Strategies for providing upper extremity amputees ... - Semantic Scholar

Technology and Health Care 7 (1999) 401–409 IOS Press

401

Strategies for providing upper extremity amputees with tactile and hand position feedback – moving closer to the bionic arm R.R. Riso University of Aalborg, Center for Sensory–Motor Interaction, Fredrik Bajers Vej. 7, Bldg. D3, Aalborg, DK-9220 Denmark E-mail: [email protected] Abstract. A continuing challenge for prostheses developers is to replace the sensory function of the hand. This includes tactile sensitivity such as finger contact, grip force, object slippage, surface texture and temperature, as well as proprioceptive sense. One approach is sensory substitution whereby an intact sensory system such as vision, hearing or cutaneous sensation elsewhere on the body is used as an input channel for information related to the prosthesis. A second technique involves using electrical stimulation to deliver sensor derived information directly to the peripheral afferent nerves within the residual limb. Stimulation of the relevant afferent nerves can ultimately come closest to restoring the original sensory perceptions of the hand, and to this end, researchers have already demonstrated some degree of functionality of the transected sensory nerves in studies with amputee subjects. This paper provides an overview of different types of nerve interface components and the advantages and disadvantages of employing each of them in sensory feedback systems. Issues of sensory perception, neurophysiology and anatomy relevant to hand sensation and function are discussed with respect to the selection of the different types of nerve interfaces. The goal of this paper is to outline what can be accomplished for implementing sensation into artificial arms in the near term by applying what is present or presently attainable technology.

1. Introduction The need to furnish powered prosthetic arms with systems that provide the user with tactile and kinesthetic sensibilities has been recognized since their inception. Ironically, the conversion to myoelectric control and away from the shoulder activated Bowdine cable system for regulating the grasp effort removed the sensory awareness that the shoulder movement provided to the user. What was left for feedback is direct visualization of the prehensor and such subtle clues as the sounds of the speed changes of the motor and transmission. The decade of the seventies saw the development of substitute sensory input based mainly on electrocutaneous stimulation [1–5] or mechanical vibration [6] applied to the skin sense of the residual limb. Those efforts provided coded signals (such as modulation of the electrical pulse repetition rate or vibratory frequency) that the user could learn to relate to either the extent of prehensor opening or the magnitude of the grasp force. Some studies even attempted to provide both grip opening and grip force information simultaneously, but volunteer subjects usually had great difficulty trying to interpret the dual information displays [1]. 2. Direct nerve stimulation The realism of the evoked sensations is improved when the electrical stimulation is applied directly to the trunk nerves of the residual limb [7–9]. By differentially activating specific sensory fascicles of the 0928-7329/99/$8.00  1999 – IOS Press. All rights reserved

402

R.R. Riso / Sensory systems for artifical hands

Fig. 1. Digital nerves that provide cutaneous sensibility to the index finger and thumb.

Fig. 2. Areas of referred sensation in the phantom hands of three BE amputee subjects. Stimulation applied to forearm trunk nerves using percutaneous 0.1 mm stainless steel wires. (After Annani et al. ’79.)

trunk nerves, use is made of the preserved topographic map between the nerve fascicles and the body locus of the perceived sensation. Figure 1 shows the anatomical projections of the median nerve as it divides at the level of the hand to provide sensory innervation to the fingers and palm. Activation of a sensory fascicle in the median nerve (at the level of the forearm) that used to be connected, for example, to the thumb of the amputated hand evokes a sensory experience that seems to emanate from the thumb of the phantom hand (Riso and Keith, unpublished). The spatial discreteness of the evoked sensation is limited by the extent to which the electrical stimulation can be focused to activate small numbers of sensory afferents within a fascicle that projected (in the former intact hand) to adjacent areas on the skin. With the typical macroelectrodes that have been used, such as small (ca. 1 mm diameter) platinum buttons [7], or bare wires ([8], Riso and Keith – unpublished), the spatial resolution for fiber activation is poor and the sizes of the referred sensations are large encompassing an entire finger or much of the palm for example. Figure 2 shows typical results reported by Annani and his colleagues [9] who used a percutaneous wire electrode to stimulate the median nerve in amputee subjects. The quality of the perceived sensation, moreover, remains a foreign feeling resembling a vibration, taping or flutter on the skin. This results because a macroelectrode activates many different types of cutaneous afferents all at once and with an unnatural synchronicity that is phase locked to the stimulus pulse train.


403

3. Physiological substrate for tactile sensibility In principle, the neural encoding of tactile events relies on a digital communication system that has four types of labeled lines (the classic four types of cutaneous mechanoreceptors) and consists of thousands of channels who’s activity is continuously polled by the central nervous system. In an intact hand, the intensive aspects of a mechanical event impressed on the skin are signaled by modulating the frequency of the discharge of the stimulated afferents and in the case of an increasing stimulus amplitude, by recruiting additional similar afferents that have higher mechanical activation thresholds. The spatial extent of the perceived sensation (called a perceptive field) is governed by the dispersion of the skin projection fields of the activated afferents. The sizes of the skin regions for activation of FAI (Meissner) and SAI (Merkle) afferents are small punctate circles or ellipses having dimensions of about 3–4 mm. FAI afferents detect object contact/release and slippage whereas SAI afferents are more involved with details of the surface shape of the contacting object and tactile gnosis. A third class of mechanoreceptor, named the Pacinian Corpuscle (FAII) has such extremely high sensitivity to mechanical deformation that a stimulus applied to the skin several centimeters away from its location might propagate enough disturbance through vibration to cause it to discharge [10]. The final class of cutaneous mechanoreceptors, named Ruffini (SAII) endings, are spindle shaped receptors that are activated by shear forces. They attach in different directions within the skin, and a given receptor is activated when the direction of an applied stretch causes that receptor to be stretched. The receptor discharge continues as long as the stretch is present. Because of the presence of residual shear forces in the skin, an SAII receptor may exhibit ongoing activity that is diminished if the relevant skin area is stretched in the direction that acts to unload that particular receptor. Many SAII receptors are found in the vicinity of the joint creases, where they are stretched or relaxed as the joint position changes. Because of this, SAII afferents may play a role in signaling joint position [10].

4. Microstimulation of tactile afferents With the development of microneurographic techniques, it was demonstrated in normal subjects, that more natural sensations can be evoked when individual cutaneous afferent fibers are electrically activated in isolation. Thus, activation of a Meissner (FAI) afferent with a single pulse evokes the sensation of a single tap over a small skin area about 2–3 mm in diameter. A train of stimuli evokes a sensation of flutter or vibration with the frequency related to the stimulus frequency. Activation of a Merkle (SAI) afferent with a train of stimuli results in a quality of sustained indentation of the skin, and the magnitude of the indentation effect is related to the frequency of the stimulus pulses. Stimulation of SAI afferents could provide the substrate for artificial feedback of grasp force. Activation of a Pacinian (FAII) afferent behaves similarly to the FAI, with a single pulse eliciting a tap sensation, and activation with a stimulus train evoking a vibratory sensation [11,12]. A priori, electrical activation of the stretch sensitive SAII afferents would appear to be useful for the artificial feedback of hand position sense. Unfortunately, however, investigators have observed that if such an afferent is electrically activated via microneural stimulation, no sensory experience is evoked. One explanation for this finding is that it might be necessary for multiple Ruffinian afferents to be active at the same time in order for this activity to reach consciousness [11,12]. That speculation does not offer much solace for the development of a sensory feedback system that is based on micro-stimulation,

404


however, since it is unlikely that several Ruffinian afferents from the same skin area and with similar preferred direction of stretch would be able to be connected to for simultaneous electrical activation. Therefore, the sensation of skin stretch may not be able to be evoked artificially.

5. What can be done to provide an artificial arm with kinesthetic sensibility? Building on earlier work by Goodwin et al. [13], Roll and Gilhodes demonstrated [14] that if a localized vibratory stimulus is applied over the tendons at the wrist, normal subjects perceive movement of the wrist joint when in fact there is no motion. The direction of the perceived motion depends upon which tendons receive the vibration. When vibration is applied to the wrist flexor tendons, for example, the perceived motion is that of wrist extension. When vibration is applied to the wrist extensors, the imagined movement appears to be (just opposite) in the direction of wrist flexion. This phenomenon can be extended to include sensations of ulnar or radial deviation as well. To evoke a sensation of ulnar movement, the vibration is applied to the radial side of the wrist, while movement in the radial direction is evoked by vibration of the tendons that attach at the ulnar side of the wrist. Finally, a perception of oblique motion can be evoked by simultaneously vibrating flexor and ulnar tendons. To explain these effects, it has been theorized that vibrating the flexor tendons induces neural activity in the stretch receptor afferents of the flexor muscles [13]. These afferents would ordinarily be active when the parent muscles are stretched by wrist extension, and therefore, the illusionary movement is perceived as wrist extension. If the flexor and extensor tendons are vibrated simultaneously, the direction of the perceived movement is governed by which set of muscle afferents has the greater activity. In these experiments, the illusionary movement direction is also influenced by adjusting the relative amplitude and frequency of the vibratory stimuli used. The subject’s brain appears to apply a kind of vectorial summation of these inputs. One additional characteristic is that the velocity of the movement can be increased by increasing the frequency of the vibration [14]. Such findings suggest that if the appropriate spindle afferents could be stimulated in an amputee’s sensory nerve, then feedback of hand position might be possible.

6. Protective and temperature sensibility A warning could be provided that an object was too warm (or cold) or that the pressures exerted on the ‘skin’ of the prosthetic hand were excessive. In the former case, thermal sensors within the prosthesis finger skin could monitor the contact temperature. If it was not within a safe range, the afferents that subserved the most relevant skin region could be driven at an intensity level that activated A-delta and/or C fibers to evoke a sensation of pain. Activation of the pain afferents would require elevated stimulation amplitudes since (owing to their small fiber size) their electrical activation thresholds are substantially higher than for tactile afferents [15]. It is unlikely that electrical stimulation could create a natural sensation of temperature, because this would require activation of (A-delta) temperature afferents without activating the pain fibers of the same axonal caliber. However, a Peltier device applied against the stump skin might be useful to apply heating or cooling as a kind of remote feedback site for temperature sensing. The device could be self activating to warn against extreme temperatures or switched on voluntarily to provide graded temperature information.


405

7. Nerve interfaces Figure 3 illustrates four types of nerve interfaces under development in diverse laboratories that could find application for prosthesis sensory feedback. The type of interface depicted in Fig. 3(a) is popularly called a “sieve” electrode [16–19]. The example shown is based on a 10 µm thick polyimide sheet containing a matrix of small holes each of which represents an addressable electrode contact. When the interface is installed at the end of a transected nerve, some of the fibers succeed in regenerating through the via holes making intimate contact with the electrode contacts. Presently, the sizes of the holes are such as to allow small bundles of axons to grow through. While high densities of holes are possible, if the hole size is reduced sufficiently to favor the inclusion of a single axon, successful regeneration seems to be severely impeded [19]. For this reason, the sieve type of interface is presently a viable candidate for nerve activation at the level of small bundles of axons. Another concept (Fig. 3(b)) for a regeneration type of nerve interface has been to position 25 µm platinum-iridium wires with exposed tips co-axially into the lumen of a quartz regeneration tube [20]. This simple design affords good neural regeneration. Figure 3(c) depicts a “brush” array of micro-machined needle electrodes (interdistance = 120 µm) that is inserted into a peripheral nerve. The structure [21] is realized using aligned X-ray lithography and galvanic shaping (LIGA) on a silicon substrate. The interconnect wiring consists of 8 µm Cu. Finally, it has also been shown possible to interface with peripheral nerves by threading fine wires (25 µm dia.) into a nerve fascicle [22], as depicted in Fig. 3(d). The wire is insulated except for a small window

Fig. 3. Illustrating four approaches toward developing a peripheral nerve interface. The interfaces in (a) and (b) require that a transected nerve regenerate through them. The devices in (c) and (d) can be installed into an intact section of peripheral nerve. In the drawing of device (d), the left boarder of the nerve is shown opened up to reveal the wire stimulation sites which are otherwise hidden within the nerve. A common return electrode site is positioned laterally and on the outside of the nerve. The illustrations in (c) and (d) are modified from original artwork kindly provided by W. Rutten [21] and by K. Yoshida [22], respectively.

406


of about 1 mm length which serves as the electrode active site. Because of the stiffness of the metal filament relative to the nerve tissue, long term stability of this type of electrode (called a Longitudinal Intrafascicular Electrode or ‘LIFE’) is still an issue. Ongoing efforts to develop conductive fibers from ultra compliant polymer filaments may minimize this potential problem. It seems unlikely, however, that vast numbers of LIFES could be installed into any given nerve fascicle, and that this approach should probably also be regarded as a means to activate small bundles of afferents within selected fascicles. 8. How many channels of stimulation are needed? Tactile feedback to provide a prosthesis user with sensations similar to a natural hand would require independent micro-stimulation of hundreds of ‘on line’ tactile afferents in order to include the various modalities and to access most of the important skin areas contacted during object manipulation. The median nerve is estimated to contain about 14 000 tactile afferents subserving the hand. The greatest density is from the finger tips, which contain about 141, 70 and 9/cm2 , FAI, SAI and SAII afferents, respectively. The palm in comparison, has only about 25, 8 and 16 afferents/cm2 [10]. Fortuitously, this difference implies that an afferent that by happen-stance was able to be activated by a neural interface would be an FAI or SAI type subserving a finger tip (the most important area involved in manipulation!). Another bonus for the developer is that activation of a single afferent (except SAII) can evoke a sensory percept of reasonable magnitude. Thus, interfacing with just a single FAI or (better) SAI afferent mapped to each finger tip could provide useful sensory feedback. Interfaces needed for single fiber activation, however, still require substantial developmental effort. In a series of chronic implantation studies with the rabbit tibial nerve as a model, the neural interface consisting of the quartz chamber with axially placed fine wires that was described in Fig. 3(b) was demonstrated to evoke sensory percepts for more than three years with stimulation currents on the order of 30 µA [20]. Such low activation thresholds, suggest that the stimulation could be focused to activate very few fibers. With the behavioral eye blink test paradigm that was used to detect the nerve activation, however, it is not possible to know how many fibers were being activated. 9. Summary and conclusions This paper has attempted to describe the developmental status of sensory systems for artificial hands, and to provide a perspective about what could be implemented in the near term using presently available technology. Figure 4 presents a schematic of an envisioned sensory feedback system. Implanted components consist of a multichannel stimulator, telemetry receiver and a neuro-interface. A signal conditioner and stimulator controller module with telemetry is placed into the prosthesis shell and powered by a battery. Tactile and joint position sensors are integrated into the finger joints and skin like covering of the hand to provide good cosmesis. Present FSR technology could be molded into the soft hand covering in the region of the finger tips and possibly also the palm, with thin ribbon conductors serving as leads. Hall effect sensors or fiber optic systems crossing the finger joints could provide finger position signals. Prior developments such as the SOUTHAMPTON and MARCUS hands have already demonstrated the utility of such sensors in providing automatic grasp control function [23–25] (and see [26] for a general discussion of sensor technology). A transcutaneous link provides means to power the implanted electronics and to deliver commands to the stimulator channels. There are no exposed wires to cause problems of lead breakage. As a result of


407

Fig. 4. Showing an envisioned powered arm prostheses with an integral system that provides cognitive sensory feedback of finger position and grasp forces.

developments in FES and Cochlear neuroprostheses, a number of commercial stimulators are presently available and the parameters for safe nerve stimulation are becoming clearly defined. With regard to the selection of an appropriate neural interface, electrical stimulation of small bundles of afferents is an easier goal presently then activation of isolated afferents, and it has the advantage that all of the grasping surfaces of the hand could be accommodated with a modest number of independent stimulation channels. The major drawback of not being able to stimulate discrete afferents is that the evoked sensations will always have an unnatural feeling since a mixture of afferents sub-serving different modalities may be co-activated indiscriminately. The observation that cutaneous afferents seem to run in separate fascicles from muscle afferents within the trunk nerves should facilitate the independent activation of tactile and muscle afferents. Prosthesis manufactures in collaboration with the robotics industry should be expected to provide additional advances in powered, sensorized manipulators. All of the components for the prosthesis system shown are within close reach. To a large extent only economic considerations separate the prosthesis in the drawing from becoming reality. Hopefully, responsible commercial partners and third party payers will now agree that it is time to implement today’s technological achievements and deliver the next generation of powered hands to the rehabilitation community.

408


Acknowledgement Support for the preparation of this manuscript was provided by the “GRIP” project (No. 26322) of the European Union Esprit Program. References [1] R.E. Prior, J. Lyman and P.A. Case, Supplemental sensory feedback for the VA/NU myoelectric hand: background and feasibility, Bulletin Prosthetics Res. 10(26) (1976), 170. [2] G.F. Shannon, A myoelectrically controlled prosthesis with sensory feedback, Med. Bio. Eng. Comput. 17 (1979), 73. [3] A.Y.J. Szeto, J. Lyman and R. Prior, Electrocutaneous pulse rate and pulse width psychometric functions, Human Factors 21(1) (1979), 241. [4] A.Y.J. Szeto and R.R. Riso, Sensory feedback using electrical stimulation of the tactile sense, in: Rehabilitation Engineering, R.V. Smith and J.H. Leslie, eds, CRC, Boca Raton, 1990, pp. 29–78. [5] R.N. Scott, R.N. Brittai, R.R. Caldwell, A.B. Cameron and V.A. Dunfield, Sensory feedback compatible with myoelectric control, Med. Biol. Eng. Comput. 18 (1980), 65. [6] R.W. Mann and S.D. Reimers, Kinesthetic sensing for the EMG controlled “Boston Arm”, IEEE Trans. Man Mach. Syst. 11(1) (1970), 110. [7] F.W. Clippinger, A system to provide sensation from an upper extremity amputation prosthesis, in: Neural Organization and Its Relevance to Prosthetics, W. Fields and L. Leavitt, eds, Intercontinental, NY, 1973, p. 165. [8] V. Mooney and J.B. Reswick, Sensory feedback myoelectric prosthesis, Final Report to VAPC, 1976. [9] A. Anani and L. Korner, Discrimination of phantom hand sensations elicited by afferent electrical nerve stimulation in below-elbow amputees, Med. Prog. Through Technol. 6 (1979), 131. [10] Å.B. Vallbo and R.S. Johansson, Tactile sensory innervation of the glabrous skin, in: Active Touch, G. Gordon, ed., Pergamon, Oxford, 1979, pp. 29–54. [11] A.B. Vallbo, K.A. Olsson, K.G. Westberg and F.J. Clark, Microstimulation of single tactile afferents from the human hand, Brain 107 (1984), 727–749. [12] J. Ochoa and E. Torebjork, Sensations evoked by intraneural microstimulation of single mechanoreceptor units innervating the human hand, J. Physiol. 342 (1983), 633. [13] G.M. Goodwin, D.I. McCloskey and P.B.C. Matthews, The contribution of muscle afferents to kinesthesia shown by vibration-induced illusions of movement and by effects of paralysing joint afferents, Brain 95 (1972), 326–329. [14] J.P. Roll and J.C. Gilhodes, Proprioceptive sensory codes mediating movement trajectory perception: human hand vibration-induced drawing illusions, Can. J. Physiol. Pharm. 73 (1995), 295–304. [15] R.G. Hallin and H.E. Torebjor, Electrically induced A an C fiber responses in intact human skin nerves, Exp. Brain Res. 16 (1973), 309. [16] D.J. Edell, A peripheral nerve information transducer for amputees. Long term multichannel recording from rabbit peripheral nerve, IEEE Trans. Biomed. Eng. 33 (1986), 203–214. [17] G.T. Kovacs, C.W. Storment, M. Halks-Miller, C.R. Belczynski, C.D. Santina, E.R. Lewis and N.I. Maluf, Siliconsubstrate microelectrode arrays for parallel recording of neural activity in peripheral and cranial nerves, IEEE Trans. Biomed. Eng. 41 (1994), 567–577. [18] T. Stieglitz, H. Beutel and J.-U. Meyer, A flexible, light-weight multichannel sieve electrode with integrated cables for interfacing regenerating peripheral nerves, Sensors and Actuators 60 (1997), 240. [19] T. Laurell, J. Drott, Q. Zhao, L. Wallman, L.M. Bjursten, G. Lundborg and N. Danielsen, Perforated silicon membranes as an artificial neural contact pad-chip design and tissue implant response, in: Proceedings IEEE/EMBS Annual Meeting, paper #330, 1996. [20] D.J. Edell, R.R. Riso, L. Devaney, B. Lartsen, M. Koris and D. DeLorenzo, Intraneural microstimulation for enhanced prosthetic control using a peripheral nerve interface, in: Proceedings IEEE/EMBS Annual Meeting, paper #1124, 1996. [21] W.L.C. Rutten, J.P.A. Smit, T.A. Frieswijk, J.A. Bielen, A.L.H. Brouwer, J.R. Buitenweg and C. Heida, Neuro-electronic interfacing with multi electrode arrays and with cultured multi electrode plates: force recruitment, selectivity, microfabrication, cell adhesion and cell containment, Electronics in Medicine and Experimental Biology (May 1999) (in press). [22] K. Yoshida and K. Horch, Selective stimulation of peripheral nerve fibers using dual intra-fascicular electrodes, IEEE Trans. Biomed. Eng. 40 (1993), 492–494. [23] P.J. Kyberd, N. Mustapha, F. Carnegie and P.H. Chappel, A clinical experience with a hierarchically controlled myoelectric hand prosthesis with vibrotactile feedback, Prosthetics and Orthotics International, 1993. [24] M.C.F. Castro and A. Cliquet, A low-cost instrumented glove for monitoring forces during object manipulation, IEEE Trans. Rehab. Eng. 5(2) (1997).


409

[25] P.J. Kyberd and E. Owen, Holland, MARCUS: A two degree of freedom hand prosthesis with hierarchical grip control, IEEE Trans. Rehab. Eng. 3(1) (1995). [26] J.G. Webster, Artificial sensors suitable for closed-loop control of FNS, Chapter 4, in: Neural Prostheses: Replacing Motor Functions After Disease or Disability, R.B.Stein, P.H. Peckham and D.P. Popovic, eds, Oxford University Press, 1992.