SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

NEUROPROSTHETICS Copyright © 2018

The Authors, some
Proprioception from a neurally controlled rights reserved;
exclusive licensee
lower-extremity prosthesis American Association
for the Advancement
of Science. No claim
Tyler R. Clites,1,2 Matthew J. Carty,1,3 Jessica B. Ullauri,1 Matthew E. Carney,1,4 Luke M. Mooney,1 to original U.S.
Jean-François Duval,1 Shriya S. Srinivasan,1,2 Hugh. M. Herr1,2,3,4* Government Works

Humans can precisely sense the position, speed, and torque of their body parts. This sense is known as proprio-
ception and is essential to human motor control. Although there have been many attempts to create human-­
mechatronic interactions, there is still no robust, repeatable methodology to reflect proprioceptive information
from a synthetic device onto the nervous system. To address this shortcoming, we present an agonist-antagonist
myoneural interface (AMI). The AMI is composed of (i) a surgical construct made up of two muscle-tendons—an
agonist and an antagonist—surgically connected in series so that contraction of one muscle stretches the other
and (ii) a bidirectional efferent-afferent neural control architecture. The AMI preserves the dynamic muscle rela-
tionships that exist within native anatomy, thereby allowing proprioceptive signals from mechanoreceptors within
both muscles to be communicated to the central nervous system. We surgically constructed two AMIs within the
residual limb of a subject with a transtibial amputation. Each AMI sends control signals to one joint of a two-degree-­
of-freedom ankle-foot prosthesis and provides proprioceptive information pertaining to the movement of that
joint. The AMI subject displayed improved control over the prosthesis compared to a group of four subjects having

Downloaded from https://www.science.org on August 12, 2024

traditional amputation. We also show natural reflexive behaviors during stair ambulation in the AMI subject that
do not appear in the cohort of subjects with traditional amputation. In addition, we demonstrate a system for
closed-loop joint torque control in AMI subjects. These results provide a framework for integrating bionic systems
with human physiology.

INTRODUCTION isometrically to create padding for a prosthetic socket (14), severing

Proprioception is the sense of the relative spatial positioning of one’s the dynamic relationship between agonist-antagonist muscle pairs
body parts and of the amount of force exerted on the environment and limiting the ability of the muscle spindles and Golgi tendon
(1). It is essential to human motor control, gait adaptation, and joint organs within these tissues to communicate meaningful informa-
stability (2, 3). In humans, proprioceptive feedback is primarily me- tion to the central nervous system. Muscles, tendons, skin, bones,
diated by a complex relationship between sensory organs within and other tissues distal to the amputation site are typically dis-
muscles and tendons (1). Although a large body of literature exists carded, despite their potential capacity to contribute to reconstruc-
surrounding the biological structures involved in proprioceptive sen- tion of the amputated residuum. Nerves that cross the amputation
sation, including stretch receptors in the skin (4–6) and movement boundary are cut under tension and then buried into fat tissue or
receptors in the joints (7–9), there is substantial evidence highlighting deep in the residuum in an effort to prevent the formation of inap-
muscle spindles and Golgi tendon organs (10) as the predominant propriate nerve tissue growths, called neuromas, which can cause
mediators of joint proprioception (1). Muscle spindles and Golgi pain or other phantom sensations (15). Although sometimes effec-
tendon organs represent only a portion of the larger proprioceptive tive in preventing neuropathic pain, this technique creates a hurdle
system; however, studies in vibration-induced illusory kinesthesia for neural interfaces because of the limited longitudinal viability of
have indicated that isolated activation of muscle afferent receptors direct contact between synthetic interfaces and peripheral nerve tis-
is sufficient to promote sensations of joint position, movement, and sues, especially for neural recording (16, 17). Robotic prostheses
torque (11, 12). Further evidence indicates that the dynamic rela- have been designed around these limitations and fall short of repro-
tionships within agonist-antagonist muscle pairs are fundamental to ducing the biological control experience. The present state-of-the-
natural sensations of joint movement (13). The complexity of this art commercial technology for persons with below-knee amputation
afferent (neural pathways that relay information from a muscle or is a powered ankle joint that is unable to fully reproduce the mo-
other end organ to the central nervous system) feedback system tions of the biological ankle and subtalar joints and does not have
poses a challenging hurdle for the development of bionic limbs that any direct connection to the nervous system (18).
benefit from bidirectional neural communication. Many attempts have been made to overcome the limitations of state-
The clinical standard of care for limb amputation surgery has not of-the-art amputation surgical techniques and commercial prosthetic
changed in almost two centuries and is not currently optimized to systems. Direct stimulation of upstream peripheral nerves through im-
facilitate neural integration with bionic limbs. In a typical amputa- plantable electrodes has shown great promise in restoring cutaneous
tion procedure, muscle tissues in the residual limb are configured touch perception and, in some cases, isolated kinesthetic sensations
(19–25). However, partly because of a mismatch between the complex-
1
Center for Extreme Bionics, Massachusetts Institute of Technology (MIT) Media ity of proprioceptive afferent signaling and the relatively low resolu-
Lab, Cambridge, MA 02139, USA. 2Harvard–MIT Division of Health Sciences and tion and precision of implantable stimulation methodologies, none of
Technology, MIT, Cambridge, MA 02139, USA. 3Division of Plastic and Reconstruc- these approaches is engineered to provide, with high-probability, sta-
tive Surgery, Brigham and Women’s Hospital, Boston, MA 02115, USA. 4Depart-
ment of Media Arts and Sciences, MIT, Cambridge, MA 02139, USA. ble and natural proprioceptive percepts. Vibration-induced illusory
*Corresponding author. Email: hherr@media.mit.edu kinesthesia (12) has been explored as a means of providing joint state

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 1 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

information through activation of cutaneous stretch receptors; un- A3 and A4). In each AMI, each muscle was mechanically linked to
fortunately, translation of this approach has been a major hurdle. Re- its partner via a tendon, which passed through a synovial canal, har-
generative peripheral nerve interfaces have emerged as a means of vested from the amputated ankle joint at the time of amputation.
stifling neuroma formation, preventing phantom pain, increasing One synovial canal was anchored to the medial flat of the tibia for
the number of independent neural control targets, and conveying cu- each AMI and served as a biological pulley for that AMI, enabling the
taneous sensory information (26, 27). Targeted muscle reinnervation tendon to slide relative to the anchored sheath such that force pro-
has a strong track record of improving controllability of myoelectric duction in one muscle caused stretch in the other. The AMI muscles
prostheses but is not designed to close the control loop with proprio- were surgically coapted (connected via suture) with each muscle set
ceptive sensation (28, 29). Prosthetic hardware has also seen dramatic at its resting tension, such that the default sensory state of each AMI
improvement in recent years (29–35), including substantial improve- reflected a neutral joint position.
ments to myoelectric control architectures (20, 27–29, 34, 36). At about 1-year postoperation, ultrasound imaging was used to
As a methodology of improving efferent (neural pathways that re- interrogate motion of each AMI during volitional cyclical movement
lay commands from the central nervous system to a muscle or other of the phantom limb. Movement commands were communicated to
end organ) prosthetic control and providing afferent proprioceptive the subject in terms of phantom limb motion (“dorsiflex your ankle”),
sensation, we present an agonist-antagonist myoneural interface rather than activation of a specific muscle (“contract your tibialis
(AMI). An AMI is made up of an agonist and an antagonist muscle-­ anterior”). Fascicle strains were estimated from ultrasound video,
tendon connected mechanically in series: When the agonist con- recorded from the antagonist muscle as the subject volitionally con-
tracts, the antagonist is stretched and vice versa (37, 38). The purpose tracted the agonist. Electromyography (EMG) was simultaneously
of an AMI is to control and interpret proprioceptive feedback from recorded from the contracting agonist. Ultrasound fascicle data
a bionic joint. This approach was first validated in several experiments showed physiologically relevant strains (up to 16%) in the antago-

Downloaded from https://www.science.org on August 12, 2024

using animal models. An AMI was constructed from musculature nist during volitional activation of the agonist. Cross-correlation of
with intact innervation and vascular supply in a rat distal hindlimb the agonist-­integrated EMG signal and antagonist fascicle strain
(39). Graded afferent signals were then recorded from the agonist showed a strong relationship between agonist muscle activation and
muscle’s innervation nerve during closed-loop functional electrical antagonist muscle stretch, with a correlation coefficient of 0.94 for
stimulation of the antagonist, demonstrating the capacity of the AMI inversion (Fig. 1B) and 0.91 for plantar flexion (Fig. 1C). Coupled
to provide natural proprioceptive feedback. In another murine study, motion was also preserved during cyclical alternating contraction
we demonstrated that a functional AMI could be constructed from of the agonist and the antagonist muscles at both low and high fre-
small denervated and devascularized muscle grafts placed in the vi- quencies (movie S1). Additional video was recorded with the ultra-
cinity of transected motor nerves (40). A caprine experiment fur- sound probe positioned adjacent to the synovial canals; these videos
ther validated that the principles demonstrated in (39) are scalable showed sliding along the medial tibia, confirming coupled move-
to larger animal models (41). ment within each AMI (movie S1).
On the basis of these previous studies, we hypothesized that the
AMI procedure and rehabilitation protocol would enable improved Control architecture: Prosthesis not in the loop
control of a multi-degree-of-freedom prosthesis while reflecting nat- All experiments were carried out using a prototype ankle-foot pros-
ural proprioceptive sensation pertaining to each prosthetic joint onto thesis with powered ankle and subtalar joints (see the “Prosthetic hard-
the central nervous system. In the case study presented herein, we test ware design” section in the Supplementary Materials for a description
this hypothesis in a human subject having a unilateral transtibial am- of the robotic hardware). When the AMI subject (subject A) wishes
putation. We first describe the implementation of two AMIs within to move the bionic limb, he contracts the AMI muscles associated
the subject’s residuum and connection via synthetic electrodes to with his intended motion. Muscle activation is estimated from EMG
an external prosthetic leg with powered artificial ankle and subtalar collected via four bipolar surface electrodes on the surface of the skin,
joints. The ability of this AMI patient (subject A) to volitionally and where each electrode is affixed adjacent to one of the four muscles
reflexively control the prosthesis in free space was evaluated and com- comprising the two AMIs. These estimates are used to independently
pared to a cohort of four participants having traditional transtibial control position and impedance (mechanical stiffness) of the pros-
amputation (group T, composed of subjects T1 to T4). We conclude thetic ankle and subtalar joints (for a complete discussion of EMG
with an evaluation of a functional electrical stimulation (FES) meth- processing and efferent control architecture, see the “Efferent con-
odology for providing torque feedback from a prosthesis to the pe- trol architecture” section and fig. S1). Because the AMI agonist and
ripheral nervous system of subject A. antagonist muscles are mechanically coupled within the residual limb
of our subject, volitional contraction of an agonist passively stretches
that muscle’s antagonist. The natural neural responses from muscle
RESULTS spindles within both muscles are then interpreted by the central ner-
AMI surgical construction vous system as sensations of joint position and speed, associated with
Two AMIs were constructed in the residual limb of a 53-year-old movement of the prosthesis (Fig. 2A). During volitional movement
male at the time of his elective unilateral transtibial amputation. One of his phantom limb, subject A reported natural proprioceptive sen-
AMI, composed of the tibialis posterior and the peroneus longus, sation throughout his phantom joint space, closely matching move-
was designed to control the bionic subtalar joint responsible for ment of the prosthesis.
prosthetic inversion and eversion movements (Fig. 1, A1 and A2). In this free-space control architecture, there is no direct feedback
A second AMI, composed of the lateral gastrocnemius and the tibialis line from the prosthesis to the AMI (hence, “prosthesis not in the
anterior, was designed to control the bionic ankle joint, responsible loop”). However, the subject receives proprioceptive afferent feedback
for prosthetic plantar flexion and dorsiflexion movements (Fig. 1, describing his intended movement command through agonist-antagonist

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 2 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

Fig. 1. Agonist-antagonist myoneural interface. (A) Two AMIs were surgically constructed within the left leg residuum of a patient to enable control of prosthetic Downloaded from https://www.science.org on August 12, 2024
subtalar and ankle joint movements. Prosthetic subtalar and ankle movements are shown in (A1) and (A2), and (A3) and (A4), respectively. In (A1), the prosthetic subtalar
joint everts (arrow) when the peroneus longus contracts, stretching the tibialis posterior; in (A2), the subtalar joint inverts (arrow) when the tibialis posterior contracts,
stretching the peroneus longus. In (A3), the prosthetic ankle joint dorsiflexes (arrow) when the tibialis anterior contracts, stretching the lateral gastrocnemius; in (A4),
the ankle joint plantar-flexes (arrow) when the lateral gastrocnemius contracts, stretching the tibialis anterior. Dashed arrows indicate muscle contraction and stretch.
(B) Ultrasound strain and EMG data for the subtalar AMI, showing coupled motion when the peroneus longus is stretched during volitional contraction of the tibialis
posterior [inversion movement (A2)]. The correlation coefficient of these two signals is 0.94. (C) Ultrasound strain and EMG data for the ankle AMI, showing coupled
motion when the tibialis anterior is stretched during volitional contraction of the lateral gastrocnemius [plantar flexion movement (A4)]. The correlation coefficient of
these two signals is 0.91. (B) and (C) are representative traces from subject A (n = 5 trials per motion). EMG values are normalized to calibrated maxima for each muscle.

stretch relationships within the AMI. In free space, where no external limited functional difference between sensations of intended and ac-
torques are applied to the prosthetic joints, the efferent control system tual joint motion.
is designed to ensure that movement of the prosthesis is reliably syn- Before beginning the experiments, we found that by tuning con-
chronized with these natural afferent sensations; in this way, there is troller gains, we could adjust sensitivity of the prosthesis to make it

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 3 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

Fig. 2. Volitional control of joint position and impedance. (A) Schematic showing how subject A activates the AMI muscle associated with his intended motion. This Downloaded from https://www.science.org on August 12, 2024
activation is recorded as EMG and generates a movement command for the motors within the prosthesis. The subject can stiffen a prosthetic joint by simultaneously
coactivating both the agonist and the antagonist muscles within the AMI associated with that joint. Afferent signals describing prosthetic joint movement are communicated
to the patient’s nervous system via muscle spindle response to differential stretch relationships within each AMI muscle. (B) Average performance maps for volitional
control tasks (n = 100 samples from subject A, n = 350 samples from group T). The scores for each metric are presented by target area; the location of each rectangle
within the axis represents the target area in joint space, ranging from full plantar flexion (PF) to full dorsiflexion (DF) and from full eversion (EV) to full inversion (IN). The
shade of the rectangle indicates the subject’s score in that target area, where lighter shades are indicative of better performance. (C) Representative sample traces of joint
position (angle), EMG, and ankle stiffness during free-space volitional control experiments for subject A (n = 100 total samples) and one subject from group T (subject T2,
n = 50 total samples). Dashed vertical lines divide the trial into segments by target motion, indicated by the text at the top of each segment. The shaded region of each
plot represents the portion of that trial in which the subject was instructed to stiffen the joint. The range of ankle angles shown is the full range of the prosthetic ankle:
from 15 degrees of PF to 10 degrees of DF. The range of subtalar angles shown is the full range of the prosthetic subtalar: from 15 degrees of EV to 15 degrees of IN. Ankle
and subtalar angle plots show target position (black) and actual position (purple). The ankle EMG plot shows signal recorded from the lateral gastrocnemius (light blue)
and the tibialis anterior (dark blue). The subtalar EMG plot shows signal recorded from the tibialis posterior (light green) and the peroneus longus (dark green). EMG values
are normalized to calibrated maxima for each muscle. Stiffness values are normalized such that a value of 1 represents coactivation of the tibialis anterior and the lateral
gastrocnemius at each muscle’s calibrated maximum.

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 4 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

more or less reactive to muscle activity than the phantom limb. If the target locations consistently more difficult than others (Fig. 2B; per-
gains were too high, our subject described movement of the prosthesis formance metrics for each individual subject are reported in table S1).
as “jumpy.” Conversely, if they were too low, he described the pros- Path nonideality indicates the distance in angle space traversed
thesis as “sluggish” and “nonresponsive.” Once the gains were well by the prosthetic joints during the initial movement of the prosthesis
tuned, movement of the prosthesis and perceived movement of the from the rest angle to the target angle (task 1), normalized to the
phantom limb came into alignment. Control subjects having tradi- ideal distance from the rest angle to the center of the target square.
tional unilateral transtibial amputation (group T), using the same pros- Better performance in this metric is indicated by a lower score. Sub-
thesis under identical conditions, did not report similar sensations. ject A’s path nonideality score was 1.65 compared to an average score
Despite ample gain adjustments and tuning, none of the control sub- of 2.7 (±0.45) for subjects in group T. This represents a 39% im-
jects felt that motion of the prosthetic joints closely matched sensation provement in performance. The average path nonideality score for
in the phantom limb. One subject (subject T2) specifically attributed the intact limb cohort was 1.68 (±0.87), and the score for subject A’s
this discrepancy in part to an unintended simultaneous antagonis- unaffected limb was 1.56.
tic cocontraction during volitional activation of muscles within his Time in target indicates ability to hold the prosthesis in the tar-
residuum. All subjects in group T described a perception of limited get window and is reported as the total time for which each subject
motion throughout their phantom joint space. maintained the prosthesis within the target during the 3-s hold task
Independent control of joint position and impedance (task 2). Better performance in this metric is indicated by a higher
For all subjects, volitional control experiments were carried out after score. Subject A’s time in target score was 2.04 s compared to an aver-
about 1 hour of tuning and free control of the device. These experi- age score of 1.53 s (±0.30) for subjects in group T. This represents a
ments evaluated each subject’s ability to independently modulate 33% improvement in performance. The average time in target score
prosthetic joint position and impedance while performing volitional for the intact limb cohort was 2.16 s (±0.33), and the score for subject

Downloaded from https://www.science.org on August 12, 2024

control tasks with the prosthesis in free space. A graphical interface A’s unaffected limb was 2.49 s.
was generated to visualize prosthesis joint angles in real time as a Total wasted motion provides insight into stability and movement
point location in two-dimensional joint space (plantar flexion and efficiency during joint motion and stiffening throughout all active
dorsiflexion on the vertical axis and inversion and eversion on the portions of the trial (tasks 1 to 3). This is reported as the total angle-­
horizontal). While wearing the prosthesis and watching the graphi- space distance traversed by the prosthetic joints, normalized to the
cal interface, each subject was instructed to complete the following minimum travel distance required to complete the tasks. Better per-
tasks: (i) move the prosthesis to within a predetermined window of formance in this metric is indicated by a lower score. Subject A’s total
joint angles, represented graphically as a rectangle in joint angle wasted motion score was 7.45 compared to an average score of 21.74
space; (ii) hold the prosthesis within this joint angle window for 3 s; (±2.68) for subjects in group T. This represents a 66% improvement
(iii) stiffen the prosthesis by cocontracting the AMI muscles, while in performance. The average total wasted motion score for the intact
maintaining the joint position within the joint angle window; and limb cohort was 8.79 (±3.17), and the score for subject A’s unaffected
(iv) return the prosthesis to its rest position. limb was 7.33.
After being introduced to the experimental paradigm and allowed The representative sample traces in Fig. 2C qualitatively highlight
one practice attempt, each subject repeatedly performed tasks 1 to 4 at the improved stability and path efficiency of subject A compared to
10 different locations across the full prosthesis joint angle space, in a subjects in group T during these volitional control experiments. These
randomly generated order. All trials were carried out within the same trends were also apparent in gait-related tasks requiring volitional
2-hour session. Three metrics were selected to evaluate performance control (Fig. 3 and movie S2). While wearing the prosthesis, subject A
during volitional control experiments. For each metric, each subject’s and each subject from group T were asked to step on the side of a 4-cm
average performance was calculated across all trials performed by block placed in their path, such that the lateral edge (outside) of the
that subject to give an overall subject score. Scores for subject A (n = prosthetic foot was in contact with the block, whereas the medial edge
100 samples from 10 trials) were compared to the average of the four (inside) of the foot remained in contact with the floor. The block was
subject scores from group T (n = 350 samples from 35 trials). placed on the floor at the location of expected foot strike of the sub-
Identical experiments were also performed on a cohort of limbs ject’s affected leg to force the prosthetic foot into an everted position
with intact biological anatomy (n = 390 samples from 39 trials). This (Fig. 3). All subjects were uniformly instructed to volitionally move
cohort included three of the four unaffected limbs from subjects in the prosthetic ankle and subtalar joints during the swing phase of
group T, as well as one limb of an additional subject with two intact a single step such that the prosthetic subtalar would be everted ap-
biological limbs. For these experiments, subjects controlled the pros- propriately for contact with the block. In subject A, we observed
thesis with EMG signals measured from surface electrodes placed volitional repositioning of the subtalar into full eversion during the
over the muscles in their intact biological limb. In this way, the per- swing phase, consistent across all trials (n = 10). Swing-phase behav-
formance metrics for this cohort are not meant to characterize the iors within group T were nonuniform, with high intersubject vari-
fully biological control system, but instead to capture performance ability (n = 32 trials). Late swing eversion, defined as the maximum
of ideal muscular anatomy working in concert with the designed eversion angle achieved between 80 and 100% of the swing phase,
robotic platform. In addition, scores from subject A’s unaffected leg was calculated for each trial. These values were then averaged to give
(n = 100 samples from 10 trials) were compared with scores from the an overall subject score. Subject A averaged 8.8 degrees of eversion,
intact limb cohort to account for the possibility that subject A was whereas the average score for group T was 4.8 (±5.9) degrees of in-
uncharacteristically skilled at the particular experimental task. version. Summary data are reported in Table 1.
The performance metrics used to assess volitional prosthetic con- Reflexive behaviors
trol were path nonideality, time in target, and total wasted motion. Reflexive activity was evaluated during stair ascent and descent tasks,
All subjects successfully completed all tasks at all locations, with some in which humans reflexively modulate swing-phase joint angle (42, 43).

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 5 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

Downloaded from https://www.science.org on August 12, 2024

Fig. 3. Simultaneous subtalar and ankle control during a gait task requiring volitional eversion. Joint position and EMG during the swing phase of gait, as subject
A (n = 10 trials) and each subject from group T (n = 32 trials) step onto the side of a block positioned on the floor to require eversion (arrow) of the prosthetic subtalar joint,
as shown in the schematic. Shaded traces indicate mean ± 1 SD. Positive and negative subtalar angles correspond to eversion (EV) and inversion (IN), respectively. Positive
and negative ankle angles correspond to dorsiflexion (DF) and plantar flexion (PF), respectively. The subtalar EMG plot shows signal recorded from the peroneus longus
(light green) and the tibialis posterior (dark green). The ankle EMG plot shows signal recorded from the lateral gastrocnemius (light blue) and the tibialis anterior (dark
blue). EMG values are normalized to calibrated maxima for each muscle.

Subject A and all subjects from group T were instructed to walk as (T3) consistently plantar-flexed beginning before toe-off, and the
naturally as possible and to avoid active volitional movement of the degree of plantar flexion lessened as the subject moved through the
prosthesis. Instructions were carefully designed and delivered uni- swing phase. Late swing plantar flexion was defined as the maximum
formly with intent to be clear, concise, consistent, and free from bias. plantar flexion angle achieved between 80 and 100% of the swing
While ascending stairs, subject A (n = 10 trials) first reflexively plantar-­ phase. Subject A averaged 11.9 degrees of plantar flexion compared
flexed the prosthetic ankle as the prosthesis left the ground and then to 2.3 (±3.2) degrees of plantar flexion in group T. Summary data are
dorsiflexed during swing to appropriately position the foot before reported in Table 1.
placing it on the step (Fig. 4A and movie S3). He described these
actions as automatic. These behaviors were not observed in subjects Control architecture: Prosthesis in the loop
from group T (n = 32 trials). Late swing dorsiflexion, defined as the The final set of experiments was designed to evaluate whether FES
maximum dorsiflexion angle achieved between 80 and 100% of the can provide usable torque information from the prosthetic device to
swing phase, was calculated for each trial, and comparisons were a subject having AMIs. To close the control loop around the pros-
made as above. Subject A averaged 7.3 degrees of dorsiflexion com- thesis, afferent feedback of prosthetic joint torque was provided to
pared to 7.0 (±3.8) degrees of plantar flexion in group T. Summary subject A through stimulation of the AMI muscles (Fig. 5A). In re-
data are reported in Table 1. sponse to torque measured on the prosthesis, microprocessors on
While descending stairs (prosthetic leg leading), subject A exhib- the bionic leg commanded artificial stimulations to the antagonist
ited plantar flexion in late swing to prepare for foot-ground contact muscle within each AMI, controlling the force borne on the mechan-
(Fig. 4B and movie S3). This behavior is fundamental to normalized ically coupled agonist. To validate this feedback modality in isola-
stair-descent gait (42, 43). Late swing plantar flexion was not appre- tion, stimulation was first applied to the tibialis anterior—the muscle
ciable in three of the four subjects from group T. The fourth subject linked to prosthetic dorsiflexion—in absence of the prosthesis. Subject

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 6 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

was asked to indicate which of the two stimuli he perceived as stronger.

Table 1. Summary data for terrain traversal trials. The metric in the Both subject A and the experimenter were blinded to stimulus ampli-
second column was calculated for each trial of the task named in the first tudes. A cumulative normal distribution was fit to the raw discrimina-
column and averaged within each subject to give an overall subject score. tion data to obtain a psychometric function. Two estimates of JND
Subject A’s overall subject score for each task (from n = 10 trials per task)
were calculated from this function as the change in stimulus amplitude
is reported in the third column. The fourth column reports mean ± 1
intersubject SD for group T (n = 4 subjects, n = 32 total trials per task). Late that resulted in 75% judgment accuracy: one for increases relative to
swing eversion, late swing dorsiflexion, and late swing plantar flexion the reference value and the other for decreases relative to the reference
were calculated as the maximum eversion, dorsiflexion, and plantar value. These two values were then averaged to give a single estimate of
flexion angles, respectively, achieved between 80 and 100% of the swing JND. The psychometric curve for subject A was smooth, and the JND
phase of the relevant task. specific to this reference amplitude was 0.065 mA (Fig. 5C).
Task Metric
Subject A
Group T (n = 4)
Closed-loop torque control
(n = 1) Afferent feedback of prosthetic torque through stimulation of the
Eversion block Late swing 8.8 degrees of 4.8 (±5.9) degrees AMI antagonist improved performance during torque control tasks.
eversion eversion of inversion In these experiments, subject A plantar-flexed the prosthetic ankle in
Stair ascent Late swing 7.3 degrees of 7.0 (±3.8) degrees response to verbal commands of percent effort (25, 50, 75, and 100%),
dorsiflexion dorsiflexion of plantar thereby applying torque to a linear rotary-spring foot pedal. Sample
flexion
trial plots (Fig. 5D) show the relationship between EMG activity,
Stair descent Late swing 11.9 degrees 2.3 (±3.2) degrees torque measured on the foot pedal, and stimulation amplitude. Sum-
plantar of plantar of plantar
mary results from these experiments are shown in Fig. 5E. With stim-
flexion flexion flexion
ulation, subject A consistently generated four distinct torques at each

Downloaded from https://www.science.org on August 12, 2024

of the four effort levels (P < 0.025, Tukey-Kramer, n = 79). Without
muscle stimulation feedback, the torques generated at 50 and 75%
effort and 75 and 100% effort did not differ significantly (P > 0.1,
A qualitatively described the sensation associated with this stimulation Tukey-Kramer, n = 79). Torques produced at 25, 50, and 100% effort
as “standing at the edge of a step, with [his] weight pushing down,” were significantly different between the stimulation on and stimula-
forcing his phantom ankle into a dorsiflexed state. By volitionally ac- tion off cases (P < 0.02, t test), whereas those produced at 75% effort
tivating his calf muscles, he counteracted the perceived dorsiflexion did not show a significant difference (P = 0.73). In his unaffected
and felt the phantom ankle return to a neutral position, as if he had limb, subject A generated significantly different torques at each of
done a calf-raise exercise while still standing at the step’s edge. Subject the four effort levels (P < 0.001, Tukey-Kramer, n = 80).
A acknowledged the absence of cutaneous sensation and described
perceiving the stimulation as “involuntary contraction” in the artifi- Descriptions of the control experience
cially stimulated tibialis anterior. However, he felt that these were mi- After each session, study participants were asked to comment on the
nor distractions, to which he would grow accustomed with repeated experience of controlling the prototype prosthesis. Subject A described
use of the prosthesis. feeling as if the prosthesis was “his leg,” referring to his missing biolog-
Characterization of perception ical limb. He explained, “My [conventional] prosthesis doesn’t have
Two psychometric evaluations were performed to quantify subject the same sort of animation to it. This feels like it’s alive.” Over the
A’s perception of torque intensity in absence of the prosthesis. First, course of this study, we observed subject A’s candid interactions with
a magnitude estimation experiment was carried out in a manner sim- the prosthesis during experimental downtime. On one occasion, at
ilar to experiments previously described in the literature (19, 25). the end of the first trial day, we noticed that he was unconsciously
Stimulation was delivered to the tibialis anterior at randomly selected fidgeting with the prosthetic foot while seated and engrossed in con-
current amplitudes of integer values between 0 and 4 mA. Subject A versation (movie S4). On the second trial day, after standing on the
was blinded to all stimulation parameters throughout the experiment. device for only a few minutes, we watched as he wiggled his pros-
During each trial, the subject was instructed to remain at rest until he thetic foot to dislodge a roll of tape that had adhered to the bottom
felt stimulation pulling his phantom ankle into a dorsiflexed position. of his shoe (movie S4). These small behaviors provide evidence to
He was then asked to counteract this perceived ankle torque by voli- support the subject’s claim that the prosthesis had become embodied.
tionally plantar-flexing his phantom ankle until the perceived joint Two days after the first trial day, in an email sent spontaneously to the
angle returned to its neutral state. After stimulation had subsided, the research team, subject A explained, “Two days later and what trans-
subject verbally rated the magnitude of perceived torque (Fig. 5B). pired is still slowly sinking in. I keep trying to describe the sensation
There was a significant correlation between perceived dorsiflexion to people. Then this morning [my daughter] asked me if I felt like a
torque and stimulation amplitude (P < 0.0001, R2 = 0.96, n = 25). cyborg. The answer was ‘no, I felt like I had a foot’. I think that in just
In the second psychometric evaluation, a forced-choice paradigm the short time I had it wired in and mounted to me it was quickly
was used to establish the just-noticeable difference (JND) for stimu- becoming part of me.”
lation intensity, following the protocol outlined in (25). During each Subjects from group T described a remarkably different subjec-
trial, a pair of stimuli were applied to the tibialis anterior in a pseu- tive control experience. Subjects T1 and T4 both felt that their inter-
dorandom order; one of the two stimuli was delivered at a reference action with the device was similar to the interaction one might have
amplitude (2 mA), and the other was delivered at 1 of 11 possible stim- playing a video game for the first time. Subject T2 explained that the
ulus values ranging from 0 to 4 mA. Each pair of stimuli was pre- prosthesis sometimes “behaved in a way that was somewhat surpris-
sented a total of 20 times in a pseudorandomly generated order, for a ing” and acknowledged that he felt “a bit of disconnect” with the
total of 220 individual trials. After each pair was presented, subject A device. He postulated that this disconnect would shrink over time, as

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 7 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

Downloaded from https://www.science.org on August 12, 2024

Fig. 4. Reflexive control during stair tasks. Ankle position and EMG while each subject (A) ascends and (B) descends stairs. Shaded traces indicate mean ± 1 SD for
subject A (n = 10 trials for each of ascent and descent) and each subject from group T (n = 32 trials for each of ascent and descent). The ankle EMG plots show signal re-
corded from the lateral gastrocnemius (light blue) and the tibialis anterior (dark blue). Arrow indicates direction of movement. EMG values are normalized to calibrated
maxima for each muscle.

he learned to control the joints in a more predictable way. Upon fur- constructed in his transtibial residuum. This subject (subject A) showed
ther questioning, he revealed that his connectedness with any prosthe- improved stability and motion path efficiency in free-space volitional
sis was directly linked to the “sensation he received from it.” Although control tasks, as compared to the cohort of four subjects having tradi-
this subject was pleased to be able to feel the device moving, which he tional transtibial amputation (group T). While ascending and de-
perceived through shifts in momentum and vibrations carried through scending stairs, subject A also demonstrated reflexive swing-phase
his socket, he noted that these sensations were only present while the behaviors that were absent in group T. In addition, we characterized
joints were in motion. In his words, “I can feel it in the passage from a methodology for closed-loop torque control with afferent proprio-
point A to point B, but once it’s at point B, or once it’s resting at point ceptive feedback of joint torque from a prosthetic limb in persons
A, there’s no sensation.” Subject T3 described “not really trusting” the having one or more AMIs. This feedback improved performance on
device. Universal to the correspondence of subjects from group T was torque control tasks.
a distinct lack of ownership of the prosthesis or emotion associated One possible explanation of performance gaps between subject
with controlling it. The discrepancy in experience between subject A A and group T during volitional control tasks is a lack of fine con-
and group T may highlight the fundamental role of natural afferent trol over residual muscle activation in the latter group. Several of the
sensation in prosthesis embodiment (44–47). subjects in group T described involuntary cocontraction as a prom-
inent source of efferent control difficulty; accompanying volitional
activation of a muscle in the residual limb is a consistent unintended
DISCUSSION contraction in that muscle’s antagonist. Consequently, these subjects
Proprioceptive sensation pertaining to a synthetic appendage was re- must increase the volitional activation of their agonist to overpower
flected onto the nervous system of a subject with two AMIs surgically the unintended antagonistic activation. This likely played a role in a

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 8 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

Downloaded from https://www.science.org on August 12, 2024

Fig. 5. Closed-loop torque control. (A) Schematic of the prosthesis-in-the-loop control architecture, in which afferent feedback of prosthetic joint torque is provided via
FES of the antagonist muscle. The patient perceives this stimulation as a natural sensation of ankle torque. (B) Magnitude estimation of perceived dorsiflexion torque as
a function of stimulation current delivered to the tibialis anterior. Perceived torques are normalized to the maximum reported value. For clarity in plotting, each point
represents the mean value of five independent trials. Error bars represent the SE, and the R2 coefficient reported on the plot is that of the mean values. (C) Discrimination
performance as a function of differences in stimulation current. The reference current for all forced choice trials was 2 mA. Points indicate percentage of test stimuli cor-
rectly identified as stronger or weaker than the reference over 20 pairwise trials, and the green line represents a cumulative normal distribution fit to the raw data.
(D) Representative sample traces of lateral gastrocnemius EMG (blue), torque (purple), and stimulation current (green) during closed-loop torque control trials for the
“stimulation on” (n = 79 total trials) and “stimulation off” (n = 79 total trials) cases. Numbers at the top of the plot correspond to percent effort commands. Stimulation cur-
rents are normalized to 9 mA. EMG values are normalized to calibrated maxima for each muscle. (E) Summary data for closed-loop torque control trials in each of the stim-
ulation on (n = 79 trials), stimulation off (n = 79 trials), and “unaffected limb” (n = 80 trials) cases. An asterisk above a bar indicates that the bar is significantly different from
all other bars in the plot (P < 0.025). Where no significance was seen, a P value for the comparison is shown. Error bars represent a 99.9% confidence interval on the mean.

perception that EMG output was binary (on or off) and in the instability tion is the result of an effort to stabilize the residual limb within the
that plagued all subjects in group T while attempting to generate prosthetic socket during the swing phase of gait. However, increased
graded volitional movement commands during the volitional control levels of involuntary cocontraction have also been observed in upper-­
experiments. extremity amputees during volitional control tasks (50–52). We posit
Cocontraction during gait in patients having unilateral lower-­ that these complications may be attributed, at least in part, to limita-
extremity amputations has been documented in several independent tions of the traditional clinical amputation procedure and rehabilitation
studies (48, 49). In these studies, it is hypothesized that cocontrac- protocol. Because the muscles in the residual limb of all subjects in

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 9 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

group T are anchored at fixed lengths, the dynamic muscle relation- electrodes that are fixed in place on the muscle to deliver repeatable
ships that exist within a biological limb with intact anatomy are bro- stimulation (16, 17). In addition, it is worth noting that stimulation
ken. These relationships are fundamental to fine motor control and of residual muscles may also improve performance during torque-­
functional joint stability (30, 53–55) and play a significant role in re- control tasks in persons with traditional amputation; however, the
ciprocal reflex inhibition (56–58). In their absence, traditional inhib- mechanism behind any potential improvement resulting from such
itory reflex arcs may be disrupted, which would increase unintended an approach, which would involve stimulating muscles that are fixed
antagonistic coactivation, and have a profound impact on a patient’s isometrically, would fundamentally differ from the agonist-antagonist
ability to generate independent and separable muscle commands. relationships that drive perception within the natural limb.
The AMI has the potential to resolve this limitation by restoring the Another key difference between the experiences of subject A and
agonist-antagonist muscle relationships that are essential to appro- group T is rooted in their subjective descriptions of their relation-
priate reflexive muscle activation and by providing feedback of move- ship with the prosthesis. Subject A felt an immediate and lasting
ment commands in the form of proprioceptive sensation. Supported connection with the device, whereas subjects in group T described
by ultrasound data and patient testimonials, it is our hypothesis that a distinct disconnect. On the basis of their accounts, we believe that
dynamic agonist-antagonist stretch relationships in the residuum of the difference in embodiment is attributable to two primary factors,
subject A provide a proprioceptive affirmation of muscle activity namely, (i) robustness and intuitiveness of efferent control and (ii)
within his residuum; each time he seeks to move his phantom limb, reliability of afferent feedback. It is our position that each correctly
subject A receives confirmation of correct muscle activation as stretch executed volitional or reflexive behavior, reinforced by natural pro-
receptors within the AMI muscles send signals to his brain. prioceptive sensation, has the potential to deepen the relationship
Swing-phase adjustments to joint position and impedance play a between human and machine. In this way, a bionic system that in-
critical role in the adaptation of gait to varying terrains (42, 43), and tegrates more completely with a patient’s sense of self has the poten-

Downloaded from https://www.science.org on August 12, 2024

their absence has a significant impact on gait symmetry (59, 60). Rep- tial to improve usage and satisfaction (44).
licating these adaptations has long been a goal of lower-extremity In this case study, the AMI described was implemented in an ideal
prosthetic research, with the majority of efforts focused on intrinsic surgical setting. The elective nature of the amputation made it possible
or EMG-based terrain prediction and recognition methodologies to carefully plan our surgical approach. The muscular anatomy and
(61–63). Unfortunately, both the accuracy and versatility of these limited degrees of freedom of the ankle and subtalar joints simplified
state-of-the-art approaches pale in comparison to the human central the procedure relative to what would be necessary at the above-knee
nervous system, with its unparalleled ability to synthesize data streams level or in the upper extremity. Even the patient’s indication for ampu-
from a vast array of biological sensors into a cogent motor control tation was advantageous; nonresolving bone injury represents an op-
framework. While traversing various terrains, subjects in group T did timal availability of healthy distal soft tissue. However, it is important
not consistently demonstrate reflexive swing-phase modulation of pros- to note that the benefits of the AMI are not restricted to this limited
thetic joint angle that would result in natural gait adaptations. These patient population. Research is already underway to explore con-
findings are in line with several studies examining EMG profiles in struction of AMIs at other amputation levels, as well as in the upper
persons having transtibial amputation. In one study, it was shown extremity. A recent study demonstrated the potential to leverage re-
that there is little intersubject consistency in muscle activation pro- generative capabilities of nerve and muscle tissue in the construction
files during level ground walking and that muscle recruitment pat- of AMIs in settings where distal tissues are no longer available, such
terns do not match those in persons with two intact biological limbs as traumatic amputations or revisions to existing amputations (40).
(49). In another study, it was concluded that persons having trans- It is worth noting that, even with these advancements, the imple-
tibial amputation are more prone to cocontraction within the resid- mentation of the AMI may not be appropriate in patients requiring
ual ankle musculature than healthy controls (48), which is consistent amputation due to advanced peripheral vascular disease. Patients in
with our observations in several subjects from group T. In contrast, this population typically exhibit neuropathy and microvascular com-
subject A reflexively modulated swing-phase joint angle in a manner promise, which may negate the benefits of the AMI and inhibit proper
appropriate to each terrain without training. These findings under- wound healing. Nevertheless, even if this population were excluded
score the potential of the AMI to reinstate the central nervous system entirely, a majority of the remaining estimated 46% of patients indi-
as the primary mediator of gait adaptation by providing the afferent cated for amputation (64) would be eligible for an AMI procedure. It
proprioceptive sensations that are crucial to this function. is also noteworthy that the study presented herein does not separate
Here, we characterize a methodology to communicate sensations the impact of the AMI procedure from the visualization exercises that
of joint torque from a bionic limb directly to the nervous system in a were added to the rehabilitation protocol with the intent of preserv-
patient having an AMI. This feedback is perceived by subject A as ing muscle sliding (for details, see the “Subject selection, surgery, and
natural torque about his phantom ankle and improves his perfor- rehabilitation” section). This is a feature inherent to a case study design
mance in a task requiring torque modulation. Reliable closed-loop that relies on a historic control group. Because this is a first-in-human
control of joint torque has the potential to provide an array of func- case study, the results presented herein serve to highlight the poten-
tionality to prosthetic users that was heretofore impossible. However, tial of the AMI to improve volitional and reflexive neural control of a
for these visions to become a reality, it will be necessary to improve prosthetic device; a larger trial in a greater number of patients is nec-
viability of the stimulation delivery mechanism. The fine-wire elec- essary to definitively understand the degree of improvement that can
trodes used in this study are not a feasible long-term solution, be- be attributed specifically to the AMI procedure.
cause they are placed acutely for each experimental session and are Proprioceptive insensibility has long been a stumbling block for
not sufficiently anchored within the muscle to withstand the large integration of bionic devices with human physical identity. The AMI
shear forces associated with socket use. These issues can be resolved is fundamentally distinct from other approaches in that its implemen-
with a shift to permanently implanted intramuscular or epimysial tation begins with a reengineering of the musculoskeletal anatomy

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 10 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

within the residuum. This approach is built upon an expanded under- 47 years), time since amputation (range, 1 to 24 years), and body
standing of what comprises a “neural interface” to incorporate not only mass index (range, 24 to 33 kg/m2). For more details about these
synthetic components but also biological tissues (26, 65). Because of subjects, see the “Subject selection (group T)” section in the Supple-
the inherent capacity of muscle tissue to amplify efferent neural sig- mentary Materials.
nals and mechanoreceptors within muscle and tendon to communi-
cate afferent proprioceptive information to the nervous system, these Surface electrode placement and EMG processing
native biological transducers are ideally suited to act as the bidirec- EMG was recorded via bipolar surface electrodes, placed acutely
tional interface between the nerve and the prosthesis. The AMI was over each of the four target muscles: lateral gastrocnemius for plan-
designed with the intent of optimizing this biological interface. The tar flexion, tibialis anterior for dorsiflexion, tibialis anterior for in-
results presented herein demonstrate the potential of such a bionic version, and peroneus longus for eversion. An identical electrode
system to improve functional outcomes and embodiment when com- placement protocol was followed for all experimental subjects. For
pared with a traditional approach to amputation. further details, see the Supplementary Materials.

Efferent control architecture

MATERIALS AND METHODS The efferent control paradigm explored in this study was designed to
Study design allow direct control of prosthetic joint position and impedance. In
The primary hypothesis investigated in this case study is that the AMI this control approach, EMG signal amplitudes recorded from the ag-
procedure and rehabilitation protocol (i) enables independent control onist and antagonist AMI muscles were interpreted as desired torques
of prosthetic joint angle and impedance and (ii) reflects proprioceptive produced in opposite directions about a virtual dynamic joint, which
afferent sensation pertaining to each joint of a two-degree-of-freedom was constructed with physiologically relevant values for virtual paral-

Downloaded from https://www.science.org on August 12, 2024

ankle-foot prosthesis onto the central nervous system. The experiments lel spring stiffness, virtual damping, and virtual inertia. The differ-
presented were designed to demonstrate the AMI’s potential to im- ence of these estimated torques was then applied to the virtual joint,
prove volitional, free-space control, restore swing-phase reflexes while causing it to move. The position of the virtual joint controlled the
traversing various terrains, and perform closed-loop torque control desired position of the associated prosthetic joint (fig. S1). Prosthetic
tasks. This was a first-in-human, nonrandomized case study. The ex- joint stiffness was directly modulated by the mean activation of the
perimental subject (subject A) served as his own control where possi- agonist and antagonist muscles, as modeled in (53). This control ar-
ble. Where control subjects having a traditional amputation were chitecture enables independent modulation of joint position and im-
needed, four subjects were selected (group T), representing a pedance. As with all EMG-based proportional control systems, there
wide array of patient demographics. None of the patients reported is a trade-off between joint stability and latency; typically, the partic-
or showed signs of any complicating nerve or muscle damage within ulars of this trade-off are buried in filter design. One benefit to the
the residual limb. virtual-joint architecture is that filter parameters take on intuitive
physical meaning and can be set to near-physiologic values. For a de-
Subject selection, surgery, and rehabilitation scription of controller tuning, see the Supplementary Materials.
The AMI operation and clinical follow-up were performed with informed Although stimulation was active for prosthetic joint torque feed-
consent at Brigham and Women’s Hospital, under the approval of the back, the stimulated muscle was assumed to be at zero activation,
Partner’s Health System Institutional Review Board. All other exper- and input from that muscle to the controller was blocked (fig. S1).
iments were carried out with informed consent at the Massachusetts Although this design eliminates the ability to actively move the joint
Institute of Technology (MIT), under the approval of the Committee in the same direction as an applied load, the scenarios in which this
on the Use of Humans as Experimental Subjects. Subject A was se- action would be desirable are likely to be extremely limited.
lected for participation based primarily on his need for elective uni-
lateral transtibial amputation, indicated due to a traumatic Hawkins Fine-wire electrode placement and stimulation
type 4 talus fracture and persistent nerve pain. He was 53 years old at Fine-wire electrodes were placed acutely (M.J.C.) at the start of each
the time of his primary amputation in July 2016. During this primary trial day, according to the technique presented in (66). For all trials,
amputation, two AMIs were constructed within his residual limb by the tibialis anterior muscle was stimulated with a 50-Hz, current-­
one of the study authors (M.J.C.). The amputation osteotomy was controlled, charge-balanced, asymmetric, biphasic pulse train (NL800,
performed at 12 cm distal to the patellar ligament, resulting in a Digitimer). The pulse width of the cathodic phase was 200 s and
residuum of standard length. Acute rehabilitation began at 6 weeks that of the anodic phase was 400 s. For the closed-loop torque con-
postoperation. In addition to standard rehabilitation protocols, the trol experiments, the cathodic current amplitude was modulated in
patient regularly performed exercises focused on preserving motion linear proportionality to prosthetic torque (measured or simulated)
within the AMI constructs. During these exercises, the patient was and ranged from 0 to 9 mA. For a full description of the electrode
asked to visualize his phantom limb and focus on moving his phan- placement protocol, stimulation parameters, and evidence of proper
tom foot through the four primary ankle and subtalar joint motions electrode placement, see the Supplemental Materials, fig. S2, and
(plantar flexion and dorsiflexion, inversion and eversion). No direct movie S5.
feedback of muscle activity was provided to the patient during reha-
bilitation exercises. Experimental sessions with the bionic pros- Closed-loop torque control experimental setup
thesis began in late April 2017 (9 to 10 months postoperation) and During the closed-loop torque control experiments, the prosthesis
continued through November 2017 (16 to 17 months postoperation). was mounted to an assembly that held it in contact with the foot
All subjects in group T were men with unilateral transtibial am- pedal, remote from the subject, to eliminate the possibility of con-
putation, selected to incorporate a range of patient age (range, 37 to founding force feedback through the prosthetic socket. The subject

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 11 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

was acoustically isolated with noise-canceling headphones worn 7. I. A. Boyd, T. D. M. Roberts, Proprioceptive discharges from stretch-receptors in the
knee-joint of the cat. J. Physiol. 122, 38–58 (1953).
over ear plugs and visually isolated with a sleep mask. He performed
8. S. Skoglund, Anatomical and physiological studies of knee joint innervation in the cat.
eight total trial groups under two different conditions; four of these Acta Physiol. Scand. Suppl. 36, 1–101 (1956).
trial groups were performed with torque feedback through stimula- 9. D. Burke, S. C. Gandevia, G. Macefield, Responses to passive movement of receptors in
tion of the AMI, and four were performed without such stimula- joint, skin and muscle of the human hand. J. Physiol. 402, 347–361 (1988).
tion. Under the stimulation on condition, the tibialis anterior was 10. L. Jami, Golgi tendon organs in mammalian skeletal muscle: Functional properties and
central actions. Physiol. Rev. 72, 623–666 (1992).
stimulated with a current amplitude proportional to the plantar 11. G. Eklund, Position sense and state of contraction; the effects of vibration. J. Neurol.
flexion torque measured on the prosthesis. Under the stimulation Neurosurg. Psychiatry 35, 606–611 (1972).
off condition, no such feedback was provided. Each trial group con- 12. G. M. Goodwin, D. I. McCloskey, P. B. C. Matthews, The contribution of muscle afferents to
sisted of 20 commands (5 at each effort level), presented in a ran- kinaesthesia shown by vibration induced illusions of movement and by the effects of
paralysing joint afferents. Brain 95, 705–748 (1972).
dom order. Within a single trial group, all trials were carried out
13. E. Ribot-Ciscar, J.-P. Roll, Ago-antagonist muscle spindle inputs contribute together to
under the same condition, but the order of trial group conditions joint movement coding in man. Brain Res. 791, 167–176 (1998).
was selected at random. This format was chosen to account for bias 14. B. J. Brown, M. L. Iorio, M. R. Klement, M. R. Conti Mica, A. El-Amraoui, P. D. O’Halloran,
in one condition having been consistently evaluated before the C. E. Attinger, Outcomes after 294 transtibial amputations with the posterior
other, while preventing the possible scenario in which the subject myocutaneous flap. Int. J. Low. Extrem. Wounds 13, 33–40 (2014).
15. P. Cao, P. De Rango, in Rutherford’s Vascular Surgery (Saunders Elsevier, 2010), vol. 2,
applies excess torque in anticipation of a stimulus that never comes. pp. 1469–1486.
The experiment was repeated without stimulation in subject 16. M. Ortiz-Catalan, R. Brånemark, B. Håkansson, J. Delbeke, On the viability of implantable
A’s unaffected limb to set a performance baseline. In these trials, electrodes for the natural control of artificial limbs: Review and discussion. Biomed. Eng.
subject A applied torque to the pedal with his right biological foot Online 11, 33 (2012).
17. X. Navarro, T. B. Krueger, N. Lago, S. Micera, T. Stieglitz, P. Dario, A critical review of
directly contacting the pedal. Pedal torque was streamed from the
interfaces with the peripheral nervous system for the control of neuroprostheses and

Downloaded from https://www.science.org on August 12, 2024

pedal’s load cell to a data acquisition system (USB-6009, National hybrid bionic systems. J. Peripher. Nerv. Syst. 10, 229–258 (2005).
Instruments). 18. S. K. Au, J. Weber, H. Herr, Powered ankle–foot prosthesis improves walking metabolic
economy. IEEE Trans. Robot. 25, 51–66 (2009).
Statistical analysis 19. D. W. Tan, M. A. Schiefer, M. W. Keith, J. R. Anderson, J. Tyler, D. J. Tyler, A neural
interface provides long-term stable natural touch perception. Sci. Transl. Med. 6,
Time-synchronized joint angle, EMG, and joint stiffness data were 257ra138 (2014).
collected from the prosthesis in real time for all trials. Statistical 20. M. Ortiz-Catalan, B. Hakånsson, R. Brånemark, An osseointegrated human-machine
comparisons were made within a given subject using the Tukey-­ gateway for long-term sensory feedback and motor control of artificial limbs. Sci. Transl.
Kramer multiple comparisons tests at a significance level of  = Med. 6, 257re6 (2014).
21. S. Raspopovic, M. Capogrosso, F. M. Petrini, M. Bonizzato, J. Rigosa, G. Di Pino,
0.025. In cases where a single comparison was appropriate, a t test
J. Carpaneto, M. Controzzi, T. Boretius, E. Fernandez, G. Granata, C. M. Oddo, L. Citi,
was used at a significance level of  = 0.025. All statistical analysis A. L. Ciancio, C. Cipriani, M. C. Carrozza, W. Jensen, E. Guglielmelli, T. Stieglitz,
was performed in MATLAB R2017a (The MathWorks). All EMG P. M. Rossini, S. Micera, Restoring natural sensory feedback in real-time bidirectional
signals are normalized to calibrated maxima for each muscle. All hand prostheses. Sci. Transl. Med. 6, 222ra19 (2014).
shaded traces indicate mean ± 1 SD. 22. M. A. Schiefer, D. Tan, S. M. Sidek, D. J. Tyler, Sensory feedback by peripheral nerve
stimulation improves task performance in individuals with upper limb loss using a
myoelectric prosthesis. J. Neural Eng. 13, 16001 (2016).
SUPPLEMENTARY MATERIALS 23. K. Horch, S. Meek, T. G. Taylor, D. T. Hutchinson, Object discrimination with an artificial
www.sciencetranslationalmedicine.org/cgi/content/full/10/443/eaap8373/DC1 hand using electrical stimulation of peripheral tactile and proprioceptive pathways with
Materials and Methods intrafascicular electrodes. IEEE Trans. Neural Syst. Rehabil. Eng. 19, 483–489 (2011).
Fig. S1. Control diagrams showing how EMG from the AMI muscles drives movement of the 24. G. S. Dhillon, K. W. Horch, Direct neural sensory feedback and control of a prosthetic arm.
prosthetic joint. IEEE Trans. Neural Syst. Rehabil. Eng. 13, 468–472 (2005).
Fig. S2. Raw EMG recorded from fine-wire electrodes during volitional movement of the 25. E. L. Graczyk, M. A. Schiefer, H. P. Saal, B. P. Delhaye, S. J. Bensmaia, D. J. Tyler, The neural
phantom limb. basis of perceived intensity in natural and artificial touch. Sci. Transl. Med. 8, 362ra142
Movie S1. Ultrasound video of coupled AMI motion. (2016).
Movie S2. Volitional control. 26. T. A. Kung, N. B. Langhals, D. C. Martin, P. J. M. Johnson, P. S. Cederna, M. G. Urbanchek,
Movie S3. Reflexive control. Regenerative peripheral nerve interface viability and signal transduction with an
Movie S4. Candid videos showing prosthesis embodiment. implanted electrode. Plast. Reconstr. Surg. 133, 1380–1394 (2014).
Movie S5. Visual confirmation of stimulated muscle contraction. 27. Z. T. Irwin, K. E. Schroeder, P. P. Vu, D. M. Tat, A. J. Bullard, S. L. Woo, I. C. Sando,
Table S1. Individual subject data for volitional control tasks. M. G. Urbanchek, P. S. Cederna, C. A. Chestek, Chronic recording of hand prosthesis
control signals via a regenerative peripheral nerve interface in a rhesus macaque.
J. Neural Eng. 13, 46007 (2016).
REFERENCES AND NOTES 28. T. A. Kuiken, G. Li, B. A. Lock, R. D. Lipschutz, L. A. Miller, K. A. Stubblefield, K. B. Englehart,
1. U. Proske, S. C. Gandevia, The proprioceptive senses: Their roles in signaling body Targeted muscle reinnervation for real-time myoelectric control of multifunction artificial
shape, body position and movement, and muscle force. Physiol. Rev. 92, 1651–1697 arms. JAMA 301, 619–628 (2009).
(2012). 29. L. J. Hargrove, A. M. Simon, A. J. Young, R. D. Lipschutz, S. B. Finucane, D. G. Smith,
2. B. L. Riemann, S. M. Lephart, The sensorimotor system, Part II: The role of proprioception T. A. Kuiken, Robotic leg control with EMG decoding in an amputee with nerve transfers.
in motor control and functional joint stability. J. Athl. Train. 37, 80–84 (2002). N. Engl. J. Med. 369, 1237–1242 (2013).
3. E. R. Kandel, J. H. Schwartz, T. M. Jessell, in Principles of Neural Science (McGraw-Hill, 2013), 30. E. J. Rouse, L. J. Hargrove, E. J. Perreault, T. A. Kuiken, Estimation of human ankle
vol. 4, 1414 pp. impedance during the stance phase of walking. IEEE Trans. Neural Syst. Rehabil. Eng. 22,
4. J. J. Ochoa, E. E. Torebjörk, Sensations evoked by intraneural microstimulation of single 870–878 (2014).
mechanoreceptor units innervating the human hand. J. Physiol. 342, 633–654 (1983). 31. H. M. Herr, A. M. Grabowski, Bionic ankle–foot prosthesis normalizes walking gait for
5. G. Macefield, S. C. Gandevia, D. Burke, Perceptual responses to microstimulation of single persons with leg amputation. Proc. Biol. Sci. 279, 457–464 (2012).
afferents innervating joints, muscles and skin of the human hand. J. Physiol. 429, 32. R. D. Bellman, M. A. Holgate, T. G. Sugar, SPARKy 3: Design of an active robotic ankle
113–129 (1990). prosthesis with two actuated degrees of freedom using regenerative kinetics, in
6. D. F. Collins, K. M. Refshauge, G. Todd, S. C. Gandevia, Cutaneous receptors contribute to Proceedings of the 2nd Biennial IEEE/RAS-EMBS International Conference on Biomedical
kinesthesia at the index finger, elbow, and knee. J. Neurophysiol. 94, 1699–1706 (2005). Robotics and Biomechatronics, Biorob 2008 (IEEE, 2008), pp. 511–516.

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 12 of 13
SCIENCE TRANSLATIONAL MEDICINE | RESEARCH ARTICLE

33. F. Sup, A. Bohara, M. Goldfarb, Design and control of a powered transfemoral prosthesis. 60. S. Sienko Thomas, C. E. Buckon, S. Jakobson-Huston, M. D. Sussman, M. D. Aiona, Stair
Int. J. Rob. Res. 27, 263–273 (2008). locomotion in children with spastic hemiplegia: The impact of three different ankle foot
34. S. Huang, J. P. Wensman, D. P. Ferris, An experimental powered lower limb prosthesis orthosis (AFOs) configurations. Gait Posture 16, 180–187 (2002).
using proportional myoelectric control. J. Med. Device. 8, 24501 (2014). 61. H. Huang, F. Zhang, L. J. Hargrove, Z. Dou, D. R. Rogers, K. B. Englehart, Continuous
35. H. Huang, T. A. Kuiken, R. D. Lipschutz, A strategy for identifying locomotion modes using locomotion-mode identification for prosthetic legs based on neuromuscular–mechanical
surface electromyography. IEEE Trans. Biomed. Eng. 56, 65–73 (2009). fusion. IEEE Trans. Biomed. Eng. 58, 2867–2875 (2011).
36. L. J. Hargrove, A. M. Simon, R. D. Lipschutz, S. B. Finucane, T. A. Kuiken, Real-time 62. A. J. Young, T. A. Kuiken, L. J. Hargrove, Analysis of using EMG and mechanical sensors to
myoelectric control of knee and ankle motions for transfemoral amputees. JAMA 305, enhance intent recognition in powered lower limb prostheses. J. Neural Eng. 11, 56021
1542–1544 (2011). (2014).
37. H. M. Herr, R. R. Riso, K. W. Song Jr., R. J. Casler, M. J. Carty, Peripheral Neural Interface Via 63. R. Stolyarov, G. Burnett, H. Herr, Translational motion tracking of leg joints for enhanced
Nerve Regeneration to Distal Tissues (Massachusetts Institute of Technology, 2016). prediction of walking tasks. IEEE Trans. Biomed. Eng. 65, 736–769 (2017).
38. H. M. Herr, T. R. Clites, B. Maimon, A. Zorzos, M. J. Carty, J.-F. Duval, Method and System for 64. K. Ziegler-Graham, E. J. MacKenzie, P. L. Ephraim, T. G. Travison, R. Brookmeyer,
Providing Proprioceptive Feedback and Functionality Mitigating Limb Pathology Estimating the prevalence of limb loss in the United States: 2005 to 2050. Arch. Phys. Med.
(Massachusetts Institute of Technology, 2016). Rehabil. 89, 422–429 (2008).
39. T. R. Clites, M. J. Carty, S. Srinivasan, A. N. Zorzos, H. M. Herr, A murine model of a novel 65. T. A. Kuiken, L. A. Miller, R. D. Lipschutz, B. A. Lock, K. Stubblefield, P. D. Marasco, P. Zhou,
surgical architecture for proprioceptive muscle feedback and its potential application to G. A. Dumanian, Targeted reinnervation for enhanced prosthetic arm function in a
control of advanced limb prostheses. J. Neural Eng. 14, 036002 (2017). woman with a proximal amputation: A case study. Lancet 369, 371–380 (2007).
40. S. S. Srinivasan, M. J. Carty, P. W. Calvaresi, T. R. Clites, B. E. Maimon, C. R. Taylor, 66. T. A. Park, G. F. Harris, “Guided” intramuscular fine wire electrode placement: A new
A. N. Zorzos, H. Herr, On prosthetic control : A regenerative agonist-antagonist technique. Am. J. Phys. Med. Rehabil. 75, 232–234 (1996).
myoneural interface. Sci. Robot. 2, eaan2971 (2017).
41. T. R. Clites, M. J. Carty, H. M. Herr, Plastic Surgery Research Council (PSRC, 2017), 119 pp. Acknowledgments: We thank A. Zorzos for his contribution to the design of the surgical
42. B. J. McFadyen, D. A. Winter, An integrated biomechanical analysis of normal stair ascent procedure, as well as J. Ernst for her work advancing the FES system. We acknowledge the
and descent. J. Biomech. 21, 733–744 (1988). contributions of C. Ricciardi and the MIT Clinical Research Center, both in securing institutional
43. R. Riener, M. Rabuffetti, C. Frigo, Stair ascent and descent at different inclinations. review board approval and in providing access to clinical spaces. All original figure artwork
Gait Posture 15, 32–44 (2002). was created by S. Ku. We recognize all those who have played a role in this work and especially

Downloaded from https://www.science.org on August 12, 2024

44. U. Wijk, I. Carlsson, Forearm amputees’ views of prosthesis use and sensory feedback. thank the subjects and their families for their sacrifice and dedication to improving the lives of
J. Hand Ther. 269–277 (2015). persons living with amputation. Funding: This work was supported by the MIT Media Lab
45. H. H. Ehrsson, B. Rosén, A. Stockselius, C. Ragnö, P. Köhler, G. Lundborg, Upper limb Consortia, the Stepping Strong Innovator Fund, and a generous gift from Google, Inc.
amputees can be induced to experience a rubber hand as their own. Brain 131, Prosthetic design and fabrication were funded in part by the U.S. Army Medical Research and
3443–3452 (2008). Materiel Command (W81XWH-14-C-0111). Author contributions: T.R.C. contributed to
46. A. Guterstam, V. I. Petkova, H. H. Ehrsson, The illusion of owning a third arm. PLOS ONE 6, conceptualization and preclinical validation of the AMI operation; designed and implemented
e17208 (2011). the efferent control and afferent stimulation paradigms; designed the experiments; collected,
47. M. Botvinick, J. Cohen, Rubber hands ‘feel’ touch that eyes see. Nature 391, 756 (1998). processed, and analyzed the data; and wrote the manuscript. M.J.C. contributed to
48. M. Seyedali, J. M. Czerniecki, D. C. Morgenroth, M. E. Hahn, Co-contraction patterns of conceptualization and preclinical validation of the AMI architecture and operation, performed
trans-tibial amputee ankle and knee musculature during gait. J. Neuroeng. Rehabil. 9, 29 both the traditional and AMI amputation procedures, and contributed to experimental design
(2012). and data collection. J.B.U. assisted with experimental design and data collection and
49. S. Huang, D. P. Ferris, Muscle activation patterns during walking from transtibial amputees developed experimental tools. M.E.C. assisted with implementation of the prosthetic hardware
recorded within the residual limb-prosthetic interface. J. Neuroeng. Rehabil. 9, 55 (2012). and embedded systems. L.M.M. conceived, designed, and developed the two-degree-of-
50. G. A. García, R. Okuno, in Electrodiagnosis in New Frontiers of Clinical Research freedom foot-ankle prosthesis. J.-F.D. designed and programmed the embedded system. S.S.S.
(IntechOpen, 2013), pp. 269–285. contributed to conceptualization and preclinical validation of the AMI operation, analyzed
51. E. Scheme, K. Englehart, Electromyogram pattern recognition for control of powered experimental ultrasound data, and assisted in writing the manuscript. H.M.H. conceived the
upper-limb prostheses: State of the art and challenges for clinical use. J. Rehabil. Res. Dev. AMI architecture and contributed to conceptualization and preclinical validation of the AMI
48, 643–660 (2011). operation, controller design, experimental design, and writing of the manuscript.
52. B. A. Lock, A. E. Schultz, T. A. Kuiken, Clinically practical applications of pattern Competing interests: H.M.H. and M.J.C. hold a patent on the AMI concept entitled “Peripheral
recognition for myoelectric prostheses, in Proceedings of the 2008 MyoElectric Controls/ Neural Interface via Nerve Regeneration to Distal Tissues” (U.S. Patent US20160346099).
Powered Prosthetics Symposium (MEC, 2008), pp. 0–5. H.M.H., T.R.C., and M.J.C. have filed a patent on the AMI operation and associated control
53. N. Hogan, Adaptive control of mechanical impedance by coactivation of antagonist strategies, entitled “Method and System for Providing Proprioceptive Feedback and
muscles. IEEE Trans. Automat. Contr. 29, 681–690 (1984). Functionality Mitigating Limb Pathology” (U.S. Pending Patent 62/276422). All other authors
54. H. Lee, N. Hogan, in ASME 2010 Dynamic Systems and Control Conference (ASME, 2010), declare that they have no competing interests. Data and materials availability: Please
pp. 6–8. address requests for more information to H.M.H. (hherr@media.mit.edu).
55. W. T. Edwards, Effect of joint stiffness on standing stability. Gait Posture 25, 432–439 (2007).
56. A. M. Smith, The coactivation of antagonist muscles. Can. J. Physiol. Pharmacol. 59, Submitted 6 September 2017
733–747 (1981). Resubmitted 9 December 2017
57. C. J. De Luca, B. Mambrito, Voluntary control of motor units in human antagonist Accepted 3 May 2018
muscles: Coactivation and reciprocal activation. J. Neurophysiol. 58, 525–542 (1987). Published 30 May 2018
58. M. Dimitriou, Human muscle spindle sensitivity reflects the balance of activity between 10.1126/scitranslmed.aap8373
antagonistic muscles. J. Neurosci. 34, 13644–13655 (2014).
59. M. T. Nahorniak, G. E. Gorton III, M. E. Gannotti, P. D. Masso, Kinematic compensations as Citation: T. R. Clites, M. J. Carty, J. B. Ullauri, M. E. Carney, L. M. Mooney, J.-F. Duval, S. S. Srinivasan,
children reciprocally ascend and descend stairs with unilateral and bilateral solid AFOs. H. M. Herr, Proprioception from a neurally controlled lower-extremity prosthesis. Sci. Transl.
Gait Posture 9, 199–206 (1999). Med. 10, eaap8373 (2018).

Clites et al., Sci. Transl. Med. 10, eaap8373 (2018) 30 May 2018 13 of 13