device control unit, therefore generating errors in the decoder.
(Fig. 6, blue waveform). The load modulation made by Additionally, the waveform shows low peak-to-peak vari-
the floating device creates minute changes in the cur- ations during the ACK, which hampers even more the
rent flowing through the external textile electrodes. This decoding of the data. In the case of participant 6 (Fig. 6c),
current is measured by the external system’s demodu- whose device was also implanted in the gastrocnemius
lator using a shunt resistor (Fig. 4). Figure 6 – orange medialis, the maximum success rate for pings was 80%
waveform – shows the output of the demodulation cir- (90 V, 2.1 A, 12 cm distance between textile electrodes).
cuit after filtering and amplifying the voltage across The orange uplink waveform also contains some HF com-
this shunt resistor. This output later passes by an analog ponent, but the ACK ping message differs clearly from
comparator, whose output connects to the receiver of the rest of the window.
the external system’s universal asynchronous receiver- The “Ping” requests sent to the device implanted in the
transmitter (UART) to decode the information sent by biceps brachii of participant 1 and the gastrocnemius
the semi-implantable device. In the case of participant 2 medialis of participant 4 were not properly received by

Fig. 6 Waveforms obtained with a floating oscilloscope and the probes connected to the external system. These waveforms are obtained after the
external system requests a “Ping” to a specific device. In blue, HF current delivered by external system for uplink; in orange, output of external system’s
demodulator, showing the filtered and amplified voltage across the shunt resistor when the floating device does an ACK ping message by load modu-
lating the HF current burst (i.e., shows changes in the current flowing through the textile electrodes). (a) Participant 2, biceps brachii. (b) Participant 5,
gastrocnemius medialis. (c) Participant 6, gastrocnemius medialis
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 10 of 16

the external system. In both cases it is unclear whether RMS is an amplitude estimator while ZC, calculated by
the devices were not properly powered (e.g., because finding the rate of crosses by zero in the time domain, is
the proximal electrode (‘B’ in Fig. 1c) was lying in fat tis- an indicator of the changes of the spectral content of the
sue), or whether the hardware-based demodulation and EMG [33]. The recorded signals were divided into 50 ms
decoding circuit of the external system was not able to windows, and for each window the RMS and ZC values
properly acquire and interpret the signal modulated by were calculated. Figure 7c-d show the mean RMS and
the floating devices when they were sending their ACK mean ZC respectively obtained for each raw signal. The
ping message (e.g., because the modulation index was too threshold for the offline ZC algorithm was defined as 0.
low for the digital converter to work properly). As expected, both parameters increased when the force
increased.
Electrical stimulation and EMG sensing To avoid the limitations imposed by the internal mem-
As the semi-implantable device in participant 2 was ory of the miniature electronic circuit of the floating
able to be powered by and to bidirectionally commu- device, and to be more efficient with the communica-
nicate with the external system in both short and long tion channel by minimizing the amount of data to upload
communication frames, it was possible to configure the for the control of the BHNS, the floating device could
stimulation and EMG acquisition parameters of the float- internally calculate a parametric value of the obtained
ing device. Additional file 2 shows a video of a stimula- EMG. To limit the calculation time of the control unit
tion sequence when the floating device was configured (i.e., the amount of time it must be active in its higher
to do biphasic symmetric stimulation with a cathodic- power mode), it was considered that the parameters
first pulse (i.e., a cathodal phase is followed by an anodal should be obtained from the signal in the time domain.
phase to obtain a charge-balanced pulse [32]). This One of these parameters is the rate of ZC, which is eas-
sequence is shown in the schematic representation of the ier to implement in hardware [34]. A trial of parametric
stimulation shown in Fig. 2c. The external system deliv- EMG based on ZC was done to evaluate the possibility
ered HF current bursts to generate a stimulation with a of performing longer recordings. The floating device was
frequency of 150 Hz, 120 µs pulse width and interphase configured to do EMG acquisition at a sampling rate of
dwell of 30 µs, with a maximum stimulation amplitude of 1 ksps and parametrize the results using ZC with a win-
2 mA. When the stimulation was triggered by the exter- dow size of 30 ms, a window size that could be used for
nal system, a visual feedback confirmed that the skin real-time control. The participant was asked to do three
area over the region where the thin-film electrodes were hard and short contractions, followed by relaxations,
located slightly sank, indicating muscle contraction. Ana- in cycles of 2 s in a 6 s frame (progressive contractions
lyzing the video frames of Additional file 2, it is estimated starting approximately at seconds 0, 2 and 4), while the
that the skin area sank approximately 1.2 mm during semi-implantable device was acquiring the informa-
stimulation. tion. Figure 8 shows the result of this assay, in which the
In the case of EMG acquisition, it was possible to three progressive contractions can be clearly noticed. An
obtain raw EMG. Figure 7a shows two raw waveforms offline moving average filter of 15 samples was used to
recorded by the wireless device and transmitted to the highlight the envelope of the obtained parametric EMG.
external system: the baseline (blue) and a voluntary During the trials to assess bidirectional communica-
maximum sustained contraction done by the participant tions, electrical stimulation and EMG acquisition, the
(orange). The floating device was configured to acquire participants did not report any type of perception related
raw EMG at a sampling rate of 1 ksps. The method for to the application of the HF current bursts. This aligns
blanking the artifacts during the AFE’s saturation and with the results obtained in [21].
recovery allowed us to distinguish this time window and
correctly obtain EMG recordings during the acquisition Extraction
window (Fig. 2d) despite the effect of the HF current After the assays were performed on the upper limb or
burst in the output of the AFE. lower limb of the participants, the intramuscular thin-
Using a force gauge as a reference (Fig. 4a), partici- film electrodes were gently extracted. The electrodes sled
pant 2 was asked to contract his arm to target a specific smoothly through the tissues, and both the main filament
force. Figure 7b shows raw EMG recordings obtained for with the active sites of the electrodes and the thin guiding
four trials, corresponding to no contraction (baseline), filament were extracted completely. There was no bleed-
3 kg, 5 and 10 kg of force. As the target force increases, ing, pain, or secondary effects.
the electrical activity measured from the floating device
increases. These raw signals (Fig. 7b) were used to quan-
tify offline the root mean square (RMS) value and the zero
crossings (ZC) rate to evaluate its use in a future BHNS.
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 11 of 16

Fig. 7 Examples of raw EMG obtained by the floating EMG sensor and stimulator in biceps brachii of participant 2. (A) Comparison between baseline and
sustained contraction. (b) Comparison between baseline and three different target forces exerted by the participant. (c) Offline calculation of average of
root mean square (RMS) value for the raw signals shown in b. d) Offline averaging of zero crossing (ZC) calculation for the raw signals shown in b. In both
cases, the window used for calculation was set to 50 ms

Discussion and (2) that the power picked up by the floating device
Here it is presented a brief report of the first-in-human during the assays was enough to power the miniature
demonstration of floating devices powered and con- electronic circuit. To accomplish this, the needle had to
trolled by volume conduction, which are capable of be completely inserted as longitudinal as possible to the
EMG sensing and electrical stimulation. These proof-of- muscle, in acute angle with the skin and the muscle fas-
concept semi-implantable devices were designed for the cia. Remarkably, the muscle fascia is an anisotropic con-
acute human study presented here and their technical nective tissue [40–42] that has higher stiffness and strain
implementation and in vivo validation were previously in the longitudinal than the transverse direction [43].
described [17]. This implies that the needle and the thin guiding fila-
Frequent breakage of the thin-film electrodes during ment may have faced higher mechanical resistance in this
their implantation was one of the most important limita- insertion scenario than in the typical scenario of use for
tions of the present study. In contrast to previous studies this kind of thin-film electrodes (i.e., perpendicular to the
(upper limb [1, 35], lower limb [36–38], neck [36, 39], and muscle). In the case of the lower limbs the conditions are
hand [36, 38] muscles) in which very similar thin-film more detrimental, as the fascia (e.g., crural fascia in the
electrodes were reproducibly implanted transversally to lower leg) is thicker and stiffer because it helps stabilize
and/or shallower into the muscle, here the electrodes had the limb during locomotion [44], and lower limb muscles
to be deeply inserted longitudinally to the limb to guar- are stiffer than upper limb muscles, making it even more
antee (1) that both the distal and proximal electrodes (‘A’ difficult to insert the dedicated needle. It is worth not-
and ‘B’ respectively in Fig. 1c) were inside muscle tissue, ing that the crural fascia in the anterior compartment
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 12 of 16

Fig. 8 Parametric acquisition using ZC rate. Participant 2 did three contractions during 6 s (progressive contraction followed by relaxation), which can be
clearly seen by the parametric waveform obtained. After the samples were uploaded to the external system, a moving average filter was added to show
the envelope of the contractions

(tibialis anterior muscle) is stiffer than in the posterior method based on volume conduction. However, it must
compartment (gastrocnemius medialis muscle) [45]. This be noted that the designed communication protocol
could explain why it was more difficult to implant in the allows the control of up to 256 devices per low-level con-
tibialis anterior than in the gastrocnemius muscles. It has trol unit, and the possibility to have several low-level
been also experimentally determined that the stiffness of units connected to a high-level controller. In fact, we have
the gastrocnemius lateralis muscle is higher than that of in vitro demonstrated that more than 10 floating devices
gastrocnemius medialis muscle, which could explain why can be wirelessly powered and digitally controlled from
it was easier to implant the device in the gastrocnemius the external system [50].
medialis muscle [46, 47]. In terms of the results obtained for bidirectional com-
Even though this thin-film technology has been munications, it is worth noting that the demodulation
implanted in animals for recording purposes for up to performed by the external system consists in a rather
9 weeks [48], and in humans for electrical stimulation simple hardware-based approach: the output of the ana-
studies for up to 11 weeks [49], it must be noted that log amplitude demodulator is interfaced with the UART
the devices presented here are proof-of-concept proto- of the digital unit using an analog comparator. This
types that demonstrate the circuit architecture and the approach has several limitations, including fixed thresh-
capabilities of the proposed approach in an acute study. olds and the impossibility to postprocess the digital data
They are not the final proposed implantable devices to be to recover lost bits. The limitations hinder the control
chronically implanted. The electronics are meant to be unit’s decoding performance in certain scenarios, which
integrated in the future into an ASIC to be included in a translates into very low success rates. For example, as
flexible threadlike implant with two electrodes at oppo- the electric potential across the thin-film electrodes
site ends as that shown in [22]. In other words, the limita- decreases, the modulation index seen from the exter-
tions faced with the thin-film electrodes will not be faced nal system during uplink decreases too. This can be
with a conformation based on a tubular body mostly improved by replacing the current approach used in the
made of silicone. demodulator of the external system (i.e., analog compara-
In this first-in-human study a single floating device was tor plus UART) with a more flexible approach based on
used per participant. This limitation was mainly due to digital processing: digital filters, comparators with adap-
timing restrictions in the hospitals where the study was tive thresholds and specific decoding algorithms. This
performed. It was considered sufficient to inject only one could increase the success rate for uplink in unfavorable
wireless device for EMG sensing and electrical stimu- cases where low SNR is obtained.
lation to demonstrate the WPT and communication
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 13 of 16

The muscle contractions obtained during electrical possible to substantially reduce the number of bytes to be
stimulation were weak compared to those obtained in transmitted, thus allowing real-time control. We report
previous electrical stimulation studies using this WPT an example using the rate of ZC, which is a basic para-
approach [17, 22]. Because of the limitations explained metric measurement obtained in the time domain. How-
above related to the frequent breakage of the thin-film ever, the communication protocol created allows us to
electrodes when implanting the floating device, dur- use a second parametric measurement that can be imple-
ing the procedure it was prioritized the alignment with mented in the firmware of the floating devices or in an
respect to the applied electric field and the settling in a ASIC (e.g., RMS, spike properties [55]). Using paramet-
single tissue over electrical stimulation (i.e., powering ric measurements, it is possible to acquire much more
and bidirectional communications were deemed more samples, store them in the wireless device’s memory,
important in the study). and be more efficient with the powering/communication
One limitation of using the same electrodes for power- channel to accomplish a BHNS system that can oper-
ing, electrical stimulation and EMG sensing in the wire- ate in real time to process and analyze the information
less device is that the HF current bursts applied by the obtained using the intramuscular electrodes, and activate
external system saturate its AFE, creating an artifact dur- the electrical stimulation or control external machines
ing EMG sensing. This artifact is equivalent to that gen- such as exoskeletons. By configuring the window size of
erated when electrical stimulation and EMG recording is these parametric measurements using the external sys-
done simultaneously in research and commercial systems tem and its bidirectional communication protocol, the
[51]. Several applications report this, including closed- measurements can be adapted to different applications
loop myoelectric control using electrotactile stimulation (e.g., selective and adaptive timely stimulation (SATS) for
[52], tremor suppression using EMG and electrical stim- tremor reduction using RMS at a 10 ms or 15 ms window
ulation [53], and EMG acquisition during transcutaneous size [1, 56], and autoencoding (AEN) of EMG for myo-
spinal cord electrical stimulation [54]. Strategies to avoid control using RMS at a 100 ms window size [57]).
stimulation artifacts include using sequential opera- The proof-of-concept floating AIMD described in
tions in which EMG recording windows are separated depth in [17] and demonstrated here in the upper and
from electrical stimulation windows [1, 53], and software lower limb of healthy participants cannot be used for
blanking [31]. The wireless device reported here uses a chronic studies of more than 80 days, as the thin-film
similar approach for the saturation artifacts: there are intramuscular electrodes may fail [58] and the skin access
short windows in which the EMG samples acquired do is prone to infections in the long term. However, the
not represent muscle activation as the HF current bursts architecture of the miniature electronic circuit connected
generate a saturation artifact, and so, these samples are to these intramuscular electrodes can be integrated into
blanked replacing them with a constant and known an ASIC, as it is composed of typical electronic compo-
value. This short, saturated window is followed by a long nents that can be fabricated using current microelec-
non-saturated EMG recording window, and the entire tronic technologies and manufacturing processes. This
sequence is acquired at the sampling frequency previ- would allow to obtain fully implantable flexible thread-
ously defined from the external system. like EMG sensors and stimulators that can be easily
The assays performed in the second participant were implanted by injection, that can be employed in chronic
focused on transmitting raw EMG samples obtained by studies and developed for future commercial use. We
the floating device. However, it must be noted that raw recently demonstrated in the hindlimbs of two anesthe-
EMG acquisition would not allow real-time control as the tized rabbits that fully injectable microstimulators (over-
proof-of-concept floating device has a limited memory all diameter of 0.97 mm) are able to rectify these volume
capacity of 2 kB, and all the samples must be uploaded to conducted HF bursts to cause local low frequency cur-
the external system, occupying the entire physical chan- rents capable of stimulation [22]. The microstimulators
nel during several milliseconds (approximately 1.35 ms include a hermetic capsule that houses an ASIC, two dc-
are required to upload one sample). This would limit the blocking capacitors and a capacitance for voltage stabili-
potential of the BHNS concept. It is worth noting that a zation. The implantation procedure proposed, which uses
similar uplink limitation is faced by other implants that a 14 G clinical catheter, is described in depth in [59].
do EMG acquisition (the IST-12 can upload at most 1
sample every millisecond [24]). The upload time could be Conclusions
decreased in the communication protocol by, for exam- This brief report presents the first-in-human demon-
ple, decreasing the duration of the power maintenance stration of the use of coupling by volume conduction as
burst that is sent prior to the uplink transmission burst a WPT and bidirectional communications method to
(more information in Additional file 1 of [17]). Also, by operate networks of wireless EMG sensors and stimula-
parametrizing the EMG signal in the floating devices it is tors. The proof-of-concept floating devices, consisting of
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 14 of 16

Author contributions
thin-film injectable intramuscular electrodes connected LBF, JM, MGS, AMG, FOB and AI conceived and designed the study. LBF, JM, CR
to a miniature electronic circuit, were powered by and and AI developed the electronic hardware and the software. MOK, AS and AI
communicated with an external system that delivered designed and developed the intramuscular electrodes. LBF, JM, CR, MGS, AMG,
CRG, FGH, FOB and AI conducted the experiment. LBF and JM performed data
HF current in the form of bursts via two external textile analysis and drafted the manuscript. MOK, CR, MGS, AMG, CRG, FGH, AC, AJA,
electrodes strapped around the limb where the device AGA, FG, AS, FOB and AI revised the manuscript critically. All authors read and
was deployed. In one participant out of six it was dem- approved the final manuscript.

onstrated that the external system can send digital com- Funding
mands to configure the EMG sensing and electrical This work has received funding from the European Union’s Horizon 2020
stimulation parameters, to trigger the acquisition or the research and innovation programme under grant agreement No. 779982
(Project EXTEND - Bidirectional Hyper-Connected Neural System). CR has
stimulation, and to upload samples. Strategies are pro- been also partially funded by CSIC Interdisciplinary Thematic Platform (PTI+)
posed to improve the overall performance of the system, NEURO-AGINGl+ (PTI-NEURO-AGING+). FOB thanks the financial support from
overcoming the limitations faced in the study regarding the Spanish MCIN/AEI/https://doi.org/10.13039/501100011033 and by the
“European Union NextGenerationEU/PRTR” under Grant agreement IJC2020-
uplink speed and decodification, and memory usage. 044467-I. AI gratefully acknowledges the financial support by ICREA under the
Even though the thin-film electrodes presented issues ICREA Academia programme.
related to the need to insert them longitudinally to the
Data Availability
limb, facing high mechanical resistance that led to break- The datasets used and/or analyzed during the current study are available from
age, it is worth noting that these limitations are inherent the corresponding author on reasonable request.
of these proof-of-concept prototypes. The demonstrated
miniature electronic circuit will be integrated in an ASIC Declarations
that will be included in a tubular implant with electrodes
Ethics approval and consent to participate
at opposite ends. The implantation procedure of this The procedures were conducted in accordance with the Declaration of
tubular conformation has been demonstrated to be suc- Helsinki and approved by the local Ethics Committees of the Movement
cessful in several assays, including those reported in [22]. Disorders Clinic of Gregorio Marañón Hospital – Madrid – and the
Biomechanics and Assistive Technology Unit of National Hospital for
This first-in-human demonstration opens the possibility Paraplegics – Toledo – (reference numbers 18/2020 and 565 respectively), as
of using coupling by volume conduction in networked well as by the Spanish Agency of Medicines and Medical Devices (AEMPS) –
neuroprosthetic systems. records 858/20/EC and 856/20/EC respectively. The participants volunteered
to participate in this study, were informed about the procedures and possible
List of Abbreviations adverse effects, and signed the informed consent to participate.
AIMDs Active implantable medical devices
EMG Electromyography Consent for publication
HMI Human-machine interface Not applicable.
BHNS Bidirectional Hyper-Connected Neural Systems
WPT Wireless power transfer Competing interests
HF High frequency The authors declare no competing interests.
ASIC Application-specific integrated circuit
ACK Acknowledge Author details
1
RAM Random-access memory Department of Information and Communications Technologies,
AFE Analog front-end Universitat Pompeu Fabra, Barcelona 08018, Spain
2
SNR Signal to noise ratio Research Centre for Information and Communications Technologies,
CMRR Common mode rejection ratio University of Granada, Granada 18014, Spain
3
ADC Analog-to-digital converter Department of Signal Theory, Telematics and Communications,
AEMPS Spanish Agency of Medicines and Medical Devices University of Granada, Granada 18014, Spain
4
UART Universal asynchronous receiver-transmitter Fraunhofer Institute for Biomedical Engineering IBMT, 66280 Sulzbach,
RMS Root mean square Germany
5
ZC Zero crossings Neural Rehabilitation Group, Cajal Institute, Spanish National Research
Council (CSIC), Madrid 28002, Spain
6
Systems Engineering and Automation Department, Carlos III University
of Madrid, Madrid 28903, Spain
Supplementary Information 7
Movement Disorders Unit, Department of Neurology, Hospital General
The online version contains supplementary material available at https://doi. Universitario Gregorio Marañón, Madrid 28007, Spain
8
org/10.1186/s12984-023-01295-5. Biomechanics and Assistive Technology Unit, National Hospital for
Paraplegics. Unit of Neurorehabilitation, Biomechanics and Sensory-
Additional file 1: Video of implantation procedure in biceps brachii of Motor Function (HNP-SESCAM), Unit associated to the CSIC, Toledo, Spain
9
participant 2 and gastrocnemius medialis of participant 5 School of Science and Technology, Department of Applied Mathematics,
Materials Science and Engineering and Electronic Technology, Rey Juan
Additional file 2: Video of electrical stimulation sequence in biceps brachii Carlos University, Móstoles 28933, Spain
of participant 2 10
CSIC’s Associated RDI Unit ‘Unidad De Neurorehabilitación, Biomecánica
Y Función Sensitivo-Motora’, Madrid, Spain
11
Serra Húnter Fellow Programme, Universitat Pompeu Fabra,
Acknowledgements Barcelona 08018, Spain
The authors would like to express their gratitude to the participants of this
study, and to Alejandro Pascual-Valdunciel and Juan Camilo Moreno for their Received: 3 October 2023 / Accepted: 12 December 2023
support regarding the documentation for ethical committees and the AEMPS.
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 15 of 16

References recording and stimulation. In: 2015 37th Annual International Conference of
1. Pascual-Valdunciel A, Gonzalez-Sanchez M, Muceli S, Adan-Barrientos B, the IEEE Engineering in Medicine and Biology Society (EMBC) [Internet]. IEEE;
Escobar-Segura V, Perez-Sanchez JR, et al. Intramuscular stimulation of 2015 [cited 2018 Jun 21]. p. 7135–8. Available from: http://ieeexplore.ieee.
muscle afferents attains prolonged tremor reduction in essential tremor org/document/7320037/.
patients. IEEE Trans Biomed Eng. 2020;68(6):1768–76. 24. Morel P, Ferrea E, Taghizadeh-Sarshouri B, Audí JMC, Ruff R, Hoffmann KP, et
2. Koller JR, Jacobs DA, Ferris DP, Remy CD. Learning to walk with an adaptive al. Long-term decoding of movement force and direction with a wireless
gain proportional myoelectric controller for a robotic ankle exoskeleton. J myoelectric implant. J Neural Eng. 2016;13(1):016002.
Neuroeng Rehabil. 2015;12(1):1–14. 25. Weir RF, Troyk PR, DeMichele GA, Kerns DA, Schorsch JF, Maas H. Implantable
3. Salminger S, Sturma A, Hofer C, Evangelista M, Perrin M, Bergmeister KD, Myoelectric sensors (IMESs) for intramuscular Electromyogram Recording.
et al. Long-term implant of intramuscular sensors and nerve transfers IEEE Trans Biomed Eng. 2009;56(1):159–71.
for wireless control of robotic arms in above-elbow amputees. Sci Rob. 26. Hart RL, Bhadra N, Montague FW, Kilgore KL, Peckham PH. Design and testing
2019;4(32):eaaw6306. of an Advanced Implantable Neuroprosthesis with Myoelectric Control. IEEE
4. Fleming A, Stafford N, Huang S, Hu X, Ferris DP, Huang HH. Myoelectric Trans Neural Syst Rehabil Eng. 2011;19(1):45–53.
control of robotic lower limb prostheses: a review of electromyography 27. McDonnall D, Hiatt S, Smith C, Guillory KS. Implantable multichannel wire-
interfaces, control paradigms, challenges and future directions. J Neural Eng. less electromyography for prosthesis control. In: 2012 Annual International
2021;18(4):41004. Conference of the IEEE Engineering in Medicine and Biology Society
5. Jung MK, Muceli S, Rodrigues C, Megía-García Á, Pascual-Valdunciel A, del- [Internet]. IEEE; 2012. p. 1350–3. Available from: http://ieeexplore.ieee.org/
Ama AJ, et al. Intramuscular EMG-Driven Musculoskeletal Modelling: towards document/6346188/.
implanted muscle interfacing in spinal cord Injury patients. IEEE Trans 28. Wahyunggoro O, Nugroho H. Performance analysis of the Windowing
Biomed Eng. 2022;69(1):63–74. technique on Elbow Joint Angle Estimation using Electromyography Signal.
6. Young AJ, Hargrove LJ, Kuiken TA. The effects of electrode size and orienta- In IOP Publishing; 2018. p. 012004.
tion on the sensitivity of myoelectric pattern recognition systems to elec- 29. Zardoshti-Kermani M, Wheeler BC, Badie K, Hashemi RM. EMG feature
trode shift. IEEE Trans Biomed Eng. 2011;58(9):2537–44. evaluation for movement control of upper extremity prostheses. IEEE Trans
7. Dewald HA, Lukyanenko P, Lambrecht JM, Anderson JR, Tyler DJ, Kirsch RF, Rehabilitation Eng. 1995;3(4):324–33.
et al. Stable, three degree-of-freedom myoelectric prosthetic control via 30. Kim T, Kim J, Koo B, Jung H, Nam Y, Chang Y, et al. Effects of sampling rate and
chronic bipolar intramuscular electrodes: a case study. J Neuroeng Rehabil. window length on Motion Recognition using sEMG Armband Module. Int J
2019;16(1):147. Precis Eng Manuf. 2021;22(8):1401–11.
8. Kilgore KL, Anderson KD, Peckham PH. Neuroprosthesis for individuals with 31. Hartmann C, Došen S, Amsuess S, Farina D. Closed-Loop Control of Myoelec-
spinal cord injury. Neurol Res. 2020;1–13. tric Prostheses with Electrotactile Feedback: influence of Stimulation Artifact
9. Memberg WD, Polasek KH, Hart RL, Bryden AM, Kilgore KL, Nemunaitis GA, et and Blanking. IEEE Trans Neural Syst Rehabil Eng. 2015;23(5):807–16.
al. Implanted neuroprosthesis for restoring arm and hand function in people 32. Cogan SF. Neural stimulation and Recording electrodes. Annu Rev Biomed
with high level tetraplegia. Arch Phys Med Rehabil. 2014;95(6):1201–1211e1. Eng. 2008;10(1):275–309.
10. Dinis H, Mendes PM. A comprehensive review of powering methods used in 33. Hägg G. Electromyographic fatigue analysis based on the number of zero
state-of-the-art miniaturized implantable electronic devices. Biosens Bioelec- crossings. Pflügers Archiv. 1981;391(1):78–80.
tron. 2021;172:112781. 34. Clancy EA, Negro F, Farina D. Single-channel techniques for information
11. Kane MJ, Breen PP, Quondamatteo F, ÓLaighin G. BION microstimulators: a extraction from the surface EMG signal. Surface electromyography: physiol-
case study in the engineering of an electronic implantable medical device. ogy, engineering, and applications. 2016;91–125.
Med Eng Phys. 2011;33(1):7–16. 35. Muceli S, Poppendieck W, Hoffmann KP, Dosen S, Benito-León J, Barroso FO,
12. Merrill DR, Lockhart J, Troyk PR, Weir RF, Hankin DL. Development of an et al. A thin-film multichannel electrode for muscle recording and stimulation
implantable myoelectric sensor for advanced prosthesis control. Artif Organs. in neuroprosthetics applications. J Neural Eng. 2019;16(2):026035.
2011;35(3):249–52. 36. Muceli S, Poppendieck W, Negro F, Yoshida K, Hoffmann KP, Butler JE, et al.
13. Schulman JH. The Feasible FES System: Battery Powered BION Stimulator. Accurate and representative decoding of the neural drive to muscles in
Proceedings of the IEEE. 2008;96(7):1226–39. humans with multi-channel intramuscular thin-film electrodes. J Physiol.
14. Barbruni GL, Ros PM, Demarchi D, Carrara S, Ghezzi D. Miniaturised Wireless 2015;593(17):3789–804.
Power Transfer Systems for Neurostimulation: a review. IEEE Trans Biomed 37. Muceli S, Poppendieck W, Holobar A, Gandevia S, Liebetanz D, Farina D.
Circuits Syst. 2020;14(6):1160–78. Accurate decoding of the spinal cord output in humans with implanted
15. Turner BL, Senevirathne S, Kilgour K, McArt D, Biggs M, Menegatti S, et al. high-density electrode arrays. bioRxiv. 2022.
Ultrasound-powered implants: a critical review of Piezoelectric Material 38. Negro F, Muceli S, Castronovo AM, Holobar A, Farina D. Multi-channel
Selection and Applications. Adv Healthc Mater. 2021;10(17):2100986. intramuscular and surface EMG decomposition by convolutive blind source
16. Agarwal K, Jegadeesan R, Guo YX, Thakor NV. Wireless Power Transfer Strate- separation. J Neural Eng. 2016;13(2):26027.
gies for Implantable Bioelectronics. IEEE Rev Biomed Eng. 2017;10:136–61. 39. Luu BL, Muceli S, Saboisky JP, Farina D, Héroux ME, Bilston LE, et al. Motor unit
17. Becerra-Fajardo L, Krob MO, Minguillon J, Rodrigues C, Welsch C, Tudela-Pi M territories in human genioglossus estimated with multichannel intramuscular
et al. Floating EMG sensors and stimulators wirelessly powered and operated electrodes. J Appl Physiol. 2018;124(3):664–71.
by Volume Conduction for Networked Neuroprosthetics. J Neuroeng Rehabil. 40. Pancheri FQ, Eng CM, Lieberman DE, Biewener AA, Dorfmann L. A constitu-
2022;19(57). tive description of the anisotropic response of the fascia lata. J Mech Behav
18. Ivorra A. Remote electrical stimulation by means of implanted rectifiers. PLoS Biomed Mater. 2014;30:306–23.
ONE. 2011;6(8):e23456. 41. Sednieva Y, Viste A, Naaim A, Bruyère-Garnier K, Gras LL. Strain assessment of
19. Ivorra A, Becerra-Fajardo L, Castellví Q. In vivo demonstration of injectable deep fascia of the thigh during leg movement: an in situ study. Front Bioeng
microstimulators based on charge-balanced rectification of epidermically Biotechnol. 2020;8:750.
applied currents. J Neural Eng. 2015;12(6):66010. 42. Stecco C, Pavan PG, Porzionato A, Macchi V, Lancerotto L, Carniel EL, et al.
20. Tudela-Pi M, Minguillon J, Becerra-Fajardo L, Ivorra A. Volume Conduction for Mechanics of crural fascia: from anatomy to constitutive modelling. Surg
Powering Deeply Implanted Networks of Wireless Injectable Medical Devices: Radiol Anat. 2009;31(7):523–9.
a Numerical Parametric Analysis. IEEE Access. 2021;9:100594–605. 43. Eng CM, Pancheri FQ, Lieberman DE, Biewener AA, Dorfmann L. Directional
21. Minguillon J, Tudela-Pi M, Becerra-Fajardo L, Perera-Bel E, Ama AJ, del-, differences in the biaxial material properties of fascia lata and the implica-
Gil-Agudo A, et al. Powering electronic implants by high frequency volume tions for fascia function. Ann Biomed Eng. 2014;42(6):1224–37.
conduction: in human validation. IEEE Trans Biomed Eng. 2023;70(2):659–70. 44. Stahl VA. A biomechanical analysis of the role of the crural fascia in the cat
22. García-Moreno A, Comerma-Montells A, Tudela-Pi M, Minguillon J, Becerra- hindlimb. Dissertation P, editor. Georgia Institute of Technology and Emory
Fajardo L, Ivorra A. Wireless networks of injectable microelectronic stimula- University; 2010.
tors based on rectification of volume conducted high frequency currents. J 45. Stecco C, Pavan P, Pachera P, De Caro R, Natali A. Investigation of the
Neural Eng. 2022;19(5):056015. mechanical properties of the human crural fascia and their possible clinical
23. Poppendieck W, Muceli S, Dideriksen J, Rocon E, Pons JL, Farina D et al. A implications. Surg Radiol Anat. 2014;36(1):25–32.
new generation of double-sided intramuscular electrodes for multi-channel
Becerra-Fajardo et al. Journal of NeuroEngineering and Rehabilitation (2024) 21:4 Page 16 of 16

46. Huang J, Qin K, Tang C, Zhu Y, Klein CS, Zhang Z, et al. Assessment of passive 54. Kim M, Moon Y, Hunt J, McKenzie KA, Horin A, McGuire M, et al. A novel
stiffness of medial and lateral heads of gastrocnemius muscle, Achilles technique to reject Artifact Components for Surface EMG signals recorded
tendon, and plantar fascia at different ankle and knee positions using the during walking with Transcutaneous spinal cord stimulation: a pilot study.
MyotonPRO. Med Sci Monitor: Int Med J Experimental Clin Res. 2018;24:7570. Front Hum Neurosci. 2021;15:660583.
47. Lall PS, Alsubiheen AM, Aldaihan MM, Lee H. Differences in medial and lateral 55. Gabriel DA. Reliability of SEMG spike parameters during concentric. Electro-
gastrocnemius stiffness after Exercise-Induced muscle fatigue. Int J Environ myogr Clin Neurophysiol. 2000;40:423–30.
Res Public Health. 2022;19(21). 56. Pascual-Valdunciel A, Kurukuti NM, Montero-Pardo C, Barroso FO, Pons JL.
48. Lewis S, Russold M, Dietl H, Ruff R, Audí JMC, Hoffmann KP, et al. Fully Implant- Modulation of spinal circuits following phase-dependent electrical stimula-
able Multi-channel Measurement System for Acquisition of muscle activity. tion of afferent pathways. J Neural Eng. 2023;20(1):016033.
IEEE Trans Instrum Meas. 2013;62(7):1972–81. 57. Vujaklija I, Shalchyan V, Kamavuako EN, Jiang N, Marateb HR, Farina D. Online
49. Zollo L, Di Pino G, Ciancio AL, Ranieri F, Cordella F, Gentile C, et al. Restor- mapping of EMG signals into kinematics by autoencoding. J Neuroeng
ing tactile sensations via neural interfaces for real-time force-and-slippage Rehabil. 2018;15(1):21.
closed-loop control of bionic hands. Sci Rob. 2019;4(27):eaau9924. 58. Oldroyd P, Malliaras GG. Achieving long-term stability of thin-film electrodes
50. Becerra-Fajardo L, Minguillon J, Comerma A, Ivorra A. Networks of Inject- for neurostimulation. Acta Biomater. 2022;139:65–81.
able Microdevices Powered and Digitally Linked by Volume Conduction 59. Malik S, Castellvi Q, Becerra-Fajardo L, Pi MT, Garcia-Moreno A, Baghini MS,
for Neuroprosthetics: a Proof-of-Concept. In: 11th International IEEE/EMBS et al. Injectable sensors based on Passive rectification of volume-conducted
Conference on Neural Engineering (NER), Baltimore, MD, USA, 2023. 2023. currents. IEEE Trans Biomed Circuits Syst. 2020;14(4):867–78.
51. Frigo C, Ferrarin M, Frasson W, Pavan E, Thorsen R. EMG signals detection and
processing for on-line control of functional electrical stimulation. J Electro-
myogr Kinesiol. 2000;10(5):351–60. Publisher’s Note
52. Garenfeld MA, Jorgovanovic N, Ilic V, Strbac M, Isakovic M, Dideriksen JL, et Springer Nature remains neutral with regard to jurisdictional claims in
al. A compact system for simultaneous stimulation and recording for closed- published maps and institutional affiliations.
loop myoelectric control. J Neuroeng Rehabil. 2021;18(1):87.
53. Dosen S, Muceli S, Dideriksen JL, Romero JP, Rocon E, Pons J, et al. Online
tremor suppression using Electromyography and Low-Level Electrical Stimu-
lation. IEEE Trans Neural Syst Rehabil Eng. 2015;23(3):385–95.