Kuiken, TA
A paper for submission to Journal of Technology and Disability:
Consideration of nerve-muscle grafts to improve the
control of artificial arms
Todd Kuiken, M.D., Ph.D.
Address where work was performed:
Dept. of PM&R, Northwestern University Medical School, Chicago, Illinois
Rehabilitation Institute of Chicago
345 E. Superior St.
Chicago, IL 60611
Corresponding Author:
Todd Kuiken, MD, PhD
Rehabilitation Institute of Chicago, Rm. 1124
345 E. Superior St.
Chicago, IL 60611
Tel: 312-238-8072
Fax: 312-238-1166
Email: tkuiken@rehabchicago.org
1
Kuiken, TA
Abstract
Improving the function of artificial arms remains a considerable challenge, especially for highlevel amputations where the disability is greatest. It may be possible to denervate expendable
regions of muscle in or near an amputated limb and graft the residual peripheral nerves to this
muscle. The surface EMG signals from the nerve-muscle grafts would then be used as additional
control signals for an externally powered prosthesis.
Such a system would allow the
simultaneous control of multiple degrees-of-freedom in a prosthesis and could greatly improve
the function of myoelectric prostheses. The potential advantages, requirements for successful
implementation and synergies with other research are discussed.
Key words: prosthesis, control, myoelectric
2
Kuiken, TA
1. Introduction
Improving the function of artificial arms remains a considerable challenge, especially for
high-level amputations where the disability is greatest. Externally powered hooks, hands, wrists,
and elbows are available, but existing control methods are inadequate. Currently, most powered
artificial limbs are controlled using myoelectric signals from an antagonist pair of muscles in the
amputated limb [1]. This allows only a single motion to be operated at a time, as operation of
the terminal device, wrist and elbow must be preformed sequentially. For example, in the
transhumeral amputee, the biceps and triceps may be used to operate elbow extension or flexion,
then rotation of the wrist, and finally opening or closing of the powered hand. This control
method is frustratingly slow.
Normal human arm function has coordinated simultaneous
movement of the hand, wrist and elbow. Furthermore, conventional high-level myoelectric
control methods do not have a natural feel as biceps and triceps function are not directly related
to wrist rotation or opening/closing of the human hand. A highly articulated limb is of little use
if its movements are not well coordinated or if it is difficult to operate.
2.0 Nerve-Muscle Graft Paradigm
Although the limb is lost with an amputation, the control signals to the limb remain in the
residual peripheral nerves. The potential exists to tap into these control signals using nervemuscle grafts and greatly improve the function of upper limb prostheses. It may be possible to
denervate expendable regions of muscle in or near an amputated limb and graft the residual
peripheral nerve endings to these muscles [2]. The nerves would reinnervate these muscles.
Then, the surface electromyograms (EMGs) from the nerve-muscle grafts could be used as
additional myoelectric control signals for an externally powered prosthesis. In a long
3
Kuiken, TA
transhumeral amputee, for example, the medial head of the biceps and two heads of the triceps
could be denervated. The median, ulnar and distal radial nerves would be grafted on to these
heads, respectively, and allowed to reinnervate these regions of muscle (see Fig. 1). Now, if the
amputee thought ‘close hand’ the neural control signal would travel down the median nerve and
cause the medial head of the biceps to contract. The surface EMG from the medial head of the
biceps would then be used as a control signal to close the terminal device of the prosthesis. If the
amputee thought ‘bend elbow’, the neural control signal would still travel down the
musculocutaneous nerve and cause just the lateral head of the biceps to contract. The surface
EMG from the lateral head of the biceps would then be used as a control signal to flex the
prosthetic elbow. In the same manner the distal radial nerve-muscle graft would control opening
of the hand and the intact head of the triceps (still innervated by its branch of the radial nerve)
would control elbow extension. Thus, antagonist pairs of muscles would provide simultaneous
control of a terminal device and elbow. The ulnar nerve-muscle graft could be used to control a
third degree-of-freedom such as a wrist rotation, or wrist flexion-extension.
For the shoulder disarticulation patient, a broad surface muscle such as the pectoralis
major could be used (see Fig. 2). The muscle would be denervated and separated into four
regions. Each residual nerve (the median, ulnar, musculocutaneous and radial nerves) would be
grafted on to a separate region of muscle. Each region of muscle would then generate an
independent surface EMG providing four additional myoelectric control signals. Free muscle
flaps could also be used to create nerve-muscle grafts where desired. For example, the latissimus
dorsi muscles could be moved to the lateral chest wall in the former axilla region and
reinnervated with residual nerves to produce new myoelectric control sites.
4
Kuiken, TA
With the nerve-muscle grafting technique the amputee's residual nerves would be grafted
onto ‘foreign’ regions of muscle and would cross-reinnervate these muscles. If the nerve
dominates motor control in the nerve-muscle graft technique, then using the EMG from the
nerve-muscle grafts as control signals for powered prostheses should have a very natural feel; the
nerves would be controlling movements in the prosthesis that directly relate to their normal
anatomic function. A number of animal studies have clearly shown that motor control is
dominated by the nerve’s function in cross-reinnervated muscle and not by the function of the
muscle [3,4]. Human studies have also shown that motor control is strongly dominated by the
function of the nerve in cross-reinnervated muscle. However, humans are capable of learning to
use nerve-muscle grafts in different ways with time and effort, just as they are capable of
learning how to use natural muscles in different ways [5,6]. Using nerve-muscle grafts for
amputees would take advantage of the nerve’s motor programming so that the nerves are
simultaneously controlling physiologically appropriate functions in the prosthesis. The control
of the artificial limb would be quicker and have a more natural feel than the use of conventional
transhumeral or shoulder disarticulation myoelectric prostheses.
This would reduce the
conscious effort required by the amputee, making the prosthesis easier to use and more
functional.
Shoulder motion would still be available to power and/or control additional functions in
the prosthesis.
For example, with the shoulder disarticulation amputee shoulder
elevation/depression and protraction/retraction could be used to control abduction/adduction and
flexion/extension in an externally powered shoulder, while the nerve-muscle grafts control the
powered elbow, wrist and terminal device. This would allow simultaneous control of two
additional degrees-of-freedom. Furthermore, using shoulder movement to control externally
5
Kuiken, TA
powered shoulder function would, once again, be more natural for the amputee allowing easier
operation of the prosthesis.
Existing myoelectric technologies could be applied with the nerve-muscle graft
technique; a new prosthesis would not need to be developed. Powered elbows, wrists and
terminal devices are commercially available. The circuitry is available allowing up to seven
analog inputs (e.g. myoelectric signals) and four on/off input signals that provide the control of
up to five motors [7]. However, these devices are limited by inadequate human control systems.
The nerve-muscle grafting technique would enable better control of currently available devices.
Another advantage of the nerve-muscle graft technique is that the additional control
signals are made available without percutaneous wires or implanted hardware that is required
with other proposed systems. With neuroelectric control, electrodes are directly connected to the
residual nerves of the amputee and the electroneurogram (ENG) of the nerve is used to control
the artificial limb [2,8-10]. This is an exciting research area that also offers the hope of
simultaneous control of a multifunction prosthesis with a natural feel. However, neuroelectric
controls requires either chronic percutaneous wires (which tend to become infected) or complex
transmitter-receiver systems.
Durability of the implanted hardware is another issue. These
systems would need to function for many decades and would require surgery to repair. The
nerve-muscle graft method uses muscle as a biological amplifier of the ENG signal to
circumvent the problems of neuroelectric control. EMG signals are hundreds of times larger
than the ENG signals, and easily recorded from the surface of the body; the additional control
signals would be accessible without the use of implanted nerve cuffs, implanted transmitterreceiver systems, or percutaneous devices. Muscle, as a biological amplifier, does not need an
external power source and it will never wear out or need repair.
6
Kuiken, TA
The potential gain from this procedure is high and the risk to the patient is low. The
technique would not compromise any other function in the amputee. Only the residual nerves
are used. The denervated muscles have lost their insertion point and have no mechanical effects.
No functional muscles are denervated with this technique and there would be no impairment of
shoulder biomechanics. It would take 2-6 months after surgery before the muscles became
reinnervated and produced useable myoelectric signals. During this time period the amputee
could still use a conventional prosthesis. This is also consistent with the common practice of first
fitting an upper limb amputee with a body-powered prosthesis. The primary disadvantage is that
the nerve-muscle grafts would most likely be performed as a second surgical procedure.
Although there is always some inherent risk with any surgery, this risk should be minimal in this
elective procedure.
In the examples described above, each major residual nerve is grafted onto a separate
region of muscle to provide a single new myoelectric control site.
If the somatotopic
organization of the peripheral nerve is preserved, it may even be possible to get more than one
control site from a nerve. Each major nerve contains motoneurons to several different muscles.
In the more distal portions of the nerve, the motoneurons of a muscle are grouped together in a
separate fascicles [11]. With distal nerve amputations, it may be possible to separate some of
these fascicles and graft them onto different regions of muscle to produce multiple myoelectric
control sites. For example, the nerve to the pronator teres branches off from the median nerve at
the elbow. In a long transhumeral amputee it may be possible to identify the pronator teres nerve
fascicle and dissect it free for a short distance. The pronator teres nerve fascicle could then be
grafted on to one region of muscle and used to control wrist supination, while the rest of the
median nerve is grafted on to a different region of muscle and this nerve-muscle graft is used to
7
Kuiken, TA
control closing of the terminal device. In this manner it may be possible to further refine and
improve the control of myoelectric prostheses. The process is only limited by the ability to
separate a nerve fascicle with a unique motor function and the ability to isolate the surface EMG
signal from the reinnervated muscle.
3.0 Requirements for Successful Implementation
For the nerve-muscle graft technique to be successful in amputees, multiple nerves would
need to consistently reinnervate separate regions of muscle. Previous studies [12,13] have found
that muscle recovery after nerve transection is quite variable. Such variable recovery could prove
problematic for the nerve-muscle graft technique. However, with the nerve-muscle grafting
technique we would be grafting large nerves containing many times the normal number of
motoneurons onto the muscles thus “hyper-reinnervating” the muscles. As a first step in the
development of this technique we tested the hypothesis that hyper-reinnervating muscle (grafting
an excessive number of motoneurons onto a muscle) would increase the likelihood that any given
muscle fiber would be reinnervated and this improve muscle recovery [14]. In this study rat
muscle was hyper-reinnervated by grafting additional nerves on to the medial gastrocnemius.
Five different nerve graft combinations were performed providing hyper-reinnervation with up to
12 times the number of normal motoneurons. The rats were allowed to fully recover, then
terminal experiments were performed. The maximum twitch and tetanic strengths of both
experimental and contralateral muscles were determined by electrical stimulation of the sciatic
nerve. At the end of the experiments the muscles of interest were removed and weighed. Hyperreinnervation significantly improved the recovery of the denervated muscles. The muscle mass
and muscle force increased as more motoneurons were grafted on the muscle. In the largest
8
Kuiken, TA
nerve-muscle grafts, the experimental muscle recovered to near normal levels with a relative
reinnervation ratio of 94.4+8.2% which was significantly greater than the recovery of selfreinnervated muscles (P<0.005) and was not statistically different from the contralateral
unoperated muscles. Based on these results we can be confident that there would be good muscle
recovery with the nerve-muscle graft technique for amputees.
A related issue is containment of the reinnervation field. With the nerve-muscle graft
technique multiple nerves will be grafted onto different regions of a muscle. It is important that
each nerve reinnervate only the intended muscle region. If there is significant overlap or blending
of the reinnervation fields, then it would be difficult to separate the surface EMG signals from each
nerve-muscle graft. The reinnervating motoneurons actively compete for the available muscle
tissue and it is unknown how multiple nerves grafted on to adjacent regions of muscle divide the
muscle. Animal experiments are underway to clarify these issues.
The remaining key issue for the potential success of the nerve-muscle graft technique is
myoelectric signal independence. Assuming that the nerves reinnervate discrete regions of muscle,
can independent surface EMG signals be recorded from each nerve-muscle graft? The primary
measure of myoelectric signal independence is cross-talk between the surface recording sites; the
unwanted detection of signals from muscles other than the muscle of interest [15]. For the clinical
application of myoelectric prostheses, the electrode is positioned empirically where the prosthetist
determines the strongest signal with the least cross-talk is obtained. Cross-talk can be prevented
from interfering with prosthesis operation by setting a threshold; the threshold is set above
background noise and the cross-talk from nearby muscles. The amputee must generate an EMG
signal greater than the threshold to operate the prosthesis. No published data is available as to
what cross-talk levels are acceptable for myoelectric prostheses, but clearly the greater the cross-
9
Kuiken, TA
talk, the higher the threshold must be set and the harder it is for the amputee to operate the limb. If
the threshold is high, then it is taxing for the amputee to reach EMG levels above the threshold.
This is analogous to lifting. If there is too much noise in the first 10 pounds of lifting then we only
consider the work done from lifting more than 10 pounds. It is very taxing to repeatedly lift at
least 10 pounds to do any movement.
The cross-talk between whole arm muscles, far less between regions of a single arm
muscle, has not been studied to date. Cross-talk is dependent upon a number of factors including
the geometry of the muscle, the geometry of surrounding tissues such as fat and bone, and the
surface recording technique. A finite element analysis of factors affecting surface EMG signal
independence has been performed [16].
This theoretical estimation presents the minimum
muscle size from which an independent myoelectric signal can be recorded. With large muscles
and little subcutaneous fat, a high degree of signal independence can be expected. Cross-talk
increases between smaller muscles and with thicker subcutaneous fat layers. Research is also in
progress exploring ways of increasing myoelectric signal independence with surgical
manipulations such as reducing the subcutaneous fat layer, changing muscle shape and insulating
muscles from each other.
4.0 Synergy with other prosthetic research
Successful implementation of the nerve-muscle graft technique would lead to other
advances in prosthesis design and development. The ability to simultaneously control more
degrees-of-freedom in a prosthesis would serve as an impetus to develop new prosthetic
components. No motorized shoulder components are currently available for shoulder
disarticulation amputees. This is due, at least in part, to inadequate options for controlling such
10
Kuiken, TA
devices. If one has to sequentially control the hand, then the wrist, then the elbow and then the
shoulder it takes too long and the cognitive motor planning burden is too high. The nervemuscle grafting technique would provide better control for such components encouraging their
development.
Applying advanced signal processing techniques to the EMG signal from nerve-muscle
grafts may lead to further improvements in myoelectric prosthesis control. Many different
methods of controlling multiple motions with advanced signal processing techniques have been
studied. Correlation of arm and shoulder behavior with multi-channel EMG recordings has also
been proposed as a process for improving the control of artificial arms [17]. Pattern recognition
techniques have been employed in several laboratories [18-23]. Stochastic time-series analysis of
the temporal signatures of EMG signals has been investigated to see if different muscle
activation patterns can be used as control signals for multifunction capability [24]. This singlechannel technique, however, requires a large computational effort and never progressed to
practical implementation. Recent studies have utilized neural networks [25], fuzzy logic [26] and
wavelet based classification techniques [27]. While some results have been encouraging, the
number of EMG control signals available limits these techniques, especially with high-level
amputations. The proposed nerve-muscle graft technique would increase the control information
available to these complex decoder algorithms so that they could be applied to high-level
amputations with greater success. This could eliminate the need for surface EMGs over the
nerve-muscle grafts to be completely independent of each other and/or make it possible to
control more than one function in a prosthesis with a single nerve-muscle graft.
Combining the nerve-muscle graft technique with cineplasty could also have a number of
advantages. Cineplasty is a form of direct muscle attachment that allows mechanical coupling of
11
Kuiken, TA
a muscle to a prosthesis. In biceps cineplasty, for example, the insertion of the biceps is cut, a
tunnel is made through the distal belly of the muscle and this tunnel is lined with skin. The
prosthesis is then linked directly to the biceps with the tunnel to operate the terminal device. The
brachialis and brachioradialis muscles remain to flex the elbow. Cineplasty has some important
advantages [28]. There is increased sensory feedback through the skin attachments and the
natural muscle proprioceptors that correlates appropriately with function in the prosthesis. It can
also eliminate the need for more proximal harnessing of the prosthesis. However, the technique
has several disadvantages that have limited wide spread application. Obviously, additional
surgery is required. Learning to use the biceps to operate the terminal device and not to bend the
elbow is challenging. The skin in the tunnel is prone to break down secondary to the high forces
required to operate the prosthesis. The long tunnel requires diligent hygiene or infection can
develop.
Several authors have advocated using small cineplasties as servo-controllers of
multifunction externally powered prostheses [28-30]. This would minimize the force required
from the cineplasty, hygiene would be easier and there is still increased sensory feedback. The
nerve-muscle graft technique could be implemented with cineplasty to gain some of the
advantages of direct muscle attachment. The nerve-muscle grafts combined with cineplasty
would provide additional servo-control signals in higher levels of amputation allowing the
simultaneous operation of multiple functions in the prosthesis. There would be increased sensory
feedback through the skin of the cineplasty. Finally, there may be increased feedback through
the reinnervated sensory organs of the muscle.
12
Kuiken, TA
5.0 Conclusions
Grafting the residual nerves of an upper-limb amputee to spare muscles could produce
additional myoelectric control signals.
This may allow simultaneous operation of multiple
functions in an externally powered prosthesis with a more natural feel than is possible with
conventional myoelectric prostheses. The nerve-muscle graft concept has great potential for
improving the function of people with upper-limb amputations—especially for high-level
amputations where the disability is greatest. Further research in this area is warranted.
Acknowledgements
This work was supported by a Biomedical Engineering Research Grant from the
Whitaker Foundation, the National Institute of Child and Human Development (Grant
#1K08HD01224-01A1) and the National Institute of Disability and Rehabilitation Research
(Grant #H133G990074-00).
References
[1]
Sears HH. Trends in upper-extremity prosthetics development. In: ‘Atlas of Limb
Prosthetics,’ Bowker JH, Michael JW, editors., 2nd Ed., St. Louis: Mosby,1992: 345-56.
[2]
Hoffer JA, Loeb GE. Implantable electrical and mechanical interfaces with nerve and
muscle. Ann of Biomed Eng 1980;8: 351-360.
[3]
Sperry RW. The effect of crossing nerves to antagonistic muscle in the hind limb of the rat.
J Comp Neuro 1941;75: 1-19.
[4]
Luff AR, Webb SN. Electromyographic activity in the cross-reinnervated soleus muscle of
unrestrained cats. J Phys 1985;365: 13-28.
13
Kuiken, TA
[5]
Nagano A, Tsuyama N, Ochiai N, Hara T, Takahashi M. Direct nerve crossing with the
intercostal nerves to treat avulsion injuries of the brachial plexus. J Hand Surg 1989;14A:
980-5.
[6]
Clemis JD, Gavron JP. Hypoglossal-facial nerve anastomosis: report on 36 cases with
posterior fossa facial paralysis. In: Disorders of the Facial Nerve. New York: Raven
Press,1982: 499-505.
[7]
Williams TW. New control options for upper limb prostheses. The 10th World Congress of
the Intl Soc of Prosth & Orthot 2001;M06.5.
[8]
Deluca CJ. Control of upper-limb prostheses: a case for neuroelectric control. J Med Eng
and Tech 1978;2(2): 57-61.
[9]
Edell DE. A peripheral nerve information transducer for amputees: long-term multichannel
recordings from rabbit peripheral nerves. IEEE Trans Biomed Eng 1986;33:203-214.
[10] Andrews B. Development of an implanted neural control interface for artificial limbs. The
10th World Congress of the Intl Soc of Prosth & Orthot p. T08, July, 2001.
[11] Jabaley, ME. Internal topography of peripheral nerves as related to repair. In: ‘Operative
Nerve Repair and Reconstruction,’ Gelberman, RH. editor, Philidelphia: Lippincott, 1991:
231-240.
[12] Frey M, Gruber H, Holle J, Freilinger G. An experimental comparison of the different
kinds of muscle reinnervation: nerve suture, nerve implantation, and muscular
neurotization. Plastic and Reconstr Surg 1982;69: 656-667.
[13] Gordon T, Stein RB. Time course and extent of recovery in reinnervated motor units of cat
triceps surae muscles. J Phys 1982;323: 307-323.
14
Kuiken, TA
[14] Kuiken TA, Rymer WZ, Childress DS. The hyper-reinnervation of rat skeletal muscle.
Brain Res 1995;676: 113-123.
[15] De Luca CJ, Merletti R. Surface myoelectric signal cross-talk among muscles of the leg.
Electroenceph Clin Neurophys 1988;69: 568-575.
[16] Lowery, M.M., Stoykov, N.S., Lowery, M.M., Taflove, A. and Kuiken, T.A. An Analysis
of Surface EMG Signal Independence. Subitted to IEEE BME, July 2002.
[17] Jacobsen SC, Mann RW. Control system for artificial arms, IEEE Conference on Man
System Cybernetics, Nov. 1973.
[18] Lyman J, Freedy A, Solomonow M. Studies toward a practical computer-aided arm
prosthesis system. Bulletin Prosth Res 1974;10: 213-225.
[19] Wirta RW, Taylor DR, Finley FR. Pattern recognition prostheses: a historical
perspective—final report. Bulletin Prosth Res 1978;10: 8-35.
[20] Alstrom C, Herberts P, Korner L. Experience with swedish multifunctional prosthetic
hands controlled by pattern recognition of multiple myoelectric signals. Intl Orthopedics
1981;5: 15-21.
[21] Lee SP, Park SH., Kim J, Kim I. EMG pattern recognition based on evidence
accumulation for prosthesis control. Proc Ann Intl Conf IEEE Eng Med Biol 1996;4:
1481-1483.
[22] Farry K, Fernandez J, Abramczyk R, Novy M, Atkins D. Applying genetic programming to
control of an artificial limb. Issues in Upper Limb Prosthetics, Institute of Biomedical
Engineering, University of New Brunswick 1997: 50-55.
[23] Gallant PJ, Morin EL, Peppard LE. Feature-based classification of myoelectric signals
using artificial neural networks. Med & Biol Eng & Comp 1998;36: 485-489.
15
Kuiken, TA
[24] Graupe D, Salahi J, Zhang DS. Stochastic analysis of myoelectric temporal signatures for
multifunctional single-sit activation of prostheses and orthoses. J Biomed Eng 1985;7: 1829.
[25] Hudgins B, Parker P, Scott R. A new strategy for multifunction myoelectric control,”
IEEE Trans Biomed Eng 1993;40:82-94.
[26] Chan FHY, Yang YS, Lam FK, Zhang YT, Parker PA. Fuzzy EMG classification for
prosthesis control. IEEE Trans Rehab Eng 2000;8:305 – 311.
[27] Englehart K, Hudgins B, Parker PA, A wavelet-based continuous classification scheme for
multifunction myoelectric control. IEEE Trans Biomed Eng 2001;48:302-311, 2001.
[28] Weir F, Heckathorne CW, Childress DS. Cineplasty as a control input for externally
powered prosthetic components. J Rehab Res and Devel 2001;38(4):357-363.
[29] Spittler AW, Rosen IE. Cineplastic muscle motors for prostheses of arm amputees. J Bone
Joint Surg 1951;33-A:601-12.
[30] Kessler HH. Cineplasty. Springfield: Charles C. Thomas; 1947.
Abbreviated Title
Myoelectric control with nerve-muscle grafts.
Figure Legend
Figure 1.
Nerve-muscle grafts for transhumeral amputation.
Figure 2.
Nerve-muscle graft system for shoulder disarticulation.
16
Kuiken, TA
Figure 1
17
Kuiken, TA
Figure 2
18