Improving the Human-machine Interface: Revolutionary Advancements in Upper-limb Prostheses

Content provided by The O&P EDGE
Current Issue - Free Subscription - Free eNewsletter - Advertise
cybernetic arm

There are many limitations associated with the current interface between the residual limb and the prosthesis for individuals with upper-limb deficiencies. The suspension of the device, which can be heavy, is often a product of awkward harnessing, localized pressures over bony elements in the residual arm, or sometimes unreliable suction environments. Prosthetic control is both limited and limiting. In body-powered devices, unrelated gross body motions create restricted movements of the prosthesis, while in externally powered systems, the activation of residual muscle bellies provides control of distal prosthetic joints that are often removed from their pre-amputation functions. At more proximal amputation levels, the problem is compounded by the need for single muscles to control multiple joints through a series of artificial myoelectric signals and patterns. Sensory feedback remains limited or absent. Body-powered designs offer some sense of the amount of prehensile force experienced at the terminal device, while externally powered designs fail to provide any sensory feedback. These limitations are exacerbated by their influences on one another. Compromises in suspension, for example, can alter the position of the electrodes relative to the target muscle bellies, further compromising control of the system.

Collectively, these limitations have undermined the successful use of upper-limb prostheses. Committed users accomplish less than they would like to with their devices, while others abandon the use of their upper-limb prostheses entirely. The rehabilitation community is aware of these limitations and progress toward their individual and collective resolution continues. As a striking example, in 2014 European researchers described a case study in which a transhumeral prosthesis is suspended to the residual limb through an osseointegrated abutment that also houses the electrical wires for internal electrodes that provide motor control signals and relate sensory input to the residual limb.1 This article reviews the progress of a number of revolutionary technologies and approaches that will ultimately improve the state of science in the realm of rehabilitation following upper-limb amputation.


Over the years, clinicians have developed many creative methods for suspending external prostheses from residual limbs. In upper-limb applications, principles related to harnessing, suction, and anatomic socket contours continue to be explored. However, those strategies seek to anchor the prosthesis over the limb rather than directly from the bony elements of the residual limb through osseointegration. Osseointegration is an increasingly familiar technique that began in dental and facial prostheses, found application in lower-limb prostheses, and was quickly, though less aggressively, explored in upper-limb devices.

While surgeons in Germany and Australia have begun performing upper-limb osseointegration surgeries, the pioneers in this approach have been Rickard Branemark, MD, MSc, PhD, and his colleagues in Sweden. A report published in 2011 canvasses the progress they had made to that point in fitting 37 upper-limb cases with osseointegrated prostheses.2 The first application of the technique in upper-limb deficiency was reported in association with a thumb amputation in 1990. This was followed two years later by transradial applications, with transhumeral applications beginning in 1994.2

Consistent with the more familiar transfemoral osseointegration procedure, the upper-limb procedure consists of three components installed in two stages. In the first stage, the intramedullary fixture is surgically implanted into the intramedullary canal of the residual long bone (or bones) of the limb and the surgical site is closed. A six-month healing period ensues in which the implant remains unloaded as bone grows into threads of the implant. This is followed by the second stage of the procedure in which the distal aspect of the implant fixture is exposed and a transcutaneous abutment is connected and secured to the implant fixture with an abutment screw. This distal abutment then acts as the attachment point for the external prosthesis.2

The authors report on 16 transhumeral cases, ten transradial cases (inclusive of a single bilateral case), ten thumb amputations, and two partial hand cases that occurred in the 20 years that followed the initial thumb amputation case. Of those, 30 individuals remained prosthetic users at the time of publication.2 For the more distal applications of partial hand and thumb integrations, only cosmetic prostheses were used, and the prostheses were attached shortly after the second surgical stage. At the more proximal, major amputation levels, prosthetic use began either with lightweight devices or shortened training prostheses, with a progressive transition to heavier, more functional prosthetic options. Ultimately, cosmetic, body-powered, and myoelectric devices have been successfully utilized at the transradial and transhumeral levels.2 At this point in the evolution of osseointegration, in cases of myoelectric control, external electrodes extending proximally from the prosthesis were used to position the electrodes over the distal muscle bellies of the residual limb.2

cybernetic arm


The current standard of control in externally powered upper-limb prostheses is dual-site surface myoelectric control. While this provides passable operation, it has limitations. The governing muscle bellies are often called upon to sequentially control one or more pairs of unnatural antagonistic prosthetic movements. Some of these limitations have been commercially addressed in the form of the Coapt pattern recognition system. Unlike the aforementioned direct control scheme, in which a single, isolated muscle signal is used to control one or more prosthetic movements, pattern recognition has been described as an indirect control scheme. In this approach, multiple surface electrodes are positioned around the circumference of the limb. Rather than training the user to generate isolated, nonphysiologic control signals, the computer software is trained to recognize the native muscle contraction patterns that the user associates with a series of prosthetic movements. Once the processor has been programmed, when these myoelectric patterns are recognized, the desired prosthetic movements are signaled within the prosthesis.

However, despite these advantages, the system is still subject to the inconsistencies associated with socket-mounted surface electrodes. These include their susceptibility to environmental electrical noise, movement of the surface electrodes over the skin that occurs with different orientations of the residual limb in space or incidental socket movement over the limb, and the effects of skin perspiration with its associated changes in electrical impedance and resulting signal strength.3 The solution to such limitations may be found in the form of surgically implanted electrodes. One of the more progressive approaches is the Implantable MyoElectric Sensor (IMES) System.3

IMES are small, cylindrical electrodes, 16mm long and 2.5mm in diameter, that detect and wirelessly transmit EMG data.3 Surgically placed within the muscles, potential advantages include stronger, more reliable signals that are not affected by the position of the arm relative to the body, socket rotation, or perspiration. Because they are implanted, available myoelectric signals are no longer confined to large, superficial muscles, but could include deeper muscles as well. Access to more muscles allows myoelectric control to be more physiologic (i.e., wrist supination controlled by residual wrist supinators) and simultaneous rather than sequential.3

Researchers at Walter Reed National Medical Military Center, Bethesda, Maryland, have initiated the first U. S. Food and Drug Administration (FDA)-approved human feasibility trial of the IMES System for patients with transradial-level amputations. With a one-to-one relationship between electrodes and each controlled degree of freedom, the first patient was implanted with eight electrodes. Six of the sensors were intended to control the available movements of opening and closing the hand, wrist pronation and supination, and thumb abduction and adduction. The two additional sensors were implanted to provide potential backup signals if one of the target IMES failed to produce adequate signals, or they could be used to control additional degrees of freedom should they become available within the system.3 More specifically, the extensor and flexor digitorum, extensor and flexor pollicis longus, and the pronator teres and supinator were used to control their native motions while the flexor carpi ulnaris and extensor carpi radialis were implanted but originally unassigned.3

Each IMES “acts as a wireless, independent, intramuscular differential amplifier” that reads and processes myoelectric activity, digitizes the information, and then transmits it wirelessly to other, external components.3 A coil laminated within the wall of the prosthetic frame creates a magnetic field. Forward telemetry from the field transmits power and configuration settings to the IMES while reverse telemetry into the magnetic field transfers signals from the imbedded sensors.3

This information is then transferred to the prosthetic control interface—a Walkman-size box worn on the user’s belt—that receives the original data from the sensors for additional processing and then drives the motor control of the prosthesis to produce the desired movements.3

Because of the limitations of the FDA-approved protocol, the initial reports on the IMES are limited to descriptive results, but these appear promising, with the subject demonstrating an ability to control the prosthesis almost immediately upon beginning his prosthetic training three weeks after surgery.3 All three available degrees of freedom (hand, wrist, and thumb) were controlled individually and simultaneously. Further, the subject’s control of his prosthesis was unaffected by the orientation of his limb in space, the relative position of the socket, or perspiration.3

Sensory Input

The absence of sensory input represents a fundamental limitation to externally powered upper-limb prostheses that prohibits users from associating their prostheses as part of themselves, seeing the prostheses as tools rather than an extension of themselves.4 Historically, the restoration of sensory input has been attempted through surrogate sensations, such as vibration, somewhere on the body’s surface. While this has demonstrated some benefit in laboratory settings, it has not been aggressively pursued, presumably because the surrogate sensations are not natural, and thus require additional user cognitive effort.4 Ideally, sensory feedback should be experienced more similarly to that associated by the natural limb. With the introduction of surgically implanted electrodes, the possibility of communicating sensory input to sensory nerves is currently being explored.4

cybernetic arm

The research team with the longest tenure of successful sensory implantation recently released a publication updating its status.4 Two subjects have been implanted with extraneural cuff electrodes, around the median, ulnar, and radial nerves of the forearm of one subject and the upper arm of the other subject, for 32 and 40 months, respectively. In both cases, the associated lead wires from the electrodes were routed percutaneously to exit the body at the lateral aspect of the arm.

With the cuff electrodes, several different stimulation types were found to be possible, including pressure, vibration, and paresthesia. Pressure was identified as the target sensory input. In addition, one of the subjects experienced flexion of his third digit. A prosthetic hand was instrumented with low-profile force sensors over the pads of the thumb and of the index and middle fingers. As the force applied to the sensors increased, the frequency of the sensory stimulation increased linearly. In addition, a bend sensor was installed to relay the relative opening aperture of the hand.4 For the subject who could experience third digit flexion, this sensor signaled that sensation. In the other subject, information from the bend sensor was translated into the sensation of relative pressure against the thenar eminence.4

This done, the subjects have participated in training sessions and functional tests. One such test, the object detection test, required the users to determine whether a wooden block had been placed within their grasp while they were blindfolded. As hoped, the effect was striking. While they correctly identified the presence or absence of the object just over half of the time without sensory input, they did so with an accuracy of 89 percent and 96 percent, respectively, when pressure and hand-span inputs were provided.4 Further, the confidence of these subjects in their blinded determinations increased substantially with the additional sensory input, from 35 percent to 84 percent. Improvements in the more familiar Box and Blocks Test and Southampton Hand Assessment Procedure were also reported.4

Of additional importance, these results have occurred with no complications from infection. The authors report 1,568 infection-free percutaneous lead-months (i.e., the product of the total number of percutaneous leads and the duration through which they’ve been implanted).4


Viewed independently, the advancements of osseointegrated suspension, discrete myoelectric control through internal electrodes, and the provision of sensory feedback each constitute encouraging progress toward more integrated, functional upper-limb prostheses. To view them collectively borders on unbelievable, and yet these technologies are being successfully combined by Branemark’s Swedish team as an “osseointegrated human-machine gateway.”1

In the 2014 publication previously mentioned, the team reported about a patient with a transhumeral-level amputation who underwent osseointegration early in 2013. But the advancement of his rehabilitation technology goes beyond that. Prior to the osseointegration, researchers had used surface myoelectric signals to obtain the ideal electrode sights for a pattern-recognition-based myoelectric control scheme that provided 94 percent accurate control of eight motions (opening and closing the hand, wrist pronation and supination, wrist flexion and extension, and elbow flexion and extension). At the time of surgery, epimysial electrodes (surgically attached to the epimysium of the muscle bellies) were placed at these sites, and their leads routed through the osseointegrated abutment where they could transmit directly to the prosthesis. Following a few weeks of training, the pattern-recognition-based system reading internal electrodes was functioning across all eight prosthetic movements with 99 percent accuracy.1

In addition to the advanced motor control, a single spiral cuff electrode was surgically implanted around the ulnar nerve, which allowed the subject to experience sensory feedback in the form of a “superficial tapping” in his fourth and fifth digits and hypothenar eminence.

The patient reported his ability to perform his activities of daily living and vocational responsibilities without the familiar limitations experienced in lesser prosthetic systems. Movement of the limb was no longer restricted, nor was control affected by limb position or other environmental conditions. The system is characterized as providing higher controllability with less effort, with the patient reporting no muscular fatigue despite using his prosthesis 16 to 18 hours per day.1


Branemark’s osseointegrated human-machine gateway, and the inclusive technologies that are being advanced by other researchers, suggest a future in which the fundamental limitations that characterize modern upper-limb prosthetic rehabilitation may be reasonably addressed by providing consistent connection, control, and sensory feedback.

Phil Stevens, MEd, CPO, FAAOP, is in clinical practice with Hanger Clinic, Salt Lake City. He can be reached at .


  1. Ortiz-Catalan, M., B. Hakansson, and R. Branemark. 2014. An osseointegrated human-machine gateway for long-term sensory feedback and motor control of artificial limbs. Science Translational Medicine 6 (257):257re6.
  2. Jönsson, S., K. Caine-Winterberger, and R. Branemark. 2011. Osseointegration amputation prostheses on the upper limbs: Methods, prosthetics and rehabilitation. Prosthetics and Orthotics International 35 (2):190-200.
  3. Pasquina, P. F., M. E. Evangelista, A. J. Carvalho, J. Lockhart, S. Griffin, G. Nanos, P. McKay, M. Hansen, D. Ipsen, J. Vandersea, J. Butkus, M. Miller, I. Murphy, and D. Hankin. 2015. First-in-man demonstration of a fully implanted myoelectric sensors system to control an advanced electromechanical prosthetic hand. Journal of Neuroscience Methods 244:85-93.
  4. Schiefer, M., D. Tan, S. M. Sidek, and D. J. Tyler. 2016. Sensory feedback by peripheral nerve stimulation improves task performance in individuals with upper limb loss using a myoelectric prosthesis. Journal of Neural Engineering 13 (1):06001.

Bookmark and Share