Methods of Providing Sensory Feedback to Wearers of ... · Methods of Providing Sensory Feedback to...
Transcript of Methods of Providing Sensory Feedback to Wearers of ... · Methods of Providing Sensory Feedback to...
Methods of Providing Sensory Feedback to Wearers of Prosthetics
John T. Johnson
Georgia Gwinnett College
13-Jul-2012
The hand’s utility as an end effector is not simply because it is under our command to grasp, release, and
otherwise manipulate the world around us. Part of the hand’s utility comes from its ability to provide feedback
about the object it is manipulating. Haptic feedback provides the sense of touch, texture and slippage that is
important for grasping and manipulating objects. This feedback allows secure holding of objects, while
moderating the force applied to the object (Westling, Johansson, 1984). Proprioceptive feedback provides the
peripheral and central nervous systems with location and stress information for the extremities. Common
prosthetics and many currently in development do not provide these sources of feedback to the wearer. Lack of
haptic feedback means the wearer has no sense of the pressure applied to an object in the prosthetic hand, or
even whether the object has fallen out of the hand. Lack of proprioceptive feedback means the wearer of the
prosthetic has no sense of where the prosthetic is without looking at it (Atkins, Heard & Donovan, 1996).
Atkins et al. (1996) conducted a survey of individuals with upper limb loss. Of the 1,575 surveyed,
1,020 used body-powered prosthetics, 438 used electric hands, and 117 were bilateral amputees. The survey
identified many areas in which prosthetics could be improved. Transhumeral amputees ranked, "required less
visual attention to perform certain functions,” third. Amputees with an electric prosthesis also ranked this as
their third priority (Atkins et al., 1996). Accuracy of localization of the index finger was improved with vision
and proprioceptive feedback, more so than with either vision or proprioception alone (van Beers, Sittig, Denier
van der Gon, 1996). Therefore, providing proprioceptive feedback would help fulfill the requirement of less
visual attention to perform certain functions.
One popular method of providing grip-force information to the prosthetic wearer is vibrotactile feedback.
Vibrotactile feedback is provided by a vibrating motor or piezoelectric transducer placed in proximity to the
skin of the wearer. The amount of vibration is proportional to the amount of force being applied by a
myoelectrically-controlled prosthetic hand. When used with vision, this feedback system provides better grip-
strength accuracy than vision alone. However, it has been shown that pressure applied to the user is a better
solution than vibration, perhaps because the pressure applied that is relative to the grip-strength is of the same
modality as the wearer would experience gripping an object with their now-absent hand. That is,
pressure=pressure is better than vibration=pressure (Patterson & Katz, 1992).
Targeted reinnervation is a recently developed technique for transhumeral and shoulder disarticulation
amputees that allows myoelectric control of a prosthesis using remaining nerves from the patient’s residual limb
to convey information to the wearer’s nervous system. The brachial plexus nerves that would have controlled
functions in the patient’s amputated limb are surgically moved to skin and muscle on another site of the body.
Typically this is the ipsilateral pectoral muscle.
Figure 1 Illustrating targeted reinnervation surgeries of BSD and STH. Green=arm nerves, blue=skin afferent nerves, dotted lines indicate nerves are deep to muscle. (A) illustrates patient BSD. The pectoral muscles were denervated, subcutaneous fat was removed, and the four branches of the brachial plexus nerves were attached to nerves emanating from the muscle segments. (B) illustrates STH. To promote regeneration, the skin nerves were cut and sewn to the hand nerves. To avoid disfiguring the patient, subcutaneous fat and breast tissue were not removed. (C) illustrates the skin areas serviced by the brachial plexus nerves in a normal hand (Kuiken et al., 2007).
Reinnervation surgery was performed on patient BSD, a man 54 years of age. BSD received electrical
burns on both arms requiring disarticulation at the shoulder along with a local skin graft. The patient presented
with pain from the skin grafts that required surgery. At that time, the decision was made to try targeted
reinnervation during the revision surgery. A diagram of BSD’s surgery may be found in Figure 1A. The surgery
proceeded as follows (Kuiken, Dumanian, Lipschutz, Miller, & Stubblefield, 2004).
The target pectoral muscle was first denervated which removes the existing afferent and efferent nerves.
The target muscle was then sectioned and divided to provide more muscle units that the anastomosed nerves
could control. The muscle was divided to maintain innervation and blood supply. The pectoralis major and
minor muscles were dissected to allow preservation of the lateral and medial pectoral nerves. The existing
groove between the clavicular head and sternal head of the pectoralis major was increased and the clavicular
head nerves were marked (Kuiken et al., 2004).
The median, musculocutaneous, radial and ulnar nerves were dissected free and marked for
identification. The scarred nerves were trimmed back to healthy portions. At this point the pectoralis major was
divided into three sections. The nerve fascicles were attached to the muscle sections created, and the nerves
were carefully routed to prevent any unwanted reinnervations. Beginning at the cephalic aspect of the chest and
proceeding caudally the muscle sections and their anastomosed nerves were as follows: clavicular head of
pectoralis major, musculocutaneous nerve; sternal head of pectoralis major, median nerve; lower segment of
sternal head, radial nerve; pectoralis minor, ulnar nerve (Kuiken et al., 2004).
Most subcutaneous fat was removed between the muscle and skin in the area so that electromyographic
electrodes later placed on the skin would obtain a better signal of the nerves activating the muscles. This also
helped the nerves reinnervate the skin (Kuiken et al., 2004).
Post surgery, the patient was not immediately able to control the pectoral muscle. The patient was
instructed to visualize and attempt to open and close the missing hand, and flex and extend the missing elbow
and wrist. Eventually the patient gained some control of the pectoral muscles. Flexing the elbow causes the
muscle under the clavicle to contract. Closing the hand causes the middle of the pectoralis major to contract.
Extending the elbow causes the lower pectoral muscle to contract. The test subject could never control the
pectoralis minor, meaning the anastomosis of the ulnar nerve was not successful (Kuiken et al., 2004).
In normal use, electromyographic sensors are placed on the skin above the muscles that were
reinnervated. The signals are received, amplified and interpreted by a control system that is interfaced to a
prosthetic. The control of the prosthetic seems natural, as the patient is mentally moving a portion of the
missing limb, and that portion of the prosthetic moves (Kuiken et al., 2004).
An added benefit of targeted reinnervation is that the afferent nerves in the anastomosed nerves
reinnervate the skin superficial to the target muscle. When areas of the skin above the pectoral muscles are
stimulated, the patient feels the sensation as if it is originating in the patient’s missing limb. This reinnervation
has been observed to happen through muscle, adipose tissue, and breast tissue (Kuiken, Marasco, Harden, &
Dewald, 2007).
To maximize electromyographic signals post-surgery, the subcutaneous fat was removed from his chest
above the pectoral area so that the skin touched the muscle beneath. The chest skin was not denervated, and no
neuropathy was observed prior to or after surgery. Five months post-surgery, a 15cm × 17cm area of the chest
skin had developed sensation. When stimulated, the sensation seemed to emanate from the missing limb. The
area was methodically stimulated by a cotton-tipped probe and mapped to areas of the missing limb, as
indicated by the patient. Stimulating single points on the chest mapped to large areas of the missing hand, and
were divided into palmar and dorsal side images as shown in Figure 2 (Kuiken et al., 2007).
Figure 2 Illustrating the sensations elicited by stimulation of the chest skin of patient BSD. Sensations were confined to the dorsal (green) or palmar (red) side. The targets are reference points for the diagram (Kuiken et al., 2007).
Patient STH also received targeted reinnervation surgery after losing her arm in an automobile accident.
To avoid disfiguring her, subcutaneous fat and breast tissue were not removed from STH’s chest. As illustrated
in Figure 1, STH’s brachial plexus nerve branches were anastomosed differently from BSD’s. The
supraclavicular cutaneous nerve was cut and attached to the side of the ulnar nerve to create a pathway to the
skin. By cutting the supraclavicular nerve, an 11cm × 9cm area of the chest was rendered numb. The cutaneous
nerve of the intercostal brachium was cut and attached to the side of the median nerve. Regeneration of
sensation did not appear to occur in this branch. Four months post-surgery, regeneration had occurred in the
ulnar-supraclavicular nerve pair and sensation returned to the previously neuropathic area of the chest (Kuiken
et al., 2007).
Mapping of STH’s sensate area proceeded as had been done with BSD. The stimulus points and
corresponding apparent sensation in the missing limb are shown in Figure 3.
Figure 3 Sensation mapping of STH's reinnervated chest area. The chest area was probed with a cotton-tipped probe with 300g of force. Upon stimulus, the patient indicated the area of apparent stimulation of the missing hand using her intact, contralateral limb (Kuiken, et al., 2007).
The sensations felt by patient STH were richer and more detailed than BSD’s. STH reported one area of
her chest that, when stimulated, conveyed proprioceptive information about her missing fourth finger. Upon
stimulation, the patient indicated that it felt like her fourth finger was being bent back. Other areas produced a
knife-edge sensation on the ulnar aspect of the missing hand. Another sensation associated with proprioception
is skin stretch. STH experienced this sensation as a pulling sensation on the second digit, and in another area as
a stretching sensation on the web between digits one and two (Kuiken et al., 2007).
Regeneration of connections to thermoreceptors was also observed in both patients. BSD’s threshold for
cold, warmth and heat pain were 24.3 ± 2.2 °C, 38.4 ± 0.6 °C, and 49.9 ± 0.1 °C on his unaffected side. On his
chest in the area that provided sensation in the missing limb, the thresholds were 28.8 ± 0.1 °C, 36.3 ± 0.2 °C,
and 45.4 ± 0.9 °C. STH’s results were similar. On her normal chest side, cold, warm and heat pain thresholds
were 30.3 ± 0.1 °C, 35.2 ± 0.1 °C, and 43.7 ± 1.2 °C. On the reinnervated side, the thresholds were 31.0 ±
0.2 °C, 32.9 ± 0.0 °C, and 40.1 ± 0.7 °C (Kuiken et al., 2007).
Electrical stimulation also evoked sensation in the reinnervated areas that were comparable in threshold
to the contralateral side. BSD was tested at five sites on the reinnervated side, and one site on the contralateral
side. The testing was done at sites that had previously been determined to provide sensation in the missing limb.
Non-painful sensation was observed at an average current of 3.7 ± 0.4 mA. The contralateral side’s threshold
was 2.7 ± 2.7 mA. STH was tested at four sites on her reinnervated chest, and one site on her contralateral side
and index finger. The average threshold was 4.9 ± 1.3 mA on the reinnervated side, 2.0 ± 0.0 mA on the
contralateral side, and 3.0 ± 0.0 mA on the index finger (Kuiken et al., 2007).
In both patients BSD and STH, there were some areas of overlap regarding sensation. Stimulating an
area of the chest caused the patient to feel the touch in both the chest and the missing hand. It is possible that the
brain will resolve this apparent conflict by shutting off referred sensation to the missing limb, or shutting off
sensation in the chest. At the time interval of the study (five years post surgery for BSD and 1.5 years post
surgery for STH), neither patient had experienced loss of either sensation (Kuiken et al., 2007). This indicates
targeted reinnervation could be a viable long-term solution for the restoration of functional haptic feedback.
Also apparent from this study is that nerve functionality in these two patients was not lost after
prolonged inactivity. Both patients received targeted reinnervation over a year after their amputations. The
regeneration shows that the nerves remained viable throughout this time span. One concern in patients with
nerves that have not been stimulated for a period of time is that healthy sections of the brain and spinal cord will
grow into the areas that have lacked stimulation. This can result in phantom limb pain when the healthy area is
stimulated (Kuiken et al., 2007).
After targeted reinnervation surgery, and the requisite recovery period, a patient may be fitted with a
prosthetic that incorporates a small load cell in the end effector. A haptic robot such as the G10 tactor from
Kinea Design, LLC shown in Figure 4 can be affixed in a non-permanent fashion to the patient such that the
tactor stimulates the reinnervated area (Marasco, Kim, Colgate, Peshkin, & Kuiken, 2011). It is important that
the tactor is matched in both modality as well as somatotopically. That is, if the load sensor is attached to the
index finger of the prosthetic, the tactor should be positioned over a reinnervated area that produces sensation in
the patient’s missing index finger. The stimulus should also be matched in modality. For instance if the sensor
on the prosthetic is a load cell that senses pressure, the tactor should press on the reinnervated area with force
proportional to that being exerted on the load cell (Kim, Colgate, Santos-Munné, Makhlin, & Peshkin, 2009).
Figure 4 Three tactor models from Kinea Design, LLC, Evanston, Il. USA (Kim, et al., 2009).
Experiments have been conducted on targeted reinnervation subjects that have been fitted with a tactor
and a prosthetic with a load cell using a variation of the Rubber Hand Illusion. In the classical version of the
Rubber Hand Illusion, the subject’s hand is hidden behind a screen, and a rubber hand is placed on a table in a
position that is near anatomically correct. A researcher simultaneously touches the rubber hand and the subject’s
hidden hand. Seeing the rubber hand being touched while feeling the sensation on the hand makes the subject
involuntarily think of the rubber hand as his or her own (Welch, 1971). In the experiment conducted with the
prosthetic patients, the researchers performed a series of tests involving variations on the Rubber Hand Illusion.
The researchers conducted tests using a vibratory transducer taped to the subject’s skin, or the aforementioned
G10 tactor to apply pressure. Subjects filled out a questionnaire relating their experienced sensation throughout
the tests.
According to the reseachers Marasco et al. (2011):
The results of this study provide evidence that a robotic touch interface, linking a prosthetic arm to the previously amputated cutaneous sensory nerves of a missing limb, can be used to elicit a shift in perception towards incorporation of the artificial limb into the self-image of two targeted reinnervation amputees. Taken collectively, these results suggest that providing physiologically and anatomically appropriate direct sensory feedback for a prosthetic limb creates a vivid sense of ownership of the device (p. 754).
Based on this study, it appears that the combination of electromyographic control of a prosthesis and
haptic feedback from the prosthesis would provide users with a limb that seems more like a part of them, rather
than a device attached to them. As the next step after haptic feedback, proprioceptive feedback would provide
users with a sense of the location of the limb without having to constantly monitor its position visually. This
would result in less mental effort to control the device. To accomplish feedback from prosthesis, there are
several techniques in use, and several hypothetical strategies. Techniques are as follows.
Direct stimulation of the brain is one means of providing sensory feedback from prostheses. In research
conducted by Davis, Kiss, Luo, Tasker, Lozano, and Dostrovsky (1998), areas of the thalamus responsible for
processing afferent stimulus in missing limbs was shown to remain functional. The research indicates that the
still-active areas may become susceptible to stimulation of other body areas, which is felt as phantom limb pain.
Studies that indicate the thalamus remains receptive to input for the missing limb are promising for
future research in brain-computer interfaces. Brain-computer interfaces connect electronics directly to the user’s
brain. Similar technology is used for deep-brain stimulation for relief of symptoms such as bradykinesia,
rigidity, and tremor related to Parkinson’s disease and other neurologic diseases (Perlmutter & Mink, 2006). By
using brain-computer interfaces, the thalamus could be stimulated to evoke sensation in the missing limb.
However, the aforementioned cross-stimulation from thalamic signals from disparate body areas could be a
confounding factor in that the other senses could override signals being applied through the brain-computer
interface. Since the thalamus is a deep structure in the brain, microstimulation of the surface would be less
invasive to implement.
Microstimulation is the stimulation of small areas of the brain. Microstimulation of S1, the primary
somatosensory cortex through multiple electrodes has been shown to provide sensory input to owl monkeys.
Each monkey was implanted with microwire arrays in the primary somatosensory (S1), primary motor (M1),
dorsal premotor (PMd), and posterior parietal (PP) cortices as shown in Figure 5. The microwire arrays were of
two types having either 2×8 or 4×8 steel wires. The wires were 50µm Teflon-coated steel. Dental acrylic was
used to fix connectors in the head cap. The monkeys were implanted four years prior to the S1 microstimulation
research, indicating long-term viability of implanting microwire arrays into the cortical areas (Fitzsimmons,
Drake, Hanson, Lebedev, & Nicolelis, 2007).
Figure 5 Sites where microwire arrays were implanted. (A) indicates that either two or three electrodes were stimulated for the time discrimination and side reversal tasks. (B) shows that four electrodes were used in a more complex spatiotemporal tasks. (Fitzsimmons, et al., 2007)
In the experiment, the stimulus applied to the S1 area simulated haptic sensation from the monkeys’
hands. The stimulus signals were 100-150 µA, the frequency was 100 Hz, and the duration was 0.1 ms. In
monkey one, apparent stimulation appeared to come from the palm of the right hand, as well as the glabrous
skin of the second and third digits of the hand. In monkey two, the stimulation appeared to come from the hand,
but was not precisely located. Prior to the microstimulation experiment, the monkeys had been trained to
perform a reaching task, and were prompted by vibrotactile feedback on their shoulder. The monkeys were able
to discern increasingly complex microstimulation patterns. This indicates more than simple on/off recognition
of the stimulus provided (Fitzsimmons et al., 2007). Due to the similarity of structure and function of simian
and human central nervous systems, it is possible this technique could be applied to human subjects.
Nerve regeneration through nerve guides with a multi-electrode array mounted within (See Figure 6) is
another promising technique for restoring sensory information to amputees. The technique involves sectioning a
nerve such as one of the afferent nerves found in the brachial plexus nerve bundle, then inserting the two ends
into a nerve guide housing the multi-electrode array. In as little as eight days, the nerves will begin regenerating
through the nerve guide, and will come into contact with the electrodes therein. As shown in the figure, the
electrodes are of different length and thus will contact different nerves when they regenerate through the nerve
guide (Garde, Keefer, Botterman, Galvan, & Romero, 2009).
Figure 6 Multi-electrode array shown mounted in a nerve guide (Garde, et al., 2009).
Similar to the nerve guide tube, a microelectrode sieve (See Figure 7) has been used to monitor nerve
activity from regenerated transectioned nerves. The sieve consists of a flat silicon substrate through which holes
have been micro-machined. Some of the holes were surrounded by electrodes of iridium attached to silicon
leads, which ended in a connector mounted to the subject’s (in this case fish) cranium. Nerves were observed to
regenerate around the implanted sieve, rather than through it (Mensinger, Anderson, Buchko, Johnson, Martin,
Tresco, Silver & Highstein, 1999).
To promote growth through the sieve, a coating of ProNectin L was added to the electrode heads.
ProNectin L has laminin functionality and is a silk-like protein that promotes nerve growth. The electrodes were
also fitted with a nerve guide tube made from poly-(acrylonitrile-vinyl chloride). Just before implantation, the
assembly received multiple coats of neural adhesive solution consisting of a 0.4 % solution of protamine sulfate
and poly-d, l-lysine (Mensinger et al., 1999).
Nerve activity through the sieve was only observed in certain cases, and was believed to be at those
electrodes that were close to nodes of Ranvier. The myelin sheath of nerves provided insulation at other sites
and prevented monitoring electrical activity (Mensinger et al., 1999). Demonstrating the ability to monitor
nerve signals with a microelectrode array should indicate the ability to inject signals into the afferent nerves that
have regenerated through the sieve. By extension, this indicates the possibility of providing sensory information
to the nervous system from prosthetics equipped with appropriate sensors.
Still at issue would be conveying those signals to the microelectrode array. It should be possible to use
inductive coupling to magnetically couple signals to an implanted device. The device could extract power
parasitically from the signal, as is done in RFID devices. This technique has been successfully demonstrated by
Sauer, Stanacevic, Cauwenberghs and Thakor (2005) to couple signals through water-bearing colloids similar to
human tissue.
Figure 7 (A) Scanning electron micrograph of the microelectrode. (B) 8 µm iridium lined electrode. (C) nerve guide tube inserted into an electrode. (D) nerve regeneration 60 days post implantation. Tissue fills greater than 50% of the pore. (E) neural activity 45 days post implantation. Bars indicate scale in µm: A=100, B=8, C=200, D=20 (Mensinger, 1999).
Other research with possible application to sensory feedback for prosthetics is that of implantable myoelectric sensors (IMES). IMESs are self-contained sensors that can be implanted subcutaneously and superficial to muscle that is being monitored. A diagram of the system is
shown in
Figure 8.
The IMESs receive power and receive and transmit data inductively through a coil built into the
prosthesis, and a coil built into each IMES. Each IMES has a unique 64-bit serial number and an 8-bit address
that allows discriminating the origin of signal data when multiple implants are placed within the residual limb.
These signals are subsequently processed by control electronics in the prosthetic and used to control the
functions of the prosthetic (Troyk, DeMichele, Kerns, & Weir, 2007).
Figure 8 IMES as used to control a prosthetic (Troyk, 2007)
One issue of concern with implantable devices such as the IMES is tissue damage in the event of
catastrophic failure. Failure in the IMES is unlikely, but could be caused by static discharge at the time of
implantation, or automated external defibrillator use after implantation. In order for a voltage to be induced
across the two electrodes of the IMES, a simultaneous failure of the PN junction of a p-channel field-effect
transistor, as well as an AC coupling capacitor, and a resistor would have to occur. Two sets of clamping diodes
are in place that will limit voltage output in the event of failure. The Very Large Clamp diodes shown in Figure
9 will limit voltage between the electrodes (and thus conveyed to tissue) to 500mV. 500mV is less than the
voltage required to dissociate water from the tissue, and therefore should be harmless (Troyk et al., 2007).
Figure 9 Tissue protection circuitry of the IMES (Troyk, 2007)
Rather than being used to monitor efferent nerve activity as amplified by skeletal muscle, the IMESs
could provide the framework for a device to send signals to afferent nerves. One possibility is combining either
the microelectrode array or the nerve guide tube with electrodes, both mentioned earlier, with a modified IMES
to stimulate one or more nerve endings of afferent nerves.
Interfacing electronics with afferent nerve endings is a challenge. Since the firing of nerves involves the
release of calcium ions, it should be possible to use any non-destructive method of calcium ion release to
stimulate nerves. Stanley, Gagner, Damanpour, Yoshida, Dordick, and Friedman (2012) have used a novel
method of stimulating insulin release in mice. The technique involves attaching antibodies for His that bind to a
modified TRPV1 channel that has the epitope tag TRPV1His to iron nanoparticles. TRPV1 forms a temperature-
sensitive calcium channel in the cell membrane. By applying a radio frequency field around the nanoparticle,
they can be made to heat up, thus causing the temperature sensitive calcium channel to open. In the experiment
by Stanley et al., this phenomenon was used to release insulin. By using iron nanoparticles of different sizes
attached to the antibodies, selective heating of cells can be achieved by using different frequencies for the RF
field.
Using this technique as a framework, it might be possible to design antibodies to epitopes found on the
surface of afferent neurons such as Synapsin I. The antibodies could be used to bind iron nanoparticles as
outlined be Stanley et al. (2012). Selective RF heating would allow the afferent neurons to be heated, possibly
triggering the release of calcium ions. The release of calcium ions could be seen as the nerve firing by more
distal neurons in the peripheral nervous system.
Another area of research that could help provide the feeling of a prosthetic limb as being “self” is
proprioceptive feedback. Proprioceptive feedback, mentioned only briefly herein, is exceedingly complex and
involves the interrelated senses provided by Golgi organs, muscle spindle fibers, joint angle receptors and
mechanoreceptors in the skin that are stimulated by stretch (Burke, Gandevia, & Macefield, 1988).
In experiments by Bark, Wheeler, Premakumar, and Cutkosky (2008), proprioceptive feedback by use of
a skin stretching device resulted in better positioning accuracy while positioning a virtual cursor than
vibrotactile feedback. This again indicates that feedback having the same modalitiy as the natural stimulus
produces superior integration into the body’s self-image and better control of the attached device. In this
experiment a virtual cursor, but results would indicate better performance in a prosthetic device could be
expected.
The loss of functionality due to the absence of a sense of touch for prosthetics has been documented as
early as 1918 by Silver. Of the techniques and theories presented herein, targeted reinnervation seems to be the
most practical. Although the surgery and reinnervation are invasive, further use of efferent nerves and the
minimally functional afferent nerves requires only external devices applied to the skin to sense or stimulate the
nerves. Other techniques such as the microelectrode array and the neuron tube have promise for increased
functionality in the future. Finally my hypothetical techniques of using modified IMESs and iron nanoparticles
hold promise and could influence my research into haptic and proprioceptive feedback in the future.
References
Atkins, D. J., Heard, D. C., & Donovan, W. H. (1996). Epidemiologic overview of individuals with upper-limb
loss and their reported research priorities. Journal of Prosthetics and Orthotics, 8(1). Retrieved June 29,
2012, from http://www.oandp.org/jpo/library/1996_01_002.asp
Burke, D., Gandevia, S., & Macefield, G. (1988). Responses to passive movement of receptors in joint, skin and
muscle of the human hand. Journal of Physiology, 402, 347-361.
Davis, K., Kiss, Z., Luo, L., Tasker, R., Lozano, A., & Dostrovsky, J. (1998). Phantom sensations generated by
thalamic microstimulation. Nature, 391, 385-387.
Fitzsimmons, N., Drake, W., Hanson, T., Lebedev, M., & Nicolelis, M. (2007). Primate reaching cued by
multichannel spationtemporal cortical microstimulation. The Journal of Neuroscience, 27, 5593-5602.
Garde, K., Keefer, E., Botterman, B., Galvan, P., & Romero, M. (2009). Early interfaced neural activity from
chronic amputated nerves. Frontiers in Neuroengineering, 2, 1-11.
Kim, K., Colgate, J., Santos-Munné, J., Makhlin, A., & Peshkin, M. (2009). On the design of miniature
haptic devices for upper extremity prosthetics. IEEE-ASME Transactions on Mechatronics, Submitted,
1-12.
Kuiken, T., Dumanian, G., Lipschutz, R., Miller, L., & Stubblefield, K. (2004). The use of targeted muscle rein
nervation for improved myoelectric prosthesis control in a bilateral shoulder disarticulation amputee.
Prosthetics and Orthotics International, 28(3), 245-253.
Kuiken, T., Marasco, P., Lock, B., Harden, R., & Dewald, J. (2007). Redirection of cutaneous sensation from
the hand to the chest skin of human amputees with targeted reinnervation. Proceedings of the National
Academy of Sciences, 104(50), 20061-20066.
Marasco, P., Kim, K., Colgate, J., Peshkin, M., & Kuiken, T. (2011). Robotic touch shifts perception of
embodiment to a prosthesis in targeted reinnervation amputees. Brain: a Journal Of Neurology, 134,
747-758.
Mensinger, A., Anderson, D., Buchko, C., Johnson, M., Martin, D., Tresco, P., et al. (1999). Chronic recording
of regenerating VIIIth nerve axons with a sieve electrode. Journal of Neurophysiology, 83, 611-615.
Patterson, P., & Katz, J. (1992). Design and evaluation of a sensory feedback system that provides grasping
pressure in a myoelectric hand. Journal Of Rehabilitation Research And Development, 29(1), 1-8.
Perlmutter, J., & Mink, J. (2006). Deep brain stimulation. The Annual Review of NeuroScience, 29, 229-257.
Sauer, C., Stanacevic, M., Cauwenberghs, G., & Thakor, N. (2005). Power harvesting and telemetry in CMOS
for implanted devices. IEEE: Transactions on Circuits and Systems, 52(12), 2605-2613.
Silver, D. (1918). The problem of the artificial arm. Journal of the American Medical Association, 71(3), 181-
183.
Stanley, S., Gagner, J., Damanpour, S., Yoshida, M., Dordick, J., & Friedman, J. (2012). Radio-wave heating of
iron oxide nanoparticles can regulate plasma glucose in mice. Science, 336, 604-608.
Troyk, P., DeMichele, G., Kerns, D., & Weir, R. (2007). IMES: An implantable myoelectric sensor.
Proceedings of the 29th Annual International Conference of the IEEE EMBS, FRA08.6, 1730-1733.
Welch, R. (1971). The effect of experienced limb identity upon adaptation to stimulated displacement of the
visual field. Perception & Psychophysics, 12(6), 453-456.
Westling, G., & Johansson, R. (1984). Factors influencing the force control during precision grip. Experimental
Brain Research, 53, 277-284.
van Beers, R. J., Sittig, A. C., & Denier van der Gon, J. J. (1996). How humans combine simultaneous
proprioceptive and visual position information. Experimental Brain Research, 111(2), 253-261.