An extensive review over current technology, possibilities and ethical implications in the area of neuronal implants.
Topics include:
- different forms of neuronal implants
- problems with current technology
- future possibilities
2. Introduction
The most incomprehensible thing about the world is that it is comprehensible.
Albert Einstein1
The central nervous system is virtually the most influential physiological system in a human
body. Several nerves contact every organ and every extremity. Almost every bodily function is
initiated and modulated by a nearby cluster of neurons. These accumulations are themselves
controlled by the central control system, the human brain.
About 1012 nerve cells2 account in this place for the majority of the nervous system. Every neu-
ron builds an activation potential and “fires” this signal along the main output line – its axon. This
process is repeated several times per second, resulting in a unique firing frequency. Neurons ac-
cept several types of input signals to change this base rate, mainly via several connection points on
the widely branched dendrites – the signal input pathways. An overview over the variety of neuro-
transmitters and signal transportation forms suggests that there are two different types of
interneuronal connections: excitatory and inhibitory synapses. Where excitatory signals are in-
creasing the firing rate of one neuron, activation from an inhibitory connection results in the
opposite effect.
During the existence of a neuron, several factors can threaten the proper function of signal
processing. Physical stress is the main cause that results in torn nerve pathways. In the case of a
severed dendrite, the injury disconnects the cell body from some of their input and control signals.
If the neuron is a part of a functional system, axon injury ceases the nerve cell from sending im-
pulses to peripheral organs or adjacent neuron clusters.
Because excitation of a neuron would be possible through an external stimulus, a damaged
neuron could theoretically be replaced - or at least mimicked - by an electrical stimulator. Howev-
er, because the targeted neuron has been embedded in a complex network of neurons, including a
variety of electrical and chemical connections, as well as feedback and control mechanisms, it
would be difficult for current technology to replicate the complete behavioral spectrum of a dam-
aged neuron. On the other hand, just because the effect of crude electrical stimulation tends to
propagate on a multitude of pathways, it might be possible to re-enable excitatory or inhibitory
impulses to stimulate healthy tissue areas again. Because of the “all-or-nothing” rule, it is some-
times possible to restore the functionality of a damaged neuronal network with just one active
contact point. It explains that the lack of a single from thousand input signals can mean the neu-
ronal threshold can no longer be reached. Action potentials are no longer formed, and as a result,
this particular neuron is seen as “silent” by their successors – a circumstance that disables the
whole network. If the missing input is replaced by a steady impulse, regular action potentials can
again be produced without significant disturbance from the artificial stimulation.
1
Vallentin A (1954) Einstein: A Biography, p. 24.
2
Poliakov GI (1972) Neuron Structure of the Brain. Harvard University Press, Cambridge, Mass
2 Introduction
3. Implementation
If you would understand anything, observe its beginning and its development.
Aristoteles
Hearing implants
In the beginning phases of neurobiological science in the late 1960, it still deemed impossible
to reactivate a lost cognitive function by applying electrodes to existing nervous tissue. Only two
decades after the first successful experiment, already 200’000 deaf people are using cochlear im-
plants, where 80-90% of them are able to recognize and understand daily speech3. The success of
these devices relies on the Volta effect, found by Alessandro Volta in 1800. It presumes that elec-
tric impulses stimulating auditory regions (inner ear or according brain areas) produce audible
noises without the necessity of actual sound waves. In the year 1930, this effect was rediscovered
and labeled as “electrophonic hearing”4. After successful experiments with a cats’ hearing nerve,
the two scientists concluded that the human auditory system was founded on electrical impulse
propagation. At this time, a telephone infrastructure had already been established, and the discov-
ery of the striking similarity between the cochlea and an electric microphone was a strong
motivational factor in the then little research on the field of neural audition. Further attempts
concentrated strongly on the transfer of sound information without the support from the middle
or even inner ear. By relying only on electric signal propagation, it would be easily possible to
bypass damaged mechanical structures that result in a permanent hearing loss.
These findings depict the foundation for the first auditory processor that split environmental
sound into its different frequencies and activated an array of 23 electrodes, which were mounted
directly on the auditory nerve, accordingly. The patient, treated by the French otologist and phy-
sicist Eyries in 1957, could hear sounds and understand speech fragments5, and the basic principle
is still used today in basic cochlear implants. After a period of skepticism - mainly due to the
patients not being able to understand speech - clinical trials in 1971 first showed positive results6.
Consequently, multichannel electrodes gained on popularity and commercial interest.
In 1979, and barely noticed first, the two surgeons William House and William Hitselberger
had a breakthrough in their clinic in Los Angeles. After having pushed an electrode directly into
the brainstem, electric impulses were able to stimulate the hearing nucleus directly7, despite the
fact that both auditory nerves were destroyed. This new hearing aid technology called the auditory
brainstem implant proved as reliable as the already established cochlear implant, and is still in use
today in case of severe nerve damage. Reliability and safety of both technologies significantly
improved when signals were transmitted over a magnetic coil in the mid 1980s, eliminating com-
mon problems such as plug failures or infections.
3
Illg A, von der Haar-Heise S, Goldring JE, Lenarz T (1999) Speech perception results for children implanted with the
CLARION cochlear implant at the Medical University of Hannover. Ann Otol Rhinol Laryngol suppl 177:93-98
4
Steven SS, Jones RC (1939) The mechanisms of hearing by electrical stimulation. J Acoust Soc Am 10:261-269
5
Djourno A, Ayries C (1957) Prothèse auditive par excitation électrique à distance du nerf sensorial à l’aide d’un bobi-
nage inclus à demeure. Presse Méd 35:1417-1423
6
Michelson RP (1971) The results of electric stimulation of the cochlea in human sensory deafness. Ann Otol Rhinol
Laryngol 80:914-919
7
Edgerton BJ, House WF, Hitselberger W (1982) Hearing by cochlear nucleus stimulation in humans. Ann Otol Rhinol
Laryngol suppl 91 (2 Pt 3):117-124
3 Implementation
4. Visual implants
Although the first experiments with visual perception implants started at the same time as
those for implanted hearing aids, this branch of applied neuroscience had to overcome greater
hurdles. The main reason for the lag in clinical application is the considerable higher amount of
active connections necessary to achieve an equal sensory experience: while a healthy auditory
nerve transfers impulses from about 30’000 receptor cells, the visual pathway accumulates infor-
mation from over 130’000’000 photoreceptors in the retina.
The early trials with direct stimulation of the visual cortex were not very successful. Partici-
pants experienced to the electrical stimulation as phosphenes - short-lived light spots – that
occurred in varying locations, independent from actual light sources in the environment8. Even
with considerable progress using modern multi-electrode arrays9, these implants not even provide
crude outlines of the environment, leaving patients unable to find visual cues relevant for orienta-
tion. The main reason for this effect is the nonconformity of the layout in the primary visual
cortex: it does not match the spatial location of retinal photoreceptors. Direct stimulation of the
optic nerve is today only possible in a very low spatial image resolution, mainly because the equiv-
alent electrodes have to be built in extremely small dimensions. Because of the structural layout of
the human eye, research about two implementations of visual implants has gained in popularity
since the 1990s: epiretinal and subretinal implants.
Epiretinal implants share their technological foundations with the auditory brainstem implant.
Environmental visual input is being recorded through a small camera, either external or integrated
in the eye lens10. The resulting signals then are converted by a microprocessor into neural frequen-
cies, either inhibiting or exciting the perceiving neurons on top of the retina. A clear limitation of
this system lies in the structural organization of the retina. Because light impulses normally would
pass a cascade of different neuron clusters, signals at the last stage (where the electrodes would
stimulate) do not represent a simple scheme of lightness and darkness any more, and can only
provide rough approximations to the natural scheme of activation levels and summation poten-
tials. According to recent findings, the according image after these early post-processing steps
would show extremely exaggerated edges. A common principle of color-coding is not elaborated
yet, which is why all artificial imagery is rendered in shades of grey.
Subretinal implants are able to bypass photoreceptors11 (which are actually located behind the
retina) by sending artificial impulses to the input layer of the intraretinal neural network. The
implant usually has the form of a microchip with one photosensitive and one output side, and has
to be placed in front of the layer with photoreceptors to ensure proper signal propagation. Two
major problems arose for the subretinal implant. Biodegradation always becomes relevant as a
silicon substrate is being enclosed in a biodynamic environment. The retina has shown to accept
the foreign body quite well, but the chip tends to take long-term damage from the direct contact
with living tissue. The other challenge is the missing power supply. Light signals carry to little
energy to power the signal processing and electric stimulation, and bodily energy sources cannot
be drained yet. Promising solutions for both problems have recently been developed. Energy
supply is currently guaranteed by using an external (and invisible) infrared light source, and bio-
logically stable envelopes ensure the long-term structural integrity of the microchip substrate12.
8
Brindley GS, Levin WS (1968) The Sensations produced by electrical stimulation of the visual cortex. J Physiol
196(2):479-93
9
Warren DJ, Normann RA (2003) Visual neuroprothesis. Handbook of neuroprosthetic methods. Finn WE, LoPresti PG
(eds) CRC Press, London, pp 261-307
10
Weiland JD, Liu W, Humayun MS (2005) Retinal prosthesis. Annu Rev Biomed Eng 7:361-401
11
Rizzo JF 3rd, Wyatt J, Humayun M, de Juan E, Liu W, Chow A, Eckmiller R, Zrenner E, Yagi T, Abrams G (2001) Retinal
prosthesis: an encouraging first decade major challenges ahead. Ophthalmology 108(1):13-14
12
Zrenner E (2002) Will retinal implants restore vision? Science 295(5557):1022-1025
4 Implementation
5. Brain-Computer Interfaces
Connecting an electronic device to the human brain not only to provide sensory input but also
to influence actual behavior is a practice that goes back to the 1950s. By installing electrodes in the
cerebral cortex of humans and animals, it was possible for José Delgado from Yale University to
evoke or attenuate a variety of reactions13. Surprisingly, not only motor responses but also primal
feelings of fear or aggression could be provoked by electric stimulation14. The researcher suggested
in 1952 that this practice could be useful with therapeutic treatment of psychotic patients, and
proposed in 1967 a device for remote controlling of emotional states15. Following the scheme of
external behavior control, another brain stimulation project accidentally showed the function of
the “reward center” in the hypothalamic region of the rat16.
The ways in which the neural networks are stimulated have gone a long way since the early
days of electrical brain stimulation. In contrast to the insertion of metal needles into the brain, it
is nowadays possible to view frequency patterns in real time and gently manipulate certain activa-
tion thresholds, without the device even touching the head. This technique which relies on
computer controlled, superconducting magnet coils is called deep brain stimulation.
Generally, every device that is able to measure neural activity as well as to influence it could be
called a human-computer interface. Typically, it consists of three modules17: Data acquisition
(electrodes collecting input from the brain), signal interpretation (determining the adequate
action from the incoming data) and output module (which executes the action).
Considering the first module, current devices can be placed in one of the two categories: inva-
sive and non-invasive. Non-invasive technology describes every implementation where the data
acquisition takes place outside the skull, i.e. through EEG measures. Because the placement of
electrodes is an easy task, this type of data acquisition allows a high amount of different test sub-
jects per period while the execution costs stay low. The biggest advantage turns out to pose a
challenge considering the quality of measurement. On the brain surface, several thousand clusters
of neurons provide different frequencies at the same time. Because of a skull thickness of 6 to 12
mm, this variety of signals becomes strongly attenuated and filtered by bone with irregular elec-
trical conduction properties. When searching for anomalies in form of a conscious impulse, brain
signals in this condition seem to show very high noise levels. Accounting for these factors, much
computing power and strong, repeated signals are needed to form a certain decision. In current
BCIs on non-invasive basis, it is possible for a trained subject to choose one letter from the alpha-
bet every two seconds on average, assuming constant full concentration on the task.
If a direct intervention should be necessary, as it is the case with epilepsy patients, the neuro-
surgeon places several small electrodes on the surface of the brain after opening the skull. This
form of data acquisition is called electrocorticography (ECoG), and the accuracy of the gained data
tops non-invasive methods by several magnitudes18. This data is used by the clinic for gaining pre-
surgical information about the condition of the specific epilepsy, and almost every invasive BCI
nowadays is a by-product of neurosurgical evaluation.
13
Delgado JM (1952) Responses evoked in waking cat by electrical stimulation of motor cortex. Am J Physiol 171(2):436-
446
14
Delgado JM, Rosvold HE, Looney E (1956) Evoking conditioned fear by electrical stimulation of subcortical structrues
in the monkey brain. J Comp Physiol Psychol 49(4):373-380
15
Delgado JM (1967) Aggression and defense under cerebral radio control. UCLA Forum Med Sci 7:171-193
16
Olds J (1956) A preliminary mapping of electrical reinforcing effects in the rat brain. J Comp Physiol Psychol 49(3):281-
285
17
Friehs GM, Zerris VA, Ojakangas CL, Fellows MR, Donoghue JP (2004) Brain-machine and brain-computer interfaces.
Stroke 35(11 suppl 1):2702-2705
18
Levine SP, Huggins JE, BeMent SL, Kushwaha RK, Schuh LA, Rohde MM, Passaro EA, Ross DA, Elisevich KV, Smith BJ
(2000) A direct brain interface based on event-related potentials. IEEE Trans Rehabil Eng 8(2):180-185
5 Implementation
6. Limitations
Never measure the height of a mountain, until you have reached the top.
Then you will see how low it was.
Dag Hammarskjold
The general disadvantage of implants in contrast to other methods of cognitive restoration or
even enhancement is obvious: this implementation always requires surgery. To understand why,
despite the great amount of professional expense, neural implants are still considered today as
veritable option for treatment, I will list the most important reasons in favor for implanted im-
plementations.
Unobtrusive: Can be placed under the skin or skull without limiting technical functions
Connectivity: Restoration of neural functions is possible even with missing essential elements
Continuity: Continuous function is ensured without any attention from the patient
Control: Desired effect can be varied or shut off immediately
Problems
Regarding current treatments, especially in the sector of epilepsy, these factors clearly put im-
plants in favor towards drug-based therapies. However, to put the advantages in relation, a closer
look towards the current limitations concerning neural implants will be necessary.
Despite cochlear implants being successfully improved over the last 40 years and neural vision
aids reaching the verge of clinical application, there is still no possibility to restore one of the
other senses. Moreover, even current devices can only deliver very poor sensory information quali-
ty. Cochlear audio quality is comparable to a mobile with bad reception. The image from a visual
implant is as blurry as under water, insensitive to low light and appears only in shades of grey.
This limitation of quality is mainly due to the small number of electrodes in an established con-
nection with functional tissue, and current research addresses this problem with larger electrode
arrays providing a much higher contact point density19.
Even if the connection between the signal processor and brain is approaching perfection, pa-
tients that are treated with this implant either have lost their sensory input in beforehand or were
even born without it. Neuroplasticity is a general process in the human brain, which tends to
strengthen heavily used connections and degenerate those that are scarcely active. Especially
when nerve fibers are severed due to physical stress, which is often the case with spinal cord inju-
ries, neural activation on the peripheral side of the nerve stops completely and causes rapid nerve
degeneration. Even with constant electrical stimulation or with the help of protective substances,
the complete connection cannot be assured because of a process called glial scarring. After the
first few months in the life of a brain, a connection between every possibly important brain region
is established via axons. After the nerve growth hormone wears off, the structural cells effectively
prevent growth of new nerve fibres with physical and chemical barriers. When a child is born, its
brain is in fact equipped with so many connections that most of them contradict each other. Neu-
roplasticity and constant exercise with the environment initiate the degeneration of unnecessary
axons, making it finally possible for the child to move both hands apart from each other. The
possibility for injured axons to regrow again, especially with old connections over long distances,
is low – decreasing the usefulness of an implant effectively.
Not only nerve connections, but also electrode matter suffers from degradation over time. Tis-
sue rejection may cause the creation of an isolating layer of scar cells, or electric stimulation could
even damage the target neurons through injudicious use. The current goal is thus to ensure the
biocompatibility of neural implants.
19
Fernandez E, Ferrandez J, Ammermuller J, Normann RA (2000) Population coding in spike trains of simultaneously
recorded retinal ganglion cells. Brain Res 887(1):222-229
6 Limitations
7. Finally, the electrode array in combination with the electrical stimulation software is only si-
mulating one aspect of a healthy neural network, which is shaped from the use of various
neurotransmitters and permanent learning effects through chemical and biological means. To
improve the performance and reliability, together with possible neural feedback, an implant
should probably connect to existing tissue in a more natural way. With the current toolset of
methods from electrical and biomechanical engineering, it is hard to imagine how a highly asso-
ciated electrochemical network of thousands of synapses could possibly be replaced by an artificial
device. Ideally, the electronic circuit would excite and inhibit all of the adjacent neurons in the
same pattern as the original neural network. To adjust for the different input conditions, “bioins-
pired” systems20 could translate certain incoming environmental signals into excitation patterns,
copying the known physiological task of the preprocessing neural network with arithmetic in-
structions. Especially in auditory and visual perception, this setup could dramatically improve the
sensory quality. A major drawback in this design involves the necessary electrode sizes to contact
all neurons on a certain surface. A density of several thousand contacts per cubic millimeter is
hardly imaginable in the context of nowadays cable-controlled arrays. Instead, it would be possible
to connect only to a small number of neuron groups on both input and output sides. A signal
processor would compare the input activity pattern with a limited set of data entries and translate
the arrangement into a predefined output pattern, ultimately simulating all major reactions of a
complex neural network. This is the level on which technological and clinical application operate
today. Several factors crucial for cognitive enhancement such as high reaction speeds, use of neu-
rotransmitters or only a system complexity density exceeding an existing neural network are still
not addressable by current technology. The very basic problems of neuroelectronic implants, such
as size, energy supply and biocompatibility seem to be solved by progress in material or computer
sciences during the next few decades. That leaves neuroscientists to concentrate on very specific
problems, and to try the most promising approaches on the way to an ingenious solution.
Approaches
Currently, the most limiting factor with a BCI is certainly interaction speed. It is possible to im-
prove the performance even without changing the hardware itself. Through creative interfaces
that try to predict the users’ actions and in combination with a high-performance 96-pin elec-
trode, it is possible to increase the letter selection speed to less than 1.2 seconds21.
Generally, researchers in the area of neural interfaces currently follow two approaches.
One possibility of improving the results is to strive for a deeper understanding of the role of
small neural networks in the signal processing chain. Recent research concerning the structural
composition of 117 neurons in the relay station of the visual processing pathway22 would be an
example for a successful pursuit of the first approach. This brain area is directly connected to the
optic nerve, and neurons are highly specialized against certain visual features. After mapping the
specific role of every neuron, researchers were able to reconstruct a natural image by “linear de-
coding” the preprocessed neural output. Similar breakthroughs recently occurred in the field of
voluntary movement23 and auditory perception24.
The second approach includes the knowledge about neurophysiological systems which have
already been studied intensively. Through a process called “neural morphing”, an attempt is made
to construct an electronic model of the cell network. Simplifications are made whenever neces-
sary; and because an engineer can build the model from scratch, even significant design
20
Fernandez E, Pelayo F, Romero S, Bongard M, Marin C, Alfaro A, Merabet L (2005) Development of a cortical visual
neuroprothesis for the blind: the relevance of neuroplasticity. J Neural Eng 2(4):R1-R12
21
Santhanam G, Ryu SI, Yu BM, Afshar A, Shenoy KV (2006) A high performance brain-computer interface. Nature
442(7099):95-198
22
Stanley GB, Li FF, Dan Y (1999) Reconstruction of natural scenes from ensemble responses in the lateral geniculate
nucleus. J Neurosci 19(18):8036-8042
23
Fogassi L, Luppino G (2005) Motor functions of the parietal lobe. Curr Opin Neurobiol 15(6):626-631
24
Ohl FW, Scheich H, Freemann WJ (2000) Topographic analysis of epidural pure-tone-evoked potentials in gerbil
auditory cortex. J Neurophysiol 83(5):3123-3132
7 Limitations
8. improvements are possible – such with as the artificial retina “Visio1” built at the University of
Pennsylvania25.
Enhancement
While treatment of patients with damages in sensory nerves has been the main motivation for
extensive research on the area of neural implants, the possibility to destroy nervous tissue at the
same has time been the main contradictor against the application in healthy individuals. However,
as soon as efficient devices for electrochemical stimulation become available, the successor of
therapy may indeed be enhancement. Currently, most sorts of cognitive enhancement are being
accomplished by careful drug use. Coffee is a prime example of a commonly accepted drug that
puts the user in control over his own wakefulness. In addition, off-label prescription drugs are
used by students and high performers to regulate their sleep cycles and remain focused for longer
periods. The definition of ‘enhancement’ is not fixed yet, neither socially or academically. From
the council of bioethics originates the statement that “’enhancement’, by contrast [to therapy], is
the directed use of biotechnical power to alter, by direct intervention, not disease processes but
the ‘normal’ workings of the human body and psyche, to augment or improve their native capaci-
ties and performances”. Most scientific authors separate between therapy and enhancement by the
actual context of implant use, where the improvement of condition counts as therapy goal setting.
Healthy persons with neural implants would thus fall into the category of enhancement. However,
this distinction only works as long as implants are not yet able to outperform their natural coun-
terparts. From this moment on, i.e. having “only” a natural hearing sense would classify for
therapy. To discuss the next section, I’d like to go along the lines of Eric Juengst26: “[The term
‘enhancement’ characterizes] interventions designed to improve human form or functioning
beyond what is necessary to sustain or restore good health”.
25
Zaghloul KA, Boahen K (2004) Optic nerve signals in a neuromorphic chip: Outer and inner retina models, testing and
results. IEEE Trans Biomed Eng 51(4):657-675
26
Juengst ET (1998) What does enhancement mean? In Parens E (ed) Enhancing Human Traits: Ethical and Social
Implications. Georgetown University Press, Washington, DC:29-47
8 Limitations
9. Ethical implications
There are no facts, only interpretations.
Friedrich Nietzsche
The role of modern medicine seems to have taken a relatively stable balance between pharma-
ceutical progress and social acceptance. Following the “Normal Function Model”, it seems that the
actual medical consensus consists of the goal to provide people with “normal” function to allow
“equal” opportunities for every member of humanity27. As both definitions of “normal” and
“health” are subject of ongoing debates, this fact should not be strongly relied on. Two main va-
riables have to be considered when predicting ethical implications on society: the risk-benefit
relation and the availability.
With current, socially accepted implants as with a pacemaker, there is no medical doubt that
the benefits of implantation outweigh the potential risks. However, with the increasing public
availability of future neural prostheses, two aspects are taken into closer consideration.
Firstly, interference with adjacent neuron clusters. As a rather simple device such as the coch-
lear implant is used to propagate signals into the main auditory nerve, no interference with other
cognitive functions would be expected - the implant only bypasses the middle and inner ear me-
chanisms. In case of a destroyed auditory nerve, the electrodes are already contacted directly with
brain matter – in this case, one small neuron cluster near the spinal cord. This area is a densely
populated zone, and because of two effects, this basic modification already results in a weak inter-
ference effect. The electric impulse from the electrode, meant to stimulate nearby neurons, has to
be one or two magnitudes higher than usual neural activation potentials because of the missing
synapse connection. Because the space between neurons is filled with liquid, this impulse is able
to propagate further than its intended working radius. In a certain distance, the voltage is not high
enough to provoke a direct reaction, but potential activation levels from adjacent neurons are
certainly increased during the impulse period. The second effect can be explained by the chaotic
connection layout between different neuron clusters. If two neurons are located in a small dis-
tance from each other, probability is high that at least one synapse connection exists – even if it is
scarcely used because both neurons belong to different task processes. Electric stimulation of this
one cluster will in this case cause these few connections to be activated, especially because the
base potential is already elevated. Indeed, reports from patients with auditory brainstem implants
support these findings with statements of “vibration” or “nausea” when their implants were acti-
vated the first time28. Adjacent nuclei to the hearing pathway would be the balance system and the
somatosensory system (sensation of vibration), amongst others. The higher the manipulated
cognitive function, the more careful one should examine potential cognitive interferences, because
an elevated base potential in higher regions would have significant and unpredictable effects on
essential details of the own self – memory, object recognition, or even personality.
Finally, neurotransmitters. Only because current stimulation occurs on an electric way, that
does not mean the effect on the affected neurons has to be purely electric in nature. Because
neurons are embedded in a network from electrical, biological and chemical reactions, it is rather
very unlikely that they would not react differently, when two of these components are missing. In
addition, the exact roles of the over 60 different neurotransmitters, not to speak of intermitting
glia cells, are largely unknown to current science. At this moment, is it thus not possible to deter-
mine potential long-term effects from artificial neural stimulation. Even the smallest of chemical
interventions could influence processes we do not understand yet – such as attention and arousal.
Before these issues are not resolved yet, both medical and social acceptability of neural im-
plants will be nonexistent, leaving it again to eager scientists to prepare the ground for further
discussion about a promising technology.
27
Daniels N (2000) Normal functioning and the treatment-enhancement distinction. Camb Q Healthc Ethics 9:309-322
28
Otto SR, Brackmann DE, Staller S, Menapace CM (1997) The multichannel auditory brainstem implant: 6-month
coinvestigator results. Adv Otol Rhinol Laryngol 52:1-7
9 Ethical implications