Content uploaded by Simon Borgognon
Author content
All content in this area was uploaded by Simon Borgognon on Nov 16, 2021
Content may be subject to copyright.
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
1 of 15
PARALYSIS
Intrafascicular peripheral nerve stimulation produces
fine functional hand movements in primates
Marion Badi1†, Sophie Wurth1†, Ilaria Scarpato1, Evgenia Roussinova1, Elena Losanno2,
Andrew Bogaard3, Maude Delacombaz3, Simon Borgognon3,4, Paul C
̆vanc̆ara5, Florian Fallegger6,
David K. Su7, Eric Schmidlin3, Grégoire Courtine4,8, Jocelyne Bloch8, Stéphanie P. Lacour6,
Thomas Stieglitz5, Eric M. Rouiller3, Marco Capogrosso3,9, Silvestro Micera1,2*
Restoring dexterous hand control is critical for people with paralysis. Approaches based on surface or intramuscular
stimulation provide limited finger control, generate insufficient force to recover functional movements, and
require numerous electrodes. Here, we show that intrafascicular peripheral electrodes could produce functional
grasps and sustained forces in three monkeys. We designed an intrafascicular implantable electrode targeting the
motor fibers of the median and radial nerves. Our interface selectively and reliably activated extrinsic and intrinsic
hand muscles, generating multiple functional grips, hand opening, and sustained contraction forces for up to
2 months. We extended those results to a behaving monkey with transient hand paralysis and used intracortical
signals to control simple stimulation protocols that enabled this animal to perform a functional grasping task. Our
findings show that just two intrafascicular electrodes can generate a rich portfolio of dexterous and functional
hand movements with important implications for clinical applicability.
INTRODUCTION
Functional neuromuscular electrical stimulation (NMES) delivered
via surface electrodes (1–3) or epimysial implants (4,5) has been
used for decades to restore grasping after paralysis due to spinal
cord injury or stroke. Although surface stimulation systems permit
the generation of a few predetermined grasp types, they are limited
by high transcutaneous activation currents introducing muscle
fatigue, poor selectivity, and the difficulty to recruit intrinsic hand
muscles. The use of intramuscular electrodes has allowed us to
target deeper muscular structures and restore additional hand func-
tions in monkeys (6,7). Such implantable NMES-based neuropros-
theses have been used in clinical studies (4,8–10). However, these
systems rely on invasive surgical procedures and require the im-
plantation of numerous muscles whose operative access can be
challenging. Therefore, the feasibility of upscaling these technologies
to large populations of patients, higher number of muscles, and,
thus, functions seems daunting. Moreover, all NMES approaches
are limited by the generation of muscle fatigue due to the inverse,
“large-to-small” recruitment order of efferent nerve fibers (11,12).
This unnatural activation results in a preferential engagement of
fast-fatiguing large-diameter muscles, which affects the forces that
can be produced by these technologies.
An alternative strategy to recruit muscles consists of targeting
branches of the peripheral nervous system (PNS). The architecture
of the PNS allows accessing several distal end-effectors from a single
proximal location, consequently bypassing the placement of mul-
tiple electrodes. For instance, epineural cuff electrodes have been
implemented in clinical neuroprostheses aiming at restoring hand
function after paralysis (13–15). However, because the contacts lie
on the surface of the nerve and thereby far from the individual nerve
fibers, epineural interfaces are prone to inverse efferent recruitment
and require relatively large currents of stimulation, leading to the
undesired activation of the deepest nerve structures (16). Intra-
fascicular stimulation offers an alternative opportunity to enhance
recruitment selectivity (17) by positioning the implant within mul-
tiple nerve fascicles near efferent axons of different muscles (18–21).
The proximity of the implant’s active sites with motor fibers reduces
the influence of the motor fiber diameter on the selective axon re-
cruitment, thereby minimizing the inverse recruitment observed
with extra neural stimulation and thus reducing the onset of fatigue
(18,19,22). However, to date, to our knowledge, no intraneural
stimulation–based neuroprosthesis for hand functional restoration
has been implemented in clinical applications.
Recent studies demonstrated the preeminent potential of intraneural
transverse intrafascicular multichannel electrodes (TIMEs) (23) to elic-
it selective sensory percepts in subjects with limb amputation, with
promising results in terms of long-term usability in humans. Moreover,
TIME implants have been used to generate specific motor responses
and permit a high-fidelity control of evoked movements in rodents (24).
Building on these results, we hypothesized that intraneural stimulation
of the median and radial nerves using a TIME-like intrafascicular elec-
trode would allow the selective recruitment of extrinsic and intrinsic
hand muscles. This ability would enable the reproducible control of rel-
evant joints to produce fine wrist and finger movements, thereby open-
ing the opportunity of restoring functionality after hand paralysis.
1Bertarelli Foundation Chair in Translational Neuroengineering, Center for Neuro-
prosthetics, and Institute of Bioengineering, École Polytechnique Fédérale de Lausann e
(EPFL), 1015 Lausanne, Switzerland. 2Biorobotics Institute and Department of
Excellence in Robotics and AI, Scuola Superiore Sant’Anna, 56025 Pisa, Italy. 3De-
partment of Neuroscience and Movement Sciences, Platform of Translational
Neurosciences, Section of Medicine, Faculty of Sciences and Medicine, University
of Fribourg, 1700 Fribourg, Switzerland. 4Center for Neuroprosthetics and Brain Mind
Institute, School of Life Sciences, EPFL, 1015 Lausanne, Switzerland. 5Laboratory for
Biomedical Microtechnology, Department of Microsystems Engineering–IMTEK,
Bernstein Center Freiburg, and BrainLinks-BrainTools Center, University of Freiburg,
79110 Freiburg, Germany. 6Bertarelli Foundation Chair in Neuroprosthetic Tech-
nology, Laboratory for Soft Bioelectronics Interface, Institute of Microengineering,
Institute of Bioengineering, Centre for Neuroprosthetics, 1202 Geneva, Switzerland.
7Neurological Surgery, Harborview Medical Center, Seattle, WA 98104, USA. 8Defitech
Center for Interventional Neurotherapies (NeuroRestore), EPFL, University Hospital
of Lausanne (CHUV), and University of Lausanne (UNIL), 1015 Lausanne, Switzerland.
9Department of Neurological Surgery, Rehabilitation and Neural Engineering Lab-
oratories, University of Pittsburgh, Pittsburgh, PA 15213, USA.
*Corresponding author. Email: silvestro.micera@epfl.ch
†These authors contributed equally to this work.
Copyright © 2021
The Authors, some
rights reserved;
exclusive licensee
American Association
for the Advancement
of Science. No claim
to original U.S.
Government Works
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
2 of 15
RESULTS
Design of the Mk-TIME: A tailored intrafascicular interface
for arm nerves in monkeys
Selective intrafascicular stimulation relies on the identification of
an optimal implantation site and a tailored implant design to target
multiple different fascicles containing motor fibers (25). Therefore,
we performed a detailed anatomical study of monkeys’ median and
radial nerves that constitute the main innervation routes for hand
flexor and extensor muscles, respectively (Fig.1A and fig. S1) (26).
Dissection of monkey arms (n=3 animals) enabled us to precisely
identify the branching points of the motor fibers innervating these
muscles (Fig.1A and fig. S2A). In all monkeys, the median and
radial ramifications innervating muscles of the wrist and the hand
emerged from the main nerve trunk distal to the elbow (Fig.1A and
fig. S2A). To determine the optimal implantation site for the
Mk-TIME above the elbow, we quantified the number and cross-
sectional area of the fascicles constituting the nerve at different levels,
labeled hereafter as cs1, cs2, and cs3, proximal to the elbow (n=6
animals; Fig.1,AandB, and fig. S1). Our analysis showed the pres-
ence of one to two larger-diameter fascicles accompanied by one
or several smaller-diameter fascicles in nerve segments harvested
3cm proximally to the epicondyle (cs1 segment). More distal nerve
portions, namely, cs2 and cs3, were characterized by a gradually in-
creasing number of progressively smaller fascicles (Fig.1B). These
anatomical features were consistent across monkeys (n=6 animals)
in both the median and radial nerves (Fig.1B and fig. S2, B to D).
Immunohistochemical labeling of choline acetyltransferase (ChAT)–
positive motor axons proximal to the elbow revealed a heterogeneous
distribution of efferent fibers within each of the larger fascicles of
the median and radial nerves, confirming the mixed modality of
these compartments (27). Motor fibers were found to be organized
in patches, dispersed throughout many fascicles, and some were
located deep within the fascicular compartments (Fig.1C). This
miscellaneous distribution of motor fibers was observed in all
three animals analyzed, in both the median and radial nerves
(fig. S2, E to I).
We used these findings and implemented an anatomically realistic
computational model of the nerve (28,29) to simulate the recruit-
ment of axons elicited by intraneural stimulation (fig. S3A). Because
of the anatomical similarities observed between the median and ra-
dial nerve segments proximal to the elbow (Fig.1B and fig. S1), we
designed a single generic model based on median nerve histology.
This model combined a volume conductor model of the nerve and
a biophysical model of motor fibers (fig. S3A) (28,29). First, we
calculated voltage distributions elicited by currents injected through
single contacts of the Mk-TIME. We then estimated which axons
were recruited by a given current amplitude (see Materials and
Methods and Supplementary Methods). We simulated multiple
possible implantation locations along the proximal-distal axis of the
nerve (cs1 to cs3), showing that implantation at a location where
efferent fibers were segregated in different fascicles led to more selec-
tive recruitment than stimulating densely packed fascicles (Fig.2A
and fig. S3, B and C). Charge injection at the most proximal level,
cs1, resulted in the simultaneous recruitment of most of the axons
and a small number of selectively recruited muscles. Intraneural
stimulation delivered at the cs2 level engaged the underlying efferent
fibers more selectively and led to a higher proportion of selectively
recruited muscles (Fig.2A and fig. S3D). Charge injected at the
most distal level, cs3, did not induce more specific recruitment of
motor fibers than at level cs2, despite the motor fibers corresponding
to different muscles being distributed in distinct fascicles (Fig.2A
and fig. S3, D and E). This phenomenon results from the linear
geometry of the Mk-TIME and the limited number of contacts on
each side of the active shaft that hinder the recruitment of highly
dispersed fibers. We then compared the essential properties of
extra- and intrafascicular recruitment for the particular level of im-
plantation, cs2 (Fig.2B), and showed that extrafascicular stimulation
resulted in a lower selectivity and displayed higher thresholds of acti-
vation and saturation than intrafascicular stimulation (Fig.2,CandD,
and fig. S3F).
The miscellaneous organization of motor fibers in the median
and radial fascicles together with our computational results suggest
that an intrafascicular interface penetrating deep neural structures
distributed across several fascicles can selectively recruit multiple
underlying muscles. Following these findings, we designed and
manufactured a new generation of Mk-TIME implants (see Supple-
mentary Methods) tailored to the dimensions of the monkey nerves
(Fig.2,EandF, and fig. S4) that we implanted about 2cm proximal to
the elbow (see Materials and Methods and Supplementary Methods).
The implant layout presented 16 channels of stimulation: 8 on each
side of the polyimide shaft. Anchoring tabs were placed on the
Elbow
PT
PL
FCR
Palm. Cut.
FDS
FDP
THE
LUM.
Wrist Med Cut.
FPL
PL
Palm. Cut.
Mk-Xe
Mk-Li
Mk-Ls
Wrist flexors
Finger flexors
Thumb flexors
1 cm
d
pr
ap
Median n.
Radial n.
A
cs3
cs2
cs1
500 µm
C
50 µm
Motor fiber
density HighLow
ChAT
Epicondyle
B
Endoneurial
area (%)
Individual
Fascicles
cs1
cs2
cs3
Mk-LsMk-Ce
...
cs1
cs2
cs3
1 mm
Serial cross sections
3D
recon-
struction
Median nerve
0
5
10
cs3
cs1cs2
0
1
2
3
cs
3
cs
1
cs
2
Mean endoneurial
area (mm2)Nb. fascicles
Fig. 1. Neuroanatomical organization of the median and radial nerves in monkeys .
(A) Left: Access to the median and radial nerves in the upper arm of the macaque
monkey (a, anterior; pr, proximal; p, posterior; d, distal). Right: Tracking of branches
originating from the median nerve to target flexor muscles (n = 3 animals). (B) Left:
Neural fascicular content in serial hematoxylin and eosin–stained cross sections at
levels cs1, cs2, and cs3 (3, 2, and 1 cm proximal to the epicondyle) along the nerve
and resulting 3D reconstruction of fascicular organization proximal to the elbow
(one representative animal). Middle: Number of fascicles and their respective
cross-sectional area at each level (two representative animals; different colors rep-
resent different fascicles). Right: Quantification of the number of fascicles and mean
endoneurial area per fascicle at each level (n = 6 animals). (C) Example of motor
axon distribution in the median nerve. The overlaid color map represents motor fiber
density over the cross section. Inset: Zoomed-in portion of fascicle with ChAT-labeled
motor fibers (black, no overlay). Error bars, SEMs. a.u., arbitrary units.
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
3 of 15
thin-film structure and ceramic adapter to secure the implant to the
epineurium (fig. S4).
Intrafascicular stimulation directly and selectively recruits
hand muscles
We inserted Mk-TIMEs in the median nerve and/or in the radial
nerve of five animals (13 implants in total; see Materials and Methods,
Supplementary Methods, and table S3) to characterize the modality
and selectivity of hand muscle recruitment using intrafascicular
stimulation. We recorded electromyographic (EMG) activity using
bipolar electrodes (see Materials and Methods) implanted in the
forearm and intrinsic hand muscles (Fig.3A). Monopolar stimula-
tion pulses delivered at increasing amplitude through individual
channels in the median nerve resulted in the specific activation of
wrist flexor, finger flexor, and thumb opposition muscles (Fig.3B,
left). Similarly, muscular recruitment curves computed for different
channels of the radial nerve interface showed a selective activation
of the wrist extensor and finger extensor muscles (Fig.3B, right).
Whether muscle activation due to intrafascicular stimulation
arises from direct or reflex responses (30,31), its functionality and
usability, in particular, in terms of generated force, need to be
guaranteed for a successful neuroprosthetic application. In the case
of spinal cord injury, perturbed spinal reflexes could impede the
ability of peripheral nerve stimulation to restore functional hand
movements if the stimulation-induced muscle activations arose ex-
clusively from indirect transsynaptic activations. In such a situation,
the correct relaying of the stimulation-induced action potentials via
the spinal cord would not necessarily be guaranteed. To understand
whether the intrafascicular recruitment of hand muscles is achieved
through direct recruitment of motor axons or afferent fiber–mediated
transsynaptic activation, we evaluated the time delay between the
stimulus and muscle response. The compound muscle action po-
tentials (CMAPs) recorded upon monopolar stimulation of the me-
dian and radial nerves occurred within 5 ms after the delivery of the
stimulation for extrinsic hand muscles and within 8 to 10 ms for
intrinsic hand muscles (fig. S5, A to C). The thenar eminence (THE)
also displayed a small proportion of late secondary muscle responses
with a delay of ~25 ms. Whereas the mostly short response latencies
suggest a direct activation of the corresponding muscles, longer
latencies of ~25 ms may result from transsynaptic Ia-mediated re-
sponses (30,31). To confirm the modality of muscle recruitment,
we investigated whether the CMAPs were altered by bursts of stim-
ulation at different frequencies (fig. S5, D to F). Changes in CMAP
peak-to-peak amplitude would indicate frequency-dependent atten-
uation or suppression (32) and confirm the presence of transsynaptic
recruitment of target muscles (33,34). We did not observe any
frequency-dependent CMAP suppression across different channels
(fig. S5, D to E) and only found a moderate attenuation in the evoked
CMAPs (<30%) at the highest frequencies tested (fig. S5F). Together,
these results suggest that, although intrafascicular stimulation likely
also recruits Ia afferents owing to their large diameter, the muscular
responses evoked by the Mk-TIME are primarily mediated by direct
neuronal pathways.
We then quantified the muscular selectivity obtained through
intrafascicular stimulation (see Materials and Methods) using two
different selectivity indexes (SIs) (Fig.3C and fig. S6, A to C). The
first metric, SI, reflects the normalized activation of a muscle of
interest compared to the average activation of all the other muscles
1 cm
E
Active sites
Active part
Ground
Conductive
tracks
Fixation ring
Implantation
needle
A B
Nb. of selectively
recruited muscles
Extrafasc. recruitment cs2
Intrafasc. recruitment cs2
cs2
cs1
cs3
Intraneural
electrode
Active site
Ceramic board
Fixation tabs
Fixation tabs
None
1-2 3-4
>5
28%
72%
cs2
16%
84%
cs1
cs3
54%46%
< 1%
Extra- Intra-
fascicular recruitment
0
1
Selectivity individual
muscles
050
1cs2
tnemtiurceR )u.a(
Charge (nC)
cs1
050
1
050
1cs3
Charge (nC)
050
1
050
1
Extra- Intra-
fascicular recruitment
0
1
Activation
threshold (mA)
C
D
Muscle 1
Muscle n
.
.
.
Muscle 1
Muscle n
.
.
.
F
Motor fibers
Fig. 2. Computational model of intraneural stimulation in the monkey median nerve and electrode design. (A) Left: Computationally modeled motor fiber distri-
butions at levels cs1, cs2, and cs3 of the median nerve. Different fiber populations assumed to innervate different muscles are color coded. Middle: Simulated recruitment
of motor fibers for each implantation level. Right: Quantification of the number of selectively recruited muscles (SI ≥ 0.85) over the entire simulated fiber dataset.
(B) Electric currents and motor fiber recruitment generated by simulated extrafascicular and intrafascicular stimulation at level cs2. (C) Maximal SI of individual muscles
for simulated extra- and intrafascicular recruitment of motor fibers at level cs2. (D) Activation threshold (10% recruitment) for simulated extra- and intrafascicular recruit-
ment. (E) Photograph of a microfabricated transverse intrafascicular multichannel electrode (Mk-TIME). (F) Placement of the electrode array transversally inside the nerve.
Shaded lines, SDs. Box plot distribution, median with 25th and 75th percentiles.
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
4 of 15
(see Materials and Methods). An SI of 0.2 thus indicates a target
muscle activation 20% larger than the average activation of the non-
targeted muscles. The second index, SIBrill, was computed as in the
study of Brill etal. (14) to allow a direct comparison with epineural
stimulation. This measure is more conservative and returns a null
selectivity if any of the nontargeted muscles display an activation
larger than 20%. The first index is more sensitive to measuring the
dominant muscle if there is one and ensures that a similar difference
in muscular activation leads to comparable selectivity measurements
(fig. S6A). In addition, because the SI is directly proportional to
muscular activation, it increases with high stimulating currents that
result in functional muscle activation (fig. S6B).
For both metrics used, our results show that intrafascicular stim-
ulation successfully engages specific muscles, namely, flexors of the
wrist and fingers; muscles opposing the thumb for the median nerve;
and extensors of the wrist, finger, and thumb for the radial nerve
(Fig.3C and fig. S6C). We observed that median nerve stimulation
through the Mk-TIME allowed the recruitment of intrinsic hand
muscles in five of seven implanted animals (Fig.3C and fig. S6D),
whereas epineural stimulation of the median nerve engaged those
muscles in one of six animals (Fig.3D) (14). In addition, although
the maximum selectivity values obtained through intrafascicular
stimulation were lower than the ones achieved using epineural stim-
ulation (Fig.3,CandD, and fig. S6C), the Mk-TIME selectively
recruited a larger number of muscles (Fig.3D and fig. S6, C and D).
To characterize the ability to engage a specific hand function, we
computed the functional selectivity (see Materials and Methods) of
each functional muscle group. Our results indicate that, for most of
the functional groups, muscular activity was 50% larger than the
activation of the other muscles or more (Fig.3D and fig. S6D). The
A
DorsalPalmar
FCR
PL
FDS
APB
ECR
EDC
APL
FDI
Radial nerve
Median nerve
ECU
FDP
B
C
Fingers flexion Wrist extension
0
1
16 24 32
ECR
Fingers extensionThumb opposition
OP
Mk-Lo
Median nerve stimulationRadial nerve stimulation
Thenar
Mk-Li
ECR
EDC
APL ECU
Mk-Lo
ECR
EDC
APL
0.20.6
FCR PL
FDS
THE
FDI
FDP
1.0
Mk-Ola
Mk-Le
Mk-Mec
Median nerve
stimulation
Radial nerve
stimulation
FCR PL
FDS
THE
FDI
FDP
FCR PL
FDS
THE
FDI
FDP
ECR
EDC
APL ECU
Mk-Li
Wrist flexion
Wrist extension
Fingers flexion
Fingers extension
Thumb opposition
Thumb extension Selectivity index epineural
Brill et
al.
(14)
Selectivity index
8412
FDS
0
1
Wrist flexion
8
012
1
16
FCR
PL
charge (nC)
Norm. recruitment (a.u.)
81216
APB
0
1
OP
Charge threshold for
maximum selectivity index
EDC
APL
16 24 32
0
1
29%
14%
14%
43%
0.20.6
FCR PL
FDS
THE
Pro-
-nator
FDP
1.0
D
Epineural cuff
stimulation selectivity
Number of selectively
recruited muscles
Mk 1 (left arm)
Mk 6
Mk 2
Mk 3
Mk 4
Mk 5
Mk 1 (right arm)
23 4
14%
57%
29%
Mk-TIMEEpineural
5
ch L2 ch L5ch R6 ch R1 ch L4
Fig. 3. Selective hand muscle recruitment using intrafascicular stimulation. (A) Anatomical representation of the flexor and extensor muscles of the monkey hand
innervated by the median and radial nerve, respectively. (B) Normalized muscle recruitment obtained by stimulation of the median nerve and radial nerve in one repre-
sentative animal. The gray dashed line depicts the stimulation charge at which the maximum SI was obtained for the most activated muscle. (C) Summary of selective
muscle activation for the median and radial nerve in five animals. Each plot represents the highest selectivity value for each muscle across the range of pulse amplitudes
tested. Selectivity for each muscle was thus achieved using different stimulation parameters (summarized in table S4). The dark line represents the SI (see Materials and
Methods), whereas the red mark depicts the SIBrill computed as in (14). All data were collected intraoperatively. The background shading of each polar plot separates the
muscles into functional groups, showing the ability to get functional selectivity in addition to muscle selectivity. (D) Top: Summary of selective muscle activation achieved
in (14) using a flattened epineural implant on the median nerve of six animals (seven implants). Each marker represents the highest selectivity value for each muscle across
the range of pulse amplitudes tested. Bottom: Number of selectively recruited muscles (SIBrill ≥ 0.2) for the epineural implant described in (14) and the Mk-TIME (n = 7
implants in both cases). PL, palmaris longus; FCR, flexor carpi radialis; FDS, flexor digitorum superficialis; FDP, flexor digitorum profundus; EDC, extensor digitorum com-
munis; APL, abductor pollicis longus; FDI, first dorsal interosseous; APB, abductor pollicis brevis; OP, opponens pollicis; THE, thenar eminence.
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
5 of 15
thumb flexion reflected a slightly lower
selectivity because two animals lacked
selective thumb flexion (Fig.3D and fig.
S6D). To facilitate the interpretation of
functional selectivity values, we reported
the threshold at which a specific move-
ment could be observed (that is, 0.2 on
the graph in fig. S6D). As a reference, the
wrist flexion movement presented in
movie S1 (Mk-Melc) was characterized by
a wrist flexor SI of 0.18 (Fig.3C, Mk-Mec).
To further compare our approach
with the high-density intrafascicular im-
plants used by Ledbetter and colleagues
(21), we computed the muscular selec-
tivity obtained through the Mk-TIME
using their definition of selectivity (see
Materials and Methods and fig. S6E).
Our data show that Mk-TIME stimula-
tion resulted in a highly selective recruit-
ment in both the median and radial
nerves (median=0.76, radial=0.83 versus
median=0.54, radial=0.32 for Ledbetter),
with about four times less channels (fig.
S6E). In addition, most Mk-TIME im-
plants recruited the flexors of the thumb,
relying on a maximum of 16 stimulation
channels per nerve, whereas Ledbetter’s
study revealed that for a 10-by-10 grid,
only a single channel elicited specific
thumb flexion (Fig.3C and fig. S6C).
Collectively, these results indicate that
our transverse electrode configuration
(fig. S6F) provides high selectivity in hand muscle recruitment
without the need to position dense three-dimensional (3D) arrays
inside the nerve.
Chronic intrafascicular implants exhibit stable
properties over time
To evaluate the properties of intrafascicular stimulation of the upper
limb nerves over several weeks, we implanted chronic tailored
Mk-TIMEs in the median and radial nerve of two monkeys, together
with chronic intramuscular leads in the hand muscles (see table S3).
The Mk-TIMEs were routed to connectors fixed on the head through
custom-made, helically wounded robust and compliant cables
(fig. S4). The EMG and Mk-TIME connectors were embedded in a
custom-designed titanium pedestal fixed to the skull (fig. S4). We
evaluated the chronic electrical, mechanical, and functional stability
of two implants (up to 2 months; Fig.4 and Materials and Methods).
Histological comparison between acutely and chronically implanted
electrodes revealed the formation of a scar tissue capsule spanning
over ~150 m on each side of the Mk-TIME, 42 days after implan-
tation (Fig.4A). This body immune response presumably contrib-
uted to the progressive increase in the activation threshold charge
measured over the first 3 weeks of the experiment (Fig.4B). After
this initial period, thresholds remained stable for the rest of the study
(Fig.4B). Functional muscular selectivity (see Materials and Methods),
as well as the proportion of functional stimulation channels, re-
mained stable until the end of the experiment (8 weeks) in the two
animals (Fig.4,CandD). The two monkeys implanted with chronic
Mk-TIME implants did not exhibit any sign of discomfort or pain
(vocalization, restlessness, clenching, or aggressiveness) during stim-
ulation even for the largest amplitudes of current. These results
confirm previous long-term experiments conducted in rodents (24)
and humans (35).
Intrafascicular stimulation elicits a variety of functional grips
We quantified the diversity and the functionality of motor outputs
obtained through intraneural stimulation. The ability of the Mk-TIMEs
to selectively recruit different muscles allowed us to generate a rich
repertoire of hand functions in intraoperative and chronic experi-
ments (Figs.5,AandB, and 6A and movie S1). In particular, median
nerve stimulation bursts delivered through different channels in
several animals elicited the selective flexion of the wrist and the
closing of the hand in three stereotypical grips. Namely, in the three
monkeys implanted, we reported the production of (i) a cylindrical
grip, or hook grip, characterized by the closure of the index, middle,
and ring fingers; (ii) a pinch-like grip, or lateral grip, identified as
the opposition of the thumb against the index and middle fingers;
and (iii) a spherical grip, or whole handgrip, in which all the fingers,
including the thumb, were flexed (Fig.5A, fig. S7A, and movie S1)
(36). Selective wrist flexion was characterized by the dominant acti-
vation of the wrist flexors, that is, the flexor carpi radialis (FCR) and
palmaris longus (PL) (Fig.5A and fig. S7A). The cylindrical grip was
generally associated with increased activation of the finger flexors
A
BCD
Fig. 4. Properties of chronically implanted intrafascicular electrodes. (A) Left and middle: Cross section of ChAT-
labeled, acutely implanted median nerve showing the proximity of the Mk-TIME (asterisk) and motor axons (arrows)
within a fascicle (Mk-Th, acute). Right: Immunological response after 42 days of Mk-TIME implantation in the median
nerve (Mk-Mec). (B) Weekly evaluations of the activation charge threshold (10% recruitment) over three active sites
(n = 2 chronic animals; Mk-Mec, 42 days; Mk-Olc, 58 days). (C) Change in functional selectivity throughout the experi-
ment (n = 2 chronic animals; Mk-Mec, 42 days; Mk-Olc, 58 days). (D) Percentage of change in the number of functional
contacts (eliciting visible movements) throughout the experiment (n = 2 chronic animals; Mk-Mec, 42 days; Mk-Olc,
58 days). Error bars, SEMs. Tuj1, neuron-specific class III -tubulin.
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
6 of 15
[flexor digitorum superficialis (FDS) and flexor digitorum profundus
(FDP)] compared to wrist flexion only (Fig.5A and fig. S7A). The
pinch-like movement recruited the flexors of the thumb (THE) and an
intrinsic muscle promoting index abduction [first dorsal interosseous
(FDI)] and thumb opposition (THE) (Fig.5A and fig. S7A). Last,
the spherical grip combined the activation of the wrist and finger
flexor muscles in all animals and engaged the thumb flexors in two
of three monkeys (Fig.5A and fig. S7A). The patterns of muscular
activation characterizing the elicited cylindrical, pinch-like, and
spherical grips can be related to the natural muscular activation ob-
served in monkeys performing similar hand movements (37). In a
work from Brochier and colleagues (37), hand muscle activity has
been recorded in monkeys executing numerous voluntary grasping
tasks. Similar to the S-ring grip described in their study, our cylin-
drical grip resulted in a large activation of finger flexors with no
thumb flexion (Fig.5A and fig. S7A). The pinch-like grip elicited by
the Mk-TIME was visually comparable to their S-precision grip re-
cruiting THE and FDI (Fig.5A and fig. S7A). Last, the spherical grip
observed in our study resembled their L-cone grip activating the
fingers, wrist, and thumb flexors (Fig.5A and fig. S7A).
Whereas patterns of muscle activation can vary from one animal
to the next owing to the differences in EMG electrode implantation,
musculature, and the number of recorded muscles, kinematic fea-
tures provide a robust way of categorizing grip movements. We
computed the variation in finger and wrist joint angles for each move-
ment elicited by intrafascicular stimulation (Fig.5B and fig. S7A)
and quantified the differences between various grips across relevant
kinematic parameters (Fig.5C and fig. S7D). Our results show that
the spherical grip was characterized by a larger flexion of the thumb
than the cylindrical grip and a stronger flexion of the middle finger
than the cylindrical and pinch-like grips (Fig.5C and fig. S7D). The
opposition of the thumb and index in the pinch-like grip resulted in
a significant decrease in the distance separating these two fingers as
compared to the spherical grip (P<0.01in one animal and P<0.05
in the other two animals; Fig.5C). The index finger was also more
flexed in the spherical grip than in the pinch-like grip. We found
these variations in kinematic characteristics to be similar across all
animals (Fig.5C and fig. S7D), and their representation in a 3D space
recapitulating the largest dimensions of variance (see Materials and
Methods) revealed a clustering of the grips of the same type (Fig.5D).
Calculation of the 3D distances in such space showed that the intra-
individual distance across different grips exceeded the interindividual
distance for the same grip (Fig.5D). These results suggest that the
grips were relatively consistent across animals and that the kinematic
A B
C D
Fig. 5. Functional grip movements achieved with the Mk-TIME. (A) Range of movements obtained by burst stimulation (500 ms) of different channels of the median
nerve (Mk-Mec, awake). Corresponding muscle activation is displayed for each movement. (B) Concurrent kinematic profiles of relevant joint angles and interjoint distances
for each movement throughout the stimulation burst. Stimulation intensities eliciting a similar range of motion were pooled together (see table S6). The gray dashed line
indicates 50% of the full movement range (that is, maximum angle; see Materials and Methods). (C) Average kinematic features measured during stimulation for each type
of grip (cylindrical, pinch, and spherical; n = 3 animals; Mk-Olc and Mk-Mec, awake; Mk-Ola, intraoperatively). Data are normalized within each animal. (D) Left: Grip move-
ments represented as least-squares ellipsoids in the 3D space created by the three first principal components (PCs) obtained by PCA of movement kinematics. Different
colors represent different grips, and different shades represent different animals. Right: Quantification of the Mahalanobis distances between different grips of a single
animal (intraindividual distance) and between the same grips for different animals (interindividual distance) along the two first PCs. *P < 0.05, **P < 0.01, and ***P < 0.001,
nonparametric Kruskal-Wallis test with Bonferroni correction for n > 2 and Wilcoxon rank sum test for unpaired data, except specified otherwise. Shaded lines, SEMs. Error
bars, SEMs. PIP, proximal interphalangeal; MCP, metacarpal interphalangeal; Data specifications (experimental parameters and conditions) for each panel are summarized
in table S6.
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
7 of 15
features characterizing different grips substantially differed within
the same monkey.
Similar to median nerve stimulation, intrafascicular bursts in the
radial nerve triggered specific hand movements such as ulnar devi-
ation and extension of the wrist, hand opening through finger ex-
tension, and abduction of the thumb (Fig.6A; fig. S7, B and C; and
movie S1). EMG analysis unveiled that ulnar deviation of the wrist
mainly engaged the extensor carpi ulnaris (ECU), whereas wrist ra-
dial extension recruited the extensor carpi radialis (ECR) (fig. S7, B
and C). Finger extension was associated with the activation of the
fingers and thumb extensors [extensor digitorum communis (EDC)
and abductor pollicis longus (APL)] (fig. S7, B and C). We computed
relevant kinematic features (fig. S7C) distinguishing various exten-
sion movements and reported the average variation in kinematic
profile for each animal (Fig.6B). Whereas the specific opposition of
the thumb was observed in one animal only, the ulnar deviation of
the wrist was observed in two of three monkeys, and the selective
extension of the wrist and fingers was present in all animals. For
these movements, kinematic features were homogeneous across
subjects (Fig.6B).
Grip force can be modulated and sustained through
intrafascicular stimulation
We then evaluated the possibility to tune the grasping force by
modulating the frequency and amplitude of intrafascicular stimula-
tion (Fig.7,AandB; fig. S8A; and movie S2) (38). Similar to surface
NMES (39,40), we found that the stimulation amplitude was directly
linked to the degree of hand muscle activation and, thus, to the grip
pressure (Fig.7,AandB). Conversely, the increase of stimulation
frequency was not associated with an increase of the magnitude of
muscle responses (fig. S8A) but only with an increase in their occur-
rence frequency. This resulted in a reduced force range compared to
amplitude modulation (Fig.7C). In addition, sinusoidal frequency
modulation delivered during a 10-s stimulation wave (see Materials
and Methods) resulted in a faster decrease of grip force, hence a
more rapid onset of fatigue, than sinusoidal amplitude modulation
(Fig.7C). By increasing the stimulation amplitude of specific channels
evoking different grips, we reached grip force values superior or
equal to those generated voluntarily by a monkey performing a func-
tional grasping task (Fig.7D and Materials and Methods). More
specifically, muscular contractions elicited through intrafascicular
median nerve stimulation achieved between 35 and 103% of the force
obtained through maximum voluntary contraction (MVC) (Fig.7D).
We did not observe any significant change in selectivity (see Mate-
rials and Methods) when modulating the stimulation charge to attain
these forces (P>0.05; Fig.7D). Similarly, increasing the stimulation
frequency did not result in any change in selectivity of muscular
recruitment (fig. S8B).
We quantified the ability of the Mk-TIME to sustain functional
amounts of force for extended periods of time by measuring the
grip pressure and wrist torques during tonic stimulation bursts (see
Materials and Methods). Our analysis demonstrated that the force
was only attenuated by 10% after 5s of stimulation and declined by
about 20 to 50% after a 20-s stimulation wave at 50Hz (Fig.7E and
fig. S8, C and D). These time periods are considerably longer than the
average execution time of reaching and grasping in intact (37,41) or
paralyzed (6) monkeys, suggesting that intrafascicular stimulation
can be used to perform such functional tasks.
We also alternately stimulated independent channels of the
median and radial nerves to reconstruct smooth sequences of hand
movements and reproduce natural patterns of muscular coactivation.
Natural grasp preparation and hand closure on an object require the
exertion of a wrist extension torque while opening and closing the
fingers (37). To achieve this, we applied synchronized stimulation
bursts on the radial nerve channels eliciting specific extension of the
wrist and fingers. We then disabled the channel eliciting finger
A
B
Mk-Mea
Ulnar deviation Fingers extens. Wrist extens. Thumb extens. Thumb-index
extens. Mk-OlcMk-MeaMk-Ola
θ Extension
wrist (°)
-20
0
20
40
**
*
Uln. dev.
Wrist ext.
θ Ulnar
deviation (°)
-20
-10
0
10
***
*
Uln. dev.
Wrist ext.
Index to
pinkie dist. (a.u.)
0
0.5
1
***
*
Uln. dev.
Fing. ext.
θ Extension
PIP pinkie (°)
***
0
10
20
*
Fing. ext.
Wrist ext.
Thumb to
index dist. (a.u.)
0.5
1*
Thumb ext.
Thumb-ind
ext.
Thumb
opoosition (°)
0
20
40
*
Fing. ext.
Thumb. ext.
Ulnar deviation Wrist extension Fingers extension Thumb-index extension
Fig. 6. Functional extension movements achieved with the Mk-TIME. (A) Types of movement obtained by stimulating different channels of the radial nerve (Mk-Mea
intraoperatively). (B) Average kinematic features measured during stimulation for each type of extension movement (n = 3 animals; Mk-Olc, awake; Mk-Mea and Mk-Ola,
intraoperatively). Data are not normalized because not all movements were observed in all animals. *P < 0.05, **P < 0.01, and ***P < 0.001, nonparametric Kruskal-Wallis
test with Bonferroni correction for n > 2 and Wilcoxon rank sum test for unpaired data, except specified otherwise. Error bars, SEMs. Data specifications (experimental
parameters and conditions) for each panel are summarized in table S6.
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
8 of 15
extension and activated a median nerve channel responsible for a
particular grip while maintaining the stimulation on the wrist ex-
tension channel. This strategy produced a natural transition be-
tween extension and flexion of the fingers while maintaining a
stabilized wrist posture (movie S1, Mk-Ola). In addition, we ob-
served that by alternately modulating the amplitude of stimulation
bursts on channels of the median and radial nerve implants, we
were able to generate smooth opening and closing of the hand, as
well as selective flexion and extension of the wrist (Fig.7F and
movie S2).
We lastly quantified the number of stimulation trials that could
be performed before muscle fatigue occurs. We first demonstrated
that after about 25 consecutive graded stimulation waves delivered
to the median nerve, the activity of the finger flexor and PL de-
creased by 50%, whereas the activation of the wrist flexor stabilized
around ~70% (fig. S8E). This corresponded to 50s of back-to-back
A
C
B
ED
FG
Fig. 7. Intrafascicular stimulation can be modulated to produce sustainable functional force. (A) Experimental setup used to record the force and muscular activity
generated by intraneural stimulation. The wrist torque is recorded in an isometric manner using a dual-range force sensor. Grip pressure is acquired using a custom-
designed grip pressure sensor. (B) Left: Traces showing the effects of stimulation charge modulation on elicited muscular activity and grip pressure over three sinusoidal
stimulation cycles (Mk-Mec, 20 days after implantation). Middle: Normalized grip pressure as a function of the injected charge (curve fit with second-order polynomial,
R2 = 0.99). Right: Corresponding increase in the EMG peak to peak for the finger flexor and THE muscles (n = 1 animal). (C) Left: Force range obtained when modulating
the amplitude or the frequency of stimulation. Force measurements are obtained by converting voltage values to newton using calibration curves (see Materials and
Methods and table S8). The grip force elicited during a maximum voluntary contraction (MVC) is reported with a dark dashed line (see Materials and Methods). Right:
Decrease of force peak after 10 s of modulation, expressed in percentage of maximal peak. (D) Left: Grip pressure elicited for different grips (full bars, different colors
indicate different monkeys) compared to the voluntary force exerted during a behavioral functional grasping and pulling task (striped bars; see Materials and Methods).
The grip force elicited during a MVC is reported with a dark dashed line (see Materials and Methods). Data labels indicate the exerted force in percentage of MVC. Right:
Difference between the SI for the charge at which a grip is observed and the SI measured when injecting enough current to produce functional forces observed on the
left (n = 7 data points; one-sample t test, P = 0.25; see Materials and Methods). (E) Evolution of grip pressure over 20 s of tonic stimulation on a single channel of the median
nerve (n = 2 monkeys). (F) Example of bidirectional stimulation achieved by alternatively stimulating one channel from the radial and median nerve, respectively. Injected
charge is shown over three stimulation cycles per electrode. Corresponding EMG envelopes are plotted for the wrist flexor and extensor muscles, and normalized wrist
torque is reported on the top right. Snapshots extracted from video recordings at different time points during the stimulation display successive closing and opening
movements of the hand (Mk-Mea, acute). (G) Average EMG peak-to-peak activity expressed as a percentage of the first data point for consecutive median nerve stimulation
bursts at different frequencies (see fig. S8F; Mk-Mec, 14 days after implantation). The dashed gray line indicates a 50% decrease in the muscle activity. The number of
stimulation bursts is indicated in light blue on top of the graph. Not significant (ns), P > 0.05; *P < 0.05, **P < 0.01, and ***P < 0.001, nonparametric Kruskal-Wallis test with
Bonferroni correction for n > 2 and Wilcoxon rank sum test for unpaired data, except specified otherwise. Shaded lines, SEMs; Error bars, SEMs. Data specifications for each
panel are summarized in table S7.
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
9 of 15
stimulation waveforms during which the amplitude was ramped up
and down in a sinusoidal shape. We also showed that short stimula-
tion bursts (~500 ms) delivered at different frequencies with 5-s
intertrial intervals resulted in homogeneous muscle peak-to-peak
activity (fig. S8F) and showed no fatigue effects after 30 trials (Fig.7G).
These results suggest that consecutive grips can be performed for
almost 1min before the onset of fatigue and that recovery periods
are necessary to avoid the exhaustion of muscle fibers.
Mk-TIME restores hand extension during a functional
grasping task: A proof of concept
In the last step, we performed a preliminary experiment aimed at
expanding these encouraging results to a behaving animal with hand
paralysis. We explored the possibility of using the Mk-TIME to en-
able a monkey to perform a voluntary functional task requiring the
activation of partially paralyzed muscles. To investigate the feasibility
of recruiting wrist and finger muscles in the absence of fully func-
tional spinal circuits, we used a temporary pharmacological nerve
block, inhibiting the transfer of afferent and efferent signals through
the nerve. For this experiment, we relied on a simple yet intuitive
stimulation control strategy whereby a single command computed
from intracortical recordings directly adjusted the relative stimula-
tion charge delivered through the radial nerve Mk-TIME (fig. S9).
To be able to map the delivered electric charge to a unique command,
we chose to use a single stimulation channel producing concomitant
wrist and finger extension. This control paradigm, although being
too simplistic to be applied to more complex grip movements or
directly used in translational studies, provided an effective and easy
way to test the performance of intrafascicular stimulation in a
behaving animal, whose muscles are partially paralyzed and spinal
reflexes are interrupted.
To selectively impair hand opening, we engineered a microfluidic
drug delivery system (42) composed of a silicone cuff wrapped
around the nerve proximally to the Mk-TIME and coupled to a
thin tube directed to the chronic assembly fixed on the skull of
the animal (figs. S4A and S10A). Lidocaine delivery to the nerve
induced a pharmacological block proximal to the injection (fig.
S10B) and increased the charge threshold for muscular activation
(fig. S10, B and C). During the experiment, lidocaine infusion inside
the pedestal port induced a selective attenuation of radial nerve
signals, resulting in specific transient paresis of the hand extensor
muscles (fig. S10D).
To build the control program, we recorded the spiking activity
in the contralateral primary motor cortex (hand area) of the mon-
key while reaching and grasping objects presented by a robotic arm
(fig. S11) (41). We then used principal components analysis (PCA)
to compute the main dimensions of variance found in the cortical
activity during the task (43). The resulting coefficient matrix, labeled
hereafter as “neural modes coefficient matrix” (fig. S11), mapped the
original brain activity to the linearly transformed PC space. Because
hand opening primarily occurs when approaching an object, we
anticipated that the PC correlating the most with movement during
the reach would provide a simple and effective control variable to
drive the stimulation in real time. Evaluation of the neural activa-
tion during movement projected on the first three PCs revealed a
preferential tuning along the second PC (PC2), accompanied by a
larger modulation depth than for the first and third PCs (fig. S12, A
and B). These properties were found to be preserved throughout
months of behavioral recordings (fig. S12C).
To perform the experiment, we first computed the modulation
depth along PC2 on data acquired before the injection of the para-
lytic drug through the microfluidic cuff. We then evaluated the
stimulation charge thresholds after the inactivation of the radial nerve
and calibrated the gain and stimulation parameters of the controller
(see fig. S11 and Materials and Methods).
Without preliminary training, the monkey was able to integrate
instantaneously the stimulation-induced hand movement to the task
behavior and efficiently and smoothly reach for the presented target
object and complete the task (fig. S13, A to C, and movie S3). Stim-
ulation trials were marked by a significantly higher success rate
(P<0.001; fig. S13B and movie S3) and a decrease in the number of
imprecise reach movements (P<0.01 for session 1; fig. S13B and
movie S3). The graded stimulation of the radial nerve based on the
PC2-related cortical dynamics potentiated the activation of the wrist
and finger extensors (fig. S13, A and D), resulting in integrated
EMG values comparable to the baseline condition (fig. S13D).
Intrafascicular stimulation applied after paralysis of the hand restored
wrist extension angles comparable to the tonic activation observed
during baseline trials (fig. S13E). This recovery coincided with an
improved kinematic strategy during the execution of the movement.
The interjoint coordination (44) measured in trials executed with
intraneural stimulation was found to be markedly closer to the
healthy condition than in the trials performed without stimulation
(fig. S13F and Supplementary Methods).
The simplicity of our control paradigm nevertheless limited the
accuracy of the intrafascicular stimulation delivery, resulting in
imprecise reaching movements in some stimulation trials (fig. S13C
and movie S3). Offline performance analysis revealed that for these
trials, the timing of wrist extensor activation was similar to the one
measured during baseline trials and dexterous reaches, but the acti-
vation of finger flexors was significantly delayed (P<0.001; fig. S13G).
This time shift was presumably caused by the incomplete capture of
the animal’s motor intent into a single PC. In these faulty trials, we
observed a large activation of the PC2 variable (fig. S13H, left),
which resulted in an increase in the stimulation amplitude and the
duration of the stimulation burst (fig. S13H), thereby impeding the
proper closing of the hand. A decoder that incorporates additional
PCs will likely encapsulate more comprehensive motor commands
and enhance the control of the stimulation to produce sophisticated
hand movements.
DISCUSSION
We introduced a tailored design and implantation paradigm for
generating dexterous and functional hand movements using intra-
fascicular stimulation in a preclinical animal model. Our approach
displayed the ability to selectively engage different muscles using a
single implant inserted in the nerve. This specific activation pro-
duced sustained, functionally relevant, and controllable actuation
of the wrist and fingers with low amounts of current in three non-
human primates. The simplicity and efficacy of the system are illus-
trated by the rich variety of grasps and extension movements evoked
by only two Mk-TIMEs using monopolar stimulation protocols.
Intramuscular stimulation has been used as an alternative to
peripheral nerve and surface stimulation to improve the selectivity
of hand muscle recruitment (5). Its implementation in a clinical
system allowed, as in the case of the Mk-TIME, the opening of the
hand and the generation of three grips per subject (lateral or pinch,
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
10 of 15
palmar, and parallel grips) (4). However, such an intramuscular
device requires a considerable number of implantable leads (10 per
subject), and researchers reported the difficulty of precisely placing
the electrodes recruiting small intrinsic muscles of the hand (45). In
addition, data showed that intramuscular stimulation of extrinsic
finger muscles is not sufficient to produce a functional opening of
the hand and necessitates the additional implantation of intrinsic
finger extensors (4). Ultimately, by increasing the number of implants
needed to be controlled simultaneously to produce a single grip
movement, intramuscular paradigms increase surgical risks and
compel to more tedious manual design of stimulation strategies than
intrafascicular implants (8).
Intrafascicular recruitment was found to reproducibly engage
the hand intrinsic muscles and selectively activate a large number of
flexor and extensor muscles. Collectively, these results suggest that
the Mk-TIME might outperform standard epineural implants for the
production of fine hand movements (14). Although epineural stim-
ulation can very selectively recruit motor fibers lying close to the
nerve surface (14), it fails at targeting muscles whose innervating
fibers lie deep within the endoneurial compartments, such as distal
intrinsic hand muscles (25). To activate as many muscles as intra-
fascicular stimulation, epineural stimulation will likely require
the implantation of additional interfaces and the design of complex
multipolar stimulation patterns that depend on the implant location
and nerve organization (14). Although both intrafascicular and epi-
neural implants require surgical implantation in the upper arm, the
lower performance of epineural interfaces compels to implant more
electrodes into the arm (13), increasing the risk to disrupt muscular
structures. In contrast, a single intrafascicular implant creates a
minimal footprint inside the nerve (24) and can produce selective
neural responses over many months (24,46).
In the context of neurological disorders, the large collection of
movements evoked by intrafascicular stimulation offers promising
therapeutic perspectives for those suffering from spinal cord injury
or stroke. Rehabilitation studies involving patients affected by
ischemic events reported that compensatory mechanisms arising in
the subacute phase more efficiently restored gross arm movements
than precise finger motion (47,48). Similarly, individuals suffering
from spinal cord lesions still lack acceptable solutions for the recovery
of fine hand motion and only partially benefit from current rehabil-
itation strategies (49,50). In patients affected by spasticity (51–53)
and in individuals exhibiting a flaccid state of the limb (54,55), the
relaxation of the fist and the stable support of the hand present
important medical challenges that could potentially be addressed
through intrafascicular stimulation of the extensor muscles.
Spinal cord injury often results in important changes in spinal
reflexes (56), which can increase the complexity of muscular responses
triggered by electrical stimulation (32–34). Here, we demonstrated
that hand muscle responses to intrafascicular stimulation primarily
derived from the direct recruitment of motor axons, thereby allow-
ing functional muscle recruitment even in the absence of fully func-
tioning spinal reflex circuits belaying transsynaptic activation of
muscle fibers. Furthermore, we showed that stimulation bursts de-
livered through the Mk-TIME in a temporarily paralyzed monkey
produced muscular activity comparable to intact animals and gen-
erated functional extension movements. Because the model of
paralysis used in our last experiment did not strictly reproduce a
spinal injury condition, the clinical translation of our approach
would nonetheless require the validation of its therapeutic benefits
in subjects suffering from spinal cord injury. Previous anatomical
examinations of the upper limb muscles and nerves in monkeys and
humans (25–27,57) revealed that the structural organization of
the arm is preserved within primates, suggesting that our approach
could be directly translated to humans. The larger number of fascicles
in the human median nerve and the interindividual variations in
motor fiber organization (25,27) may indicate the possible need for
additional Mk-TIME implants, or denser electrode arrays (21), to
achieve similar results.
We showed that simple stimulation patterns could be linked with
cortical inputs to provide a primitive low-dimensional control of the
implant, allowing a monkey with transient hand paralysis to regain
extension functionality and complete a reach-and-grasp task. Al-
though this experiment extended the applicability of restoring func-
tional and usable movements obtained in healthy, immobilized
animals to a behaving animal, it must be considered a preliminary
proof of concept. A major limitation of our approach lies in the sim-
plistic design of our cortical control strategy. Such paradigm enabling
the modulation of a single active site using a unique command
cannot be realistically scaled to more complex grip movements. The
efficient translation of intrafascicular stimulation for fine hand con-
trol requires performant algorithms for the decoding of a large
repertoire of finger movements (58). Further studies will be needed
to clarify how to convert the decoded information into relevant
stimulation patterns that can be easily tuned by the user. In this
regard, the simple synergistic activation of hand muscles triggered
by single intrafascicular channels presents encouraging even if pre-
liminary evidence for its instinctive control through a relatively low
number of instructing commands. Our results, together with the
recent clinical application of intrafascicular implants to restore
sensory feedback in upper (38,46,59) and lower limb (60,61)
amputees, suggest promising opportunities to restore fine hand
motor function in subjects with paralysis.
MATERIALS AND METHODS
Study design
This preclinical study in macaque monkeys aimed at assessing
whether intrafascicular electrodes could be used to produce precise
movements of the hand. The experiments were conducted on
10 adult macaques (Macaca fascicularis) ranging from 5 to 16 years old
(weight, between 3.1 and 8 kg) detailed in table S1. All animals were
group-housed in an enriched indoor room where they had access to
water and food ad libitum. The experimental protocol was elaborated
in compliance with the national law on animal protection and
approved by the federal and local veterinary authorities (veterinary
authorization numbers 2017_03_FR, 2017_03_E-FR, 2014_42E FR,
2017_04E_FR, 2016_09_FR, and 2017_22_FR). Randomization and
study size calculation were not applicable. To optimize the surgical
procedure and determine the dimensions of intrafascicular elec-
trodes, we performed a neuroanatomical analysis on six cadaver
arms. This preliminary step allowed characterizing interindividual
anatomical variabilities and optimizing our implantation approach
for the median and radial nerves. We conducted acute electrophys-
iology experiments on seven animals to evaluate the efficacy of
intrafascicular electrodes to selectively and reproducibly recruit flexor
and extensor muscles of the hand. We performed four repetitions
of each condition as documented in previous similar studies (24).
We characterized the longevity of our electrodes in two chronically
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
11 of 15
implanted monkeys (Mk-Mec and Mk-Olc), thus gathering data from
multiple subjects while minimizing the number of animals involved.
These two chronically implanted animals, together with two of the
acutely implanted monkeys (Mk-Mea and Mk-Ola), were involved
in the functional grip characterization and the force and fatigue ex-
periments. Because we found no reference for this type of work, we
performed as many trials as possible for each condition within the
limited time at our disposal. We determined that three subjects were
sufficient to generalize our findings despite interindividual variability
while reducing the number of involved animals and the duration of
experiments. The time dedicated to daily experimental sessions in
chronically implanted monkeys was determined by the maximum
time under anesthesia (~3 hours) or the maximum reasonable time
spent in the experimental room (~2 hours). The overall duration of
chronic experiments is detailed in table S1. Acute experiments typ-
ically lasted for a full day (~8 hours). Our last pilot experiment was
designed to demonstrate the ability of intrafascicular electrodes to
restore hand extension in a functional task. As an early proof of
concept, we performed this last study in only one of our chronically
implanted animals (Mk-Olc) and repeated over two experimental
sessions. Quantification of the dexterous, imprecise, and failed reaches
in the behavioral experiment was performed in a blind manner, that
is, without knowing whether the stimulation was triggered during
the trial. The detailed number of repetitions of each experiment is
reported in the corresponding legends, Materials and Methods,
and tables.
Dissection and tissue processing
Cadaver arms were harvested after transcardial perfusion (62,63)
and dissected. Nerve samples were collected (~3cm proximal to the
epicondyle), embedded in paraffin, and sectioned in 4-m cross
sections (Hyrax M25, Microm) before storage at +4°C.
Fascicle topography
3D reconstructions (Neurolucida 11.0, MBF Bioscience) and fascicular
topography analysis along the dissected nerves were performed
using evenly spaced hematoxylin and eosin–stained cross sections
(one slice every 40; Tissue-Tek Prisma Sakura) acquired with an
optical microscope (Olympus slide scanner VS120-CS100, Olympus
Corp.) at 10×. Fascicular composition was analyzed using the open-
source software Fiji (ImageJ, National Institutes of Health) (fig. S1).
Motor axon distribution
We performed a qualitative assessment of the motor fiber distribu-
tion on nerve cross sections labeled for ChAT (goat anti- ChAT; 1:50;
AB144P, Sigma-Aldrich), revealed with a 3,3′-Diaminobenzidine
(DAB)–peroxidase (horseradish peroxidase) kit (Vector Laboratories).
Images acquired at 20× were analyzed using a custom-written routine
in Fiji-ImageJ (fig. S1).
Mk-TIME biointegration
Slides were processed for immunohistochemical labeling against
III tubulin (axons, rabbit anti-Tuj; 1:200; ab18207) and macrophage/
monocytes (mouse anti-CD68; 1:200; MCA341GA, Bio-Rad) and
revealed using a secondary antibody solution containing goat anti- rabbit
immunoglobulin G (IgG) Alexa Fluor 488 and goat anti-mouse IgG Alexa
Fluor 555 (1:200; Thermo Fisher Scientific). Samples were mounted
with a 4′,6-diamidino-2-phenylindole (DAPI)–containing medium
(VECTASHIELD, Vector Laboratories) and acquired at 40×.
Computational model
We used histological cross sections of the median nerve at different
implantation levels to build a realistic finite element model using
previously described methodology (28,64). The neurophysical mod-
els of motor fibers were implemented in Anaconda 2.7.14 (Python
Software Foundation), using NEURON v7.4 to solve the membrane
dynamics. To represent the dynamics of each nerve fiber, we used
the McIntyre-Richardson-Grill model (64–66). The biophysical
properties of the fiber compartments were derived from previous
work (28,64,67,68). The implementation of our computational
model is detailed in Supplementary Methods (see the “Computa-
tional model implementation” section).
TIME and microfluidic drug delivery system fabrication
The TIME electrode was derived from previously published design
(69). The manufacturing of the peripheral implants is detailed in
Supplementary Methods (see the “Peripheral implants” section) and
figs. S4 and S10A.
Simulated recruitment curves and selectivity analysis
We computed recruitment curves for individual motor fibers during
40-s-long square pulses of monopolar cathodic stimulation at in-
creasing amplitudes. We considered a fiber to be recruited if an ac-
tion potential traveled along its whole length. Recruitment curves,
activation and saturation amplitude, and selectivity indices were
computed on the basis of a bootstrap analysis (with replacement) of
simulated fibers. The resulting box plot distributions were thus de-
rived from bootstrapped samples (n=10,000 populations per muscle).
To compare the specificity of stimulation at different locations along
the nerve, we computed the SI as
SI m = REC m −
∑
n≠m
M REC n
─
M − 1
where m is the muscle of interest, M is the total number of innervated
muscles, and REC is the normalized recruitment level. The SI was
computed for every active site, amplitude steps (from 20 A to 2 mA,
steps of 100 A), muscles, and bootstrapped populations. Except
when analyzing activation and saturation thresholds or SIs along the
electrode, we considered only the channel and current step producing
the highest average SI for each muscle. The analysis of the modeled
recruitment is detailed in Supplementary Methods (see the “Simu-
lated recruitment curves and selectivity analysis” paragraph in the
“Data processing and analysis” section).
Surgical procedures
All surgical procedures were conducted under aseptic conditions and
general anesthesia (63).
Brain array surgery
Mk-Olc was implanted with three microelectrode arrays. One array
with 48 channels was implanted in the right hemisphere in the M1
hand region (400-m pitch, 1.5-mm tip length; Blackrock Microsys-
tems). The primary somatosensory cortex and ventral premotor cortex
areas were also implanted but were not analyzed for this study.
Microfluidic system and Mk-TIME electrode implantation
Two custom-made Mk-TIME implants were inserted in the median
and radial nerves according to previously described methods (24,70).
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
12 of 15
The median electrode was placed ~2cm proximally to the elbow,
and the radial electrode was implanted ~2cm proximally to the
epicondyle along the humeral bone. Electrophysiological testing was
performed intraoperatively to adjust the position of the electrode
and verify that a single pulse of stimulation delivered through active
sites located at the shaft extremities induced motor responses. The
microfluidic cuff was positioned proximally to the electrode and
secured using a medical-grade staple (Premium Surgiclip, Medtronic).
Wires were routed subcutaneously along the anterior compartment
of the arm. Procedures are detailed in Supplementary Methods (see
the “Surgical procedures” section).
Intramuscular electrode implantation
We implanted two monkeys with eight pairs of Teflon-coated stain-
less steel wires to chronically record EMG activity from the flexor
and extensor muscles of the hand. The detailed implantation proce-
dure is described in (62). Acute EMG recordings were performed
using intramuscular subcutaneous needles (Ambu Neuroline). The
muscles implanted in chronic and acute experiments are summarized
in table S3. Intrinsic hand muscles were always implanted acutely.
Electrophysiology
Electrical stimuli were delivered as charge-balanced cathodic-first
biphasic pulses through a 32-channel headstage (LP32CH-32,
Tucker-Davis Technologies) using a medical-grade stimulator (IZ2H,
Tucker-Davis Technologies). Stimulation waveforms were digitally
built within the processor unit (RZ2, Tucker-Davis Technologies)
using the user programming interface OpenEx suite (Tucker-Davis
Technologies). Custom-written routines were used to communicate
with the controller through MATLAB (MATLAB, MathWorks Inc.)
or C++ (Visual Studio) API.
Muscle recruitment curves
Recruitment curves (n=6 acute and 2 chronic experiments; table S3)
were performed by delivering biphasic cathodic-first current pulses
(1 Hz) of 40- or 80-s duration at increasing current intensities
ranging from subthreshold to saturation of the CMAP (20,24). EMG
activity was filtered online (50-Hz notch; 10- to 5000-Hz band-pass).
For every channel, four repetitions of each current step were per-
formed, and the average peak-to-peak amplitude of the evoked
CMAP of each muscle was used to analyze the relationship between
stimulation-evoked muscle activity and stimulation intensity. The
recorded CMAPs were normalized to their maximal amplitude ob-
tained throughout the experiment for each muscle.
In vivo selectivity analysis
The selectivity for every muscle and each charge injection step was
assessed by calculating its SI. Similar to our simulated recruitment
curve analysis, we used the previously described SI (28)
SI m = CMAP m −
∑
n≠m
M CMAP n
─
M − 1 (1)
where m is the muscle of interest, M is the total number of implanted
muscles, and CMAP is the normalized CMAP (fig. S6A). This defi-
nition results in an SI varying from −1 to 1, where 1 means complete
activation of the mth muscle and no activation of the others and −1
means complete activation of all the muscle groups except for the
mth. This definition of selectivity intends to penalize situations in
which the activation of the targeted muscle is concomitant with the
activation of other untargeted muscles. It uses relative activation by
normalizing it by the number of muscles that varied across experi-
ments. In comparison with alternative SI computations (14), this
approach ensures that stimulation parameters producing similar
differences in muscular activations will result in comparable SI values
(fig. S6A). The SI was computed for each muscle across every active
site and amplitude step. For a given muscle, only stimulation pa-
rameters for which this muscle was the most activated were consid-
ered, constraining the SI between 0 and 1. In addition, we imposed
that the activation of the second maximally activated muscle did not
exceed 80% because the sporadic high activation of one muscle will
not necessarily decrease the SI if the activation of all the other mus-
cles remains low. The intrinsic thumb flexor (flexor pollicis brevis)
and thumb opponens (opponens pollicis) were labeled as THE. When
both were implanted, we computed THE selectivity using the most
activated one. The final selectivity for each muscle was computed as
the maximum SI achieved across contacts and charges considered
(Fig.3B and fig. S6, A and B). The functional SI was calculated by
grouping muscles according to their function, namely, wrist flexion
(FCR and PL), wrist extension (ECR and ECU), finger flexion (FDS
and FDP), finger extension (EDC), thumb/index opposition (THE
and FDI), and thumb extension (APL), and computing the highest
SI within each group (fig. S6D). To compare the selectivity achieved
using the Mk-TIME with previous studies in monkeys, we computed
the SI of each muscle across every stimulation parameter (channels
and injected charge) using the definitions of Brill etal. (14) and
Ledbetter etal. (21) (Fig.3C and fig. S6, A to E). We reported the
maximum value for each muscle in Fig.3C and fig. S6 (A, C, and E).
Electrode stability analysis
We assessed the functional stability of the chronic implants (Fig.4)
in two animals (Mk-Mec, median nerve, 42 days and Mk-Olc, radial
nerve, 59 days) by computing the weekly functional SI (defined
above) for each of the three movement groups per nerve (wrist, digit,
and thumb flexion for median nerve and wrist, digit, and thumb
extension for radial nerve; see previous paragraph). We also charac-
terized the Mk-TIME stability by computing the evolution of the
threshold charge, defined as the charge necessary to elicit 10% of the
maximally achieved muscle activation, for each muscle over the time
course of the experiment. Implant integrity was assessed by count-
ing the number of functional contacts, defined as contacts eliciting
visible movement for a charge inferior to 100 nC, and normalizing
them to the first measurement day.
Experiments in awake and anesthetized monkeys
Functional and force modulation experiments were performed in
awake or anesthetized conditions. In the awake paradigm, the animal
was sitting in a chair, slightly sedated [ketamine (3.5mg/kg) and
NaCl 1:1; 0.1 ml/15 min] to stay calm for long recording periods. In
the anesthetized condition, the animal received intravenous perfu-
sion of propofol (0.2 to 1.7 ml/kg per hour), and vitals were contin-
uously monitored. In both cases, the forearm was fixed on an
armrest, leaving the hand visible from all directions to record kine-
matics and/or kinetic signals.
Functional mapping
The characterization of the functional movements elicited through
stimulation was done by delivering 500- to 1000-ms bursts (frequency
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
13 of 15
of 50Hz and pulse width of 40 or 80 s) at different amplitudes.
Channels of the median and radial nerve electrodes were mapped
individually.
Force modulation
The effects of stimulation amplitude or frequency on the generated
wrist force and grip pressure were assessed by modulating one of
these two parameters at a time, on individual stimulation channels.
Sine-wave modulation was delivered between a minimum and a
maximum amplitude or frequency value with a 1- to 2-s period
(Fig.7,BandC, and fig. S8A). The amplitude range depended on
the active site, and the frequency range was set between 40 and
100Hz. Each trial lasted about 10s. Coactivation of flexor and ex-
tensor muscles was achieved by alternately stimulating the median
and radial nerve with a sinusoidal amplitude wave of a 2-s period
and a phase shift of 90° (Fig.7F). Fatigue experiments consisted of
20s of tonic bursts at 50Hz on one single channel (Fig.7E and
fig. S8, C and D).
EMG burst and continuous EMG recording analysis
EMG signals were band-pass filtered offline (50 to 500 Hz). In the
case of stimulation-evoked activity, a Savitzky-Golay filter (2.5-ms
smoothing window) was applied to remove stimulation artifacts.
The envelope was computed by rectifying the EMG and applying a
low-pass filter at 6Hz. Signals were normalized to the maximal
muscle activity obtained across the trials of interest and interpolated
when trials differed in length.
Hand kinematics analysis
Postprocessing of kinematic data was performed to ensure that joint
markers were properly labeled. For movements recorded intraoper-
atively under anesthesia (Mk-Mea and Mk-Ola), 2D kinematics of
each finger were extracted manually from single-camera frames us-
ing a custom-built MATLAB routine (MATLAB, MathWorks Inc.)
(table S5). In chronic awake recordings, we measured 3D kinematics
using reflective markers, converted the position data to rotational
degrees of freedom, and computed joint angles and distances sepa-
rating fingers during median and radial nerve stimulation. Stimula-
tion intensities eliciting a similar range of motion were pooled
together [table S6 for Fig.5(AandB) and captions for fig. S7 (A to C)].
To relate the elicited change in kinematic joint angles to the natural
range of movement, we computed the percentage of full movement
range covered by the largest change in elicited angle. Namely, for a
given kinematic feature, we chose the movement displaying the
largest average variation and computed it as a percentage of the
maximum “natural” angle range (Fig.5,AandB, and fig. S7, A to C).
For median nerve stimulation, because the hand was resting partially
flexed (no active extension), we divided the maximum natural angle
range by 2. We thus used 45° for finger joint angles and 22.5° for the
ulnar wrist deviation angle. For radial nerve stimulation, the maxi-
mum natural angle range was set to 90° and 45° for wrist extension
and wrist ulnar deviation angles, respectively. Kinematic features
were low-pass filtered at 10 Hz, and the mean value recorded at
maximum movement amplitude (~100-ms window) was subtracted
to the value at rest (~100-ms window before stimulation) to derive
the scatterplots in Figs.5C and 6B and fig. S7D. Outliers falling out-
side the 25th to 75th percentile range were excluded. We applied
PCA on normalized grip kinematic features (table S5) from all three
monkeys and computed the inter- and intraindividual Mahalanobis
distance along the first two PCs by pooling together the distances
obtained between PCA clusters representing different grips of the
same animal and identical grips from different animals, respectively.
Grip force analysis
Data were low-pass filtered at 5Hz. To determine the relationship
between grip force and stimulation parameters (Fig.7B and fig. S8A),
forces elicited at the same amplitude or frequency value were aver-
aged. Amplitude and frequency modulation strategies were compared
in terms of (i) generated range of force, computed as the peak to
peak of the force (Fig.7C, left), and (ii) induced fatigue, expressed
as the difference between the first and the last peaks of the force
over 10s of sine wave stimulation (Fig.7C, right). The difference
between the two modulation approaches was examined by pairwise
comparing trials with the same force sensor and the same active site.
The grip pressure corresponding to the maxima of the amplitude
sinusoidal wave was compared with the natural amount of force
exerted by a monkey (Mk-Olc) during a voluntary grasping task
(Fig.7D) (41). This functional task consisted of grasping and pull-
ing an object mounted on a haptic robot (KUKA iiwa 7, KUKA AG)
over ~20cm. It required the monkey to overcome the resistance of
the robotic joints set to 200 N/m in the x, y, and z dimensions of
space. In our case, the monkey was presented with cylindrical,
spherical, and pinch-like objects. Specifically, we compared the force
peaks generated by the Mk-TIME active site best evoking a certain
type of grasp to the voluntary exerted pressure on the same grip
sensor shape (sphere, cylinder, or pinch). In addition, to compare
the force generated using intrafascicular stimulation to a monkey
MVC, we trained an animal (Mk-Olc) to squeeze our custom-made
spherical grip pressure sensor (41) as strong as possible using operant
conditioning. Briefly, the animal was trained to fixate a screen dis-
playing a square whose size was proportional to the grip pressure
applied. The monkey progressively learned to apply a stronger and
stronger force to reach a target square size and receive a liquid food
reward. The force corresponding to the MVC was chosen as the
maximum force value obtained across 150 trials (Fig.7D). Voltage
values were converted to newton (Fig.7,CandD) using calibration
curves depicted in (41) and table S8.
Evaluation of selectivity change for strong grips
We quantified the change in selectivity when modulating the stim-
ulation charge (Fig.7D, right) by evaluating (i) the charge threshold
at which a distinct movement was observed and (ii) the charge
threshold for which we measured functional forces (Fig.7D, left).
For the two conditions, we computed the sum of SIs (according to
Eq. 1) of individual muscles having a positive SI at condition (i). We
then subtracted the sum obtained at condition (ii) from the sum
obtained at condition (i). A Kolmogorov-Smirnov test was applied
to test the null hypothesis that data come from a normal distribu-
tion (P=0.06). We then performed a one-sample t test to test the
null hypothesis of zero mean on the total distribution.
Functional grasping behavioral task paradigm and
data analysis
The computation of neural modes coefficient matrix, the behavioral
experimental paradigm, and corresponding data analysis are detailed
in Supplementary Methods (see the “Neural modes coefficients
matrix computation,” “Brain-controlled stimulation of extension,”
and “Data processing and analysis” sections and the paragraphs
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
14 of 15
“Performance analysis in behavioral functional extension task,”
“PCs modulation analysis,” “PC-based stimulation control perfor-
mance,” and “DeepLabCut tracking”).
Statistics
All computed parameters were quantified and compared between
tested groups unless otherwise specified. All data are reported as
means±SEM unless specified otherwise. When bootstrap was ap-
plied, significance was analyzed using a one-sample two-sided t test
on the distribution of mean = 1 − 2 and =
√
_
1 + 2
_
2 without
correcting for sample size. When performing unpaired comparisons,
significance was evaluated using a nonparametric Kruskal-Wallis test
with Bonferroni correction for multiple comparisons. For paired
comparisons, the Wilcoxon signed-rank test was used.
SUPPLEMENTARY MATERIALS
www.science.org/doi/10.1126/scitranslmed.abg6463
Supplementary Methods
Figs. S1 to S13
Tables S1 to S8
References (71–76)
Movies S1 to S3
View/request a protocol for this paper from Bio-protocol.
REFERENCES AND NOTES
1. G. J. Snoek, M. J. IJzerman, F. A. C. G. in’t Groen, T. S. Stoffers, G. Zilvold, Use of the NESS
Handmaster to restore handfunction in tetraplegia: Clinical experiences in ten patients.
Spinal Cord 38, 244–249 (2000).
2. A. Prochazka, M. Gauthier, M. Wieler, Z. Kenwell, The bionic glove: An electrical stimulator
garment that provides controlled grasp and hand opening in quadriplegia. Arch. Phys.
Med. Rehabil. 78, 608–614 (1997).
3. D. Popović, A. Stojanović, A. Pjanović, S. Radosavljević, M. Popović, S. Jović, D. Vulović,
Clinical evaluation of the bionic glove. Arch. Phys. Med. Rehabil. 80, 299–304 (1999).
4. P. H. Peckham, K. L. Kilgore, M. W. Keith, A. M. Bryden, N. Bhadra, F. W. Montague, An
advanced neuroprosthesis for restoration of hand and upper arm control using
an implantable controller. J. Hand Surg. Am. 27, 265–276 (2002).
5. P. Taylor, J. Esnouf, J. Hobby, The functional impact of the freehand system on tetraplegic
hand function. Clinical results. Spinal Cord 40, 560–566 (2002).
6. C. Ethier, E. R. Oby, M. J. Bauman, L. E. Miller, Restoration of grasp following paralysis
through brain-controlled stimulation of muscles. Nature 485, 368–371 (2012).
7. C. T. Moritz, S. I. Perlmutter, E. E. Fetz, Direct control of paralysed muscles by cortical
neurons. Nature 456, 639–642 (2008).
8. A. B. Ajiboye, F. R. Willett, D. R. Young, W. D. Memberg, B. A. Murphy, J. P. Miller,
B. L. Walter, J. A. Sweet, H. A. Hoyen, M. W. Keith, P. H. Peckham, J. D. Simeral,
J. P. Donoghue, L. R. Hochberg, R. F. Kirsch, Restoration of reaching and grasping
movements through brain-controlled muscle stimulation in a person with tetraplegia:
A proof-of-concept demonstration. Lancet 389, 1821–1830 (2017).
9. P. H. Peckham, M. W. Keith, K. L. Kilgore, J. H. Grill, K. S. Wuolle, G. B. Thrope, P. Gorman,
J. Hobby, M. J. Mulcahey, S. Carroll, V. R. Hentz, A. Wiegner; Implantable Neuroprosthesis
Research Group, Efficacy of an implanted neuroprosthesis for restoring hand grasp
in tetraplegia: A multicenter study. Arch. Phys. Med. Rehabil. 82, 1380–1388 (2001).
10. D. A. Friedenberg, M. A. Schwemmer, A. J. Landgraf, N. V. Annetta, M. A. Bockbrader,
C. E. Bouton, M. Zhang, A. R. Rezai, W. J. Mysiw, H. S. Bresler, G. Sharma, Neuroprosthetic-
enabled control of graded arm muscle contraction in a paralyzed human. Sci. Rep. 7, 8386
(2017).
11. F. Rattay, Analysis of models for external stimulation of axons. IEEE Trans. Biomed. Eng. 33,
974–977 (1986).
12. S. A. Binder-Macleod, L. Snyder-Mackler, Muscle fatigue: Clinical implications for fatigue
assessment and neuromuscular electrical stimulation. Phys. Ther. 73, 902–910 (1993).
13. W. D. Memberg, K. H. Polasek, R. L. Hart, A. M. Bryden, K. L. Kilgore, G. A. Nemunaitis,
H. A. Hoyen, M. W. Keith, R. F. Kirsch, An implanted neuroprosthesis for restoring arm
and hand function in people with high level tetraplegia. Arch. Phys. Med. Rehabil. 95,
1201–1211.e1 (2014).
14. N. A. Brill, S. N. Naufel, K. Polasek, C. Ethier, J. Cheesborough, S. Agnew, L. E. Miller,
D. J. Tyler, Evaluation of high-density, multi-contact nerve cuffs for activation of grasp
muscles in monkeys. J. Neural Eng. 15, 036003 (2018).
15. W. Tigra, M. Dali, L. William, C. Fattal, A. Gélis, J.-L. Divoux, B. Coulet, J. Teissier, D. Guiraud,
C. Azevedo Coste, Selective neural electrical stimulation restores hand and forearm
movements in individuals with complete tetraplegia. J. Neuroeng. Rehabil. 17, 66 (2020).
16. R. Hugosdottir, S. Þ. Jónasson, H. Sigþórsson, Þ. Helgason, Feasibility study of a novel
electrode concept for a neuroprosthesis for augmentation of impaired finger functions.
Eur. J. Transl. Myol. 24, 4671 (2014).
17. J. Badia, T. Boretius, D. Andreu, C. Azevedo-Coste, T. Stieglitz, X. Navarro, Comparative
analysis of transverse intrafascicular multichannel, longitudinal intrafascicular and multi polar
cuff electrodes for the selective stimulation of nerve fascicles. J. Neural Eng. 8, 036023 (2011) .
18. P. H. Veltink, J. A. van Alste, H. B. K. Boom, Multielectrode intrafascicular and extraneural
stimulation. Med. Biol. Eng. Comput. 27, 19–24 (1989).
19. R. A. Gaunt, J. A. Hokanson, D. J. Weber, Microstimulation of primary afferent neurons
in the L7 dorsal root ganglia using multielectrode arrays in anesthetized cats: Thresholds
and recruitment properties. J. Neural Eng. 6, 055009 (2009).
20. S. Raspopovic, M. Capogrosso, J. Badia, X. Navarro, S. Micera, Experimental validation
of a hybrid computational model for selective stimulation using transverse intrafascicular
multichannel electrodes. IEEE Trans. Neural Syst. Rehabil. Eng. 20, 395–404 (2012).
21. N. M. Ledbetter, C. Ethier, E. R. Oby, S. D. Hiatt, A. M. Wilder, J. H. Ko, S. P. Agnew,
L. E. Miller, G. A. Clark, Intrafascicular stimulation of monkey arm nerves evokes
coordinated grasp and sensory responses. J. Neurophysiol. 109, 580–590 (2013).
22. K. Yoshida, K. Horch, Selective stimulation of peripheral nerve fibers using dual
intrafascicular electrodes. IEEE Trans. Biomed. Eng. 40, 492–494 (1993).
23. T. Boretius, J. Badia, A. Pascual-Font, M. Schuettler, X. Navarro, K. Yoshida, T. Stieglitz,
A transverse intrafascicular multichannel electrode (TIME) to interface with the peripheral
nerve. Biosens. Bioelectron. 26, 62–69 (2010).
24. S. Wurth, M. Capogrosso, S. Raspopovic, J. Gandar, G. Federici, N. Kinany, A. Cutrone,
A. Piersigilli, N. Pavlova, R. Guiet, G. Taverni, J. Rigosa, P. Shkorbatova, X. Navarro,
Q. Barraud, G. Courtine, S. Micera, Long-term usability and bio-integration of polyimide-
based intra-neural stimulating electrodes. Biomaterials 122, 114–129 (2017).
25. I. Delgado-Martínez, J. Badia, A. Pascual-Font, A. Rodríguez-Baeza, X. Navarro, Fascicular
topography of the human median nerve for neuroprosthetic surgery. Front. Neurosci. 10,
286 (2016).
26. T. M. Brushart, Central course of digital axons within the median nerve of Macaca
mulatta. J. Comp. Neurol. 311, 197–209 (1991).
27. B. Gesslbauer, L. A. Hruby, A. D. Roche, D. Farina, R. Blumer, O. C. Aszmann, Axonal
components of nerves innervating the human arm. Ann. Neurol. 82, 396–408 (2017).
28. S. Raspopovic, M. Capogrosso, S. Micera, A computational model for the stimulation of rat
sciatic nerve using a transverse intrafascicular multichannel electrode. IEEE Trans. Neural
Syst. Rehabil. Eng. 19, 333–344 (2011).
29. S. Romeni, G. Valle, A. Mazzoni, S. Micera, Tutorial: A computational framework
for the design and optimization of peripheral neural interfaces. Nat. Protoc. 15,
3129–3153 (2020).
30. D. Burke, Clinical uses of H reflexes of upper and lower limb muscles. Clin. Neurophysiol.
Pract. 1, 9–17 (2016).
31. M. A. Fisher, H reflexes and F waves Fundamentals, normal and abnormal patterns.
Neurol. Clin. 20, 339–360 (2002).
32. N. Greiner, B. Barra, G. Schiavone, H. Lorach, N. James, S. Conti, M. Kaeser, F. Fallegger,
S. Borgognon, S. Lacour, J. Bloch, G. Courtine, M. Capogrosso, Recruitment of upper-limb
motoneurons with epidural electrical stimulation of the cervical spinal cord. Nat.
Commun. 12, 435 (2021).
33. A. N. Sharpe, A. Jackson, Upper-limb muscle responses to epidural, subdural
and intraspinal stimulation of the cervical spinal cord. J. Neural Eng. 11, 016005 (2014).
34. U. S. Hofstoetter, S. M. Danner, B. Freundl, H. Binder, W. Mayr, F. Rattay, K. Minassian,
Periodic modulation of repetitively elicited monosynaptic reflexes of the human
lumbosacral spinal cord. J. Neurophysiol. 114, 400–410 (2015).
35. C. M. Oddo, S. Raspopovic, F. Artoni, A. Mazzoni, G. Spigler, F. Petrini, F. Giambattistelli,
F. Vecchio, F. Miraglia, L. Zollo, G. D. Pino, D. Camboni, M. C. Carrozza, E. Guglielmelli,
P. M. Rossini, U. Faraguna, S. Micera, Intraneural stimulation elicits discrimination of
textural features by artificial fingertip in intact and amputee humans. eLife 5, e09148 (2016).
36. T. Feix, J. Romero, H.-B. Schmiedmayer, A. M. Dollar, D. Kragic, The GRASP taxonomy
of human grasp types. IEEE Trans. Human-Mach. Syst. 46, 66–77 (2016).
37. T. Brochier, R. L. Spinks, M. A. Umilta, R. N. Lemon, Patterns of muscle activity underlying
object-specific grasp by the macaque monkey. J. Neurophysiol. 92, 1770–1782 (2004).
38. G. Valle, F. M. Petrini, I. Strauss, F. Iberite, E. D’Anna, G. Granata, M. Controzzi, C. Cipriani,
T. Stieglitz, P. M. Rossini, A. Mazzoni, S. Raspopovic, S. Micera, Comparison of linear
frequency and amplitude modulation for intraneural sensory feedback in bidirectional
hand prostheses. Sci. Rep. 8, 16666 (2018).
39. T. Kesar, L.-W. Chou, S. A. Binder-Macleod, Effects of stimulation frequency versus pulse
duration modulation on muscle fatigue. J. Electromyogr. Kinesiol. 18, 662–671 (2008).
40. M. Behringer, S. Grützner, J. Montag, M. McCourt, M. Ring, J. Mester, Effects of stimulation
frequency, amplitude, and impulse width on muscle fatigue. Muscle Nerve 53, 608–616 (2016).
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Badi et al., Sci. Transl. Med. 13, eabg6463 (2021) 27 October 2021
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE
15 of 15
41. B. Barra, M. Badi, M. G. Perich, S. Conti, S. S. M. Salehian, F. Moreillon, A. Bogaard, S. Wurth,
M. Kaeser, P. Passeraub, T. Milekovic, A. Billard, S. Micera, M. Capogrosso, A versatile
robotic platform for the design of natural, three-dimensional reaching and grasping tasks
in monkeys. J. Neural Eng. 17, 016004 (2020).
42. E. A. Pohlmeyer, L. R. Jordon, P. Kim, L. E. Miller, A fully implanted drug delivery system
for peripheral nerve blocks in behaving animals. J. Neurosci. Methods 182, 165–171 (2009).
43. J. A. Gallego, M. G. Perich, L. E. Miller, S. A. Solla, Neural manifolds for the control
of movement. Neuron 94, 978–984 (2017).
44. M. C. Cirstea, A. B. Mitnitski, A. G. Feldman, M. F. Levin, Interjoint coordination dynamics
during reaching in stroke. Exp. Brain Res. 151, 289–300 (2003).
45. K. L. Kilgore, P. H. Peckham, G. B. Thrope, M. W. Keith, K. A. Gallaher-Stone, Synthesis
of hand grasp using functional neuromuscular stimulation. IEEE Trans. Biomed. Eng. 36,
761 (1989).
46. F. M. Petrini, G. Valle, I. Strauss, G. Granata, R. D. Iorio, E. D’Anna, P. Čvančara, M. Mueller,
J. Carpaneto, F. Clemente, M. Controzzi, L. Bisoni, C. Carboni, M. Barbaro, F. Iodice,
D. Andreu, A. Hiairrassary, J.-L. Divoux, C. Cipriani, D. Guiraud, L. Raffo, E. Fernandez,
T. Stieglitz, S. Raspopovic, P. M. Rossini, S. Micera, Six-month assessment of a hand
prosthesis with intraneural tactile feedback. Ann. Neurol. 85, 137–154 (2019).
47. P. Raghavan, J. W. Krakauer, A. M. Gordon, Impaired anticipatory control of fingertip forces
in patients with a pure motor or sensorimotor lacunar syndrome. Brain 129, 1415–1425 (2006 ).
48. J. W. Krakauer, S. T. Carmichael, Broken Movement: The Neurobiology of Motor Recovery
After Stroke (MIT Press, 2017).
49. M. R. Popovic, T. A. Thrasher, M. E. Adams, V. Takes, V. Zivanovic, M. I. Tonack, Functional
electrical therapy: Retraining grasping in spinal cord injury. Spinal Cord 44, 143–151 (2006).
50. S. Mangold, T. Keller, A. Curt, V. Dietz, Transcutaneous functional electrical stimulation
for grasping in subjects with cervical spinal cord injury. Spinal Cord 43, 1–13 (2005).
51. J.-M. Gracies, Pathophysiology of spastic paresis. I: Paresis and soft tissue changes. Muscle
Nerve 31, 535–551 (2005).
52. J.-M. Gracies, Pathophysiology of spastic paresis. II: Emergence of muscle overactivity.
Muscle Nerve 31, 552–571 (2005).
53. D. K. Sommerfeld, E. U.-B. Eek, A.-K. Svensson, L. W. Holmqvist, M. H. von Arbin, Spasticity
after stroke. Stroke 35, 134–139 (2004).
54. J. W. McDonald, C. Sadowsky, Spinal-cord injury. Lancet 359, 417–425 (2002).
55. G. Kwakkel, B. J. Kollen, J. van der Grond, A. J. H. Prevo, Probability of regaining dexterity
in the flaccid upper limb. Stroke 34, 2181–2186 (2003).
56. V. Dietz, S. Grillner, A. Trepp, M. Hubli, M. Bolliger, Changes in spinal reflex and locomotor
activity after a complete spinal cord injury: A common mechanism? Brain 132,
2196–2205 (2009).
57. J. M. Johnston, Y. E. Cohen, H. Shirley, J. Tsunada, S. Bennur, K. Christison-Lagay,
C. L. Veeder, Recent refinements to cranial implants for rhesus macaques (Macaca
mulatta). Lab. Anim. 45, 180–186 (2016).
58. A. K. Vaskov, Z. T. Irwin, S. R. Nason, P. P. Vu, C. S. Nu, A. J. Bullard, M. Hill, N. North,
P. G. Patil, C. A. Chestek, Cortical decoding of individual finger group motions using ReFIT
Kalman filter. Front. Neurosci. 12, 751 (2018).
59. E. D’Anna, F. M. Petrini, F. Artoni, I. Popovic, I. Simanić, S. Raspopovic, S. Micera, A
somatotopic bidirectional hand prosthesis with transcutaneous electrical nerve
stimulation based sensory feedback. Sci. Rep. 7, 10930 (2017).
60. F. M. Petrini, M. Bumbasirevic, G. Valle, V. Ilic, P. Mijović, P. Čvančara, F. Barberi, N. Katic,
D. Bortolotti, D. Andreu, K. Lechler, A. Lesic, S. Mazic, B. Mijović, D. Guiraud, T. Stieglitz,
A. Alexandersson, S. Micera, S. Raspopovic, Sensory feedback restoration in leg amputees
improves walking speed, metabolic cost and phantom pain. Nat. Med. 25, 1356–1363 (2019).
61. F. M. Petrini, G. Valle, M. Bumbasirevic, F. Barberi, D. Bortolotti, P. Cvancara,
A. Hiairrassary, P. Mijovic, A. Ö. Sverrisson, A. Pedrocchi, J.-L. Divoux, I. Popovic, K. Lechler,
B. Mijovic, D. Guiraud, T. Stieglitz, A. Alexandersson, S. Micera, A. Lesic, S. Raspopovic,
Enhancing functional abilities and cognitive integration of the lower limb prosthesis.
Sci. Transl. Med. 11, eaav8939 (2019).
62. M. Capogrosso, F. B. Wagner, J. Gandar, E. M. Moraud, N. Wenger, T. Milekovic,
P. Shkorbatova, N. Pavlova, P. Musienko, E. Bezard, J. Bloch, G. Courtine, Configuration
of electrical spinal cord stimulation through real-time processing of gait kinematics.
Nat. Protoc. 13, 2031–2061 (2018).
63. S. Badoud, S. Borgognon, J. Cottet, P. Chatagny, V. Moret, M. Fregosi, M. Kaeser, E. Fortis,
E. Schmidlin, J. Bloch, J. F. Brunet, E. M. Rouiller, Effects of dorsolateral prefrontal cortex
lesion on motor habit and performance assessed with manual grasping and control
of force in macaque monkeys. Brain Struct. Funct. 222, 1193–1206 (2017).
64. M. Capogrosso, N. Wenger, S. Raspopovic, P. Musienko, J. Beauparlant, L. Bassi Luciani,
G. Courtine, S. Micera, A computational model for epidural electrical stimulation of spinal
sensorimotor circuits. J. Neurosci. 33, 19326–19340 (2013).
65. C. C. McIntyre, W. M. Grill, Selective microstimulation of central nervous system neurons.
Ann. Biomed. Eng. 28, 219–233 (2000).
66. C. C. McIntyre, W. M. Grill, Extracellular stimulation of central neurons: Influence of stimulus
waveform and frequency on neuronal output. J. Neurophysiol. 88, 1592–1604 (2002).
67. M. Capogrosso, J. Gandar, N. Greiner, E. M. Moraud, N. Wenger, P. Shkorbatova,
P. Musienko, I. Minev, S. Lacour, G. Courtine, Advantages of soft subdural implants
for the delivery of electrochemical neuromodulation therapies to the spinal cord.
J. Neural Eng. 15, 026024 (2018).
68. S. Raspopovic, F. M. Petrini, M. Zelechowski, G. Valle, Framework for the development
of neuroprostheses: From basic understanding by sciatic and median nerves models
to bionic legs and hands. Proc. IEEE 105, 34–49 (2017).
69. T. Boretius, K. Yoshida, J. Badia, K. Harreby, A. Kundu, X. Navarro, W. Jensen, T. Stieglitz,
in Biomedical Robotics and Biomechatronics (BioRob), 2012 4th IEEE RAS & EMBS Internat ional
Conference on Biomedical Robotics and Biomechatronics (IEEE, 2012), pp. 282–287.
70. S. Raspopovic, M. Capogrosso, F. M. Petrini, M. Bonizzato, J. Rigosa, G. Di Pino,
J. Carpaneto, M. Controzzi, T. Boretius, E. Fernandez, G. Granata, C. M. Oddo, L. Citi,
A. L. Ciancio, C. Cipriani, M. C. Carrozza, W. Jensen, E. Guglielmelli, T. Stieglitz,
P. M. Rossini, S. Micera, Restoring natural sensory feedback in real-time bidirectional
hand prostheses. Sci. Transl. Med. 6, 222ra19 (2014).
71. T. Stieglitz, H. Beutel, J.-U. Meyer, “Microflex”—A new assembling technique for
interconnects. J. Intell. Mater. Syst. Struct. 11, 417–425 (2000).
72. P. Čvančara, T. Boretius, V. M. López-Álvarez, P. Maciejasz, D. Andreu, S. Raspopovic,
F. Petrini, S. Micera, G. Granata, E. Fernandez, P. M. Rossini, K. Yoshida, W. Jensen,
J.-L. Divoux, D. Guiraud, X. Navarro, T. Stieglitz, Stability of flexible thin-film metallization
stimulation electrodes: Analysis of explants after first-in-human study and improvement of
in vivo performance. J. Neural Eng. 17, 046006 (2020).
73. Paul Cvancara, thesis, (2020).
74. T. Stieglitz, M. Schuettler, B. Rubehn, T. Boretius, J. Badia, X. Navarro, in 2011 5th
International IEEE/EMBS Conference on Neural Engineering, (2011), pp. 529–533.
75. A. Mathis, P. Mamidanna, K. M. Cury, T. Abe, V. N. Murthy, M. W. Mathis, M. Bethge,
DeepLabCut: Markerless pose estimation of user-defined body parts with deep learning.
Nat. Neurosci. 21, 1281–1289 (2018).
76. M. A. Schiefer, R. J. Triolo, D. J. Tyler, A model of selective activation of the femoral nerve
with a flat interface nerve electrode for a lower extremity neuroprosthesis. IEEE
Transactions on Neural Systems and Rehabilitation Engineering 16, 195–204 (2008).
Acknowledgments: We would like to thank B. Barra for help with the experiments and the
behavioral platform design; F. Lanz for help with anesthesia and surgery preparations;
G. Kohut and S. Durand for help in the dissection, anatomy characterization, and first acute
experiment; N. Greiner for help in designing the computational model; S. Conti for help with
the animal care and experiments; J. Maillard and L. Bossy for the care provided to the animals;
A. Gaillard and A. Francovich for help in developing the experimental setup; A. Zbinden for the
veterinary survey of the monkeys; and M. Kaeser for advice for the training of the monkeys.
Funding: We would like to acknowledge the financial support from the Wyss Center grants
(WCP 008 and W015-2016 to F.F. and S.P.L.), the Bertarelli Foundation (Catalyst Fund Grant to
S.M.), an Ambizione Fellowship (no. 167912 to M.C.), and the Swiss National Science
Foundation (no. 170032 and National Competence Center Research Robotics to S.M.). Author
contributions: S.M., E.M.R., S.W., M.B., and M.C. conceived the study. M.C., M.B., I.S., E.R., and
A.B. designed and implemented the experimental setup. I.S. and M.B. programmed the
electrophysiology and stimulation control software. P.C
̆. designed and fabricated the
intraneural implants with the support and supervision of T.S. M.B. implemented the
computational model. M.B. and S.W. performed neuroanatomy and histological procedures
and used the results to determine the dimension of the intraneural implants. F.F., M.B., and
S.W. designed the microfluidic cuff system, and F.F. manufactured the system and the
encapsulating pedestal with the support and supervision of S.P.L. M.B., A.B., and M.D. trained
the animal. G.C., J.B., D.K.S., A.B., S.W., S.B., M.B., and E.S. performed the surgeries. M.B., S.W., I.S.,
E.R., E.L., A.B., and S.B. conducted the experiments. M.B., S.W., E.R., E.L., and S.B. performed the
data analysis. M.B., E.R., and S.B. preprocessed the data. M.B., S.W., E.L., and E.R. prepared the
figures. M.B., S.W., M.C., and S.M. wrote the manuscript. All authors edited the manuscript.
S.M., M.C., and E.M.R. supervised the study. Competing interests: The authors declare that
they have no competing interests. Data and materials availability: All data associated with
this study are present in the paper or the Supplementary Materials and are available at https://
doi.org/10.5281/zenodo.5346351. The code used in the study is available at https://doi.
org/10.5281/zenodo.5346351.
Submitted 19 January 2021
Resubmitted 11 March 2021
Accepted 6 September 2021
Published 27 October 2021
10.1126/scitranslmed.abg6463
Citation: M. Badi, S. Wurth, I. Scarpato, E. Roussinova, E. Losanno, A. Bogaard, M. Delacombaz,
S. Borgognon, P. C
̆vanc̆ara, F. Fallegger, D. K. Su, E. Schmidlin, G. Courtine, J. Bloch, S. P. Laco ur,
T. Stieglitz, E. M. Rouiller, M. Capogrosso, S. Micera, Intrafascicular peripheral nerve stimulation
produces fine functional hand movements in primates. Sci. Transl. Med. 13, eabg6463 (2021).
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
Use of think article is subject to the Terms of service
Science Translational Medicine (ISSN ) is published by the American Association for the Advancement of Science. 1200 New York Avenue
NW, Washington, DC 20005. The title Science Translational Medicine is a registered trademark of AAAS.
Copyright © 2021 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim
to original U.S. Government Works
Intrafascicular peripheral nerve stimulation produces fine functional hand
movements in primates
Marion BadiSophie WurthIlaria ScarpatoEvgenia RoussinovaElena LosannoAndrew BogaardMaude DelacombazSimon
BorgognonPaul C#vanc#araFlorian FalleggerDavid K. SuEric SchmidlinGrégoire CourtineJocelyne BlochStéphanie P.
LacourThomas StieglitzEric M. RouillerMarco CapogrossoSilvestro Micera
Sci. Transl. Med., 13 (617), eabg6463. • DOI: 10.1126/scitranslmed.abg6463
It is TIME to move your hand
Upper limb paralysis can develop after spinal cord injury or stroke. Electrical stimulation has been used to partially
restore hand movements; however, current approaches using surface or intramuscular stimulation require challenging
surgeries and/or have limited efficacy and are associated with important adverse effects. Here, Badi et al. developed
an intraneural transverse intrafascicular multichannel electrode (TIME) system, composed of two electrodes for
stimulation of the median and radial nerve, that was able to restore hand movements in primates. In a proof-of-
principle experiment, one paralyzed monkey was able to perform hand movements using a brain-controlled TIME.
Intrafascicular stimulation might be used for generating and allowing fine hand movements in paralyzed patients.
View the article online
https://www.science.org/doi/10.1126/scitranslmed.abg6463
Permissions
https://www.science.org/help/reprints-and-permissions
Downloaded from https://www.science.org at Bibliotheque Cantonale et Universitaire - Fribourg on October 28, 2021
