Sunteți pe pagina 1din 14

COCHLEAR IMPLANTS 885

coordination: basic remarks and experimental approach. J. Au- 29. T. Sekigami, S. Shimoda, K. Nishida, Y. Matsuo, S. Ichimori,
ton. Nerv. Syst. 1981; 3(24):335368. K. Ichinose, M. Shichiri, M. Sakakida, and E. Araki, Com-
9. M. C. K. Khoo, R. E. Kronauer, K. P. Strohl, and A. S. Slutsky, parison between closed-loop portal and peripheral venous in-
Factors inducing periodic breathing in humans: a general sulin delivery systems for an artificial endocrine pancreas. J.
model. J. Appl. Physiol. 1982; 53:644659. Artif. Organs 2004; 7(2):91100.
10. J. E. A. McIntosh and R. P. McIntosh, Mathematical Model- 30. A. Gentilini, C. Schaniel, M. Morari, C. Bieniok, R. Wymann,
ling and Computers in Endocrinology. Berlin: Springer Ver- and T. Schnider, A new paradigm for the closed-loop intraop-
lag, 1980. erative administration of analgesics in humans. IEEE Trans.
11. K. V. Baev, K. A. Greene, F. F. Marciano, J. E. Samanta, A. G. Biomed. Eng. 2002; 49(4):289299.
Shetter, K. A. Smith, M. A. Stacy, and R. F. Spetzler, Phys- 31. J. Dolensek, F. Runovc, and M. Kordas, Simulation of pulmo-
iology and pathophysiology of cortico-basal ganglia-thalamo- nary ventilation and its control by negative feedback. Com-
cortical loops: theoretical and practical aspects. Prog. put. Biol. Med. 2005; 35(3):217228.
Neuropsychopharmacol. Biol. Psychiatry 2002; 26(4):771 32. R. D. Branson, J. A. Johannigman, R. S. Campbell, and K.
804. Davis Jr., Closed-loop mechanical ventilation. Respir. Care
12. E. R. Kandel, J. H. Schwartz, and T. M. Jessel, Principles of 2002; 47(4):427451.
Neural Science, 4th ed. New York: McGraw Hill, 2000. 33. S. Jezernik, R. G. Wassink, and T. Keller, Sliding mode closed-
13. L. Ljung, System Identification: Theory for the User. Engle- loop control of FES: controlling the shank movement. IEEE
wood Cliffs, NJ: Prentice-Hall, 1999. Trans. Biomed. Eng. 2004; 51(2):263272.
14. C. W. J. Granger, Economic process involving feedback. In- 34. K. O. Johnson, D. Popovic, R. R. Riso, M. Koris, C. Van Doren,
formation and Control, Vol. 6. 1963, pp. 2848. and C. Kantor, Perspectives on the role of afferent signals in
15. P. E. Wellstead, Non-parametric methods of system identifi- control of motor neuroprostheses. Med. Eng. Phys. 1995;
cation. Automatica 1981; 17: 5569. 17(7):481496.
16. H. Akaike, Selected Papers of Hirotugo Akaike. Berlin: 35. R. Allen, D. Smith, Neuro-fuzzy closed-loop control of depth of
anaesthesis. Artif. Intell. Med. 2001; 21(13):185191.
Springer-Verlag, 1998.
36. N. Wiener, Cybernetics, or Control and Communication in the
17. J. Rissanen, Modelling by shortest data description. Automa-
tica 1978; 14:465471. Animal and the Machine. Cambridge, MA: The Technology
Press and New York: Wiley, 1948.
18. P. Van Overschee and B. DeMoor, Subspace Identification
of Linear Systems: Theory, Implementation, Applications. 37. R. A. Hess, Human in the loop control. In: W. S. Levine, ed.,
The Control Handbook. New York: CRC Press, 1996,
Dordrecht, The Netherlands: Kluwer Academic Publisher,
1996. pp. 14971505.
19. U. Forssell and L. Ljung, Closed-loop identification revisited. 38. R. Sutton, Modelling Human Operators in Control System
Automatica 1999; 35:12151241. Design. New York: Wiley, 1990.
20. P. M. J. Van den Hof and R. J. P. Schrama, An indirect method
for transfer function estimation from closed loop data. Au-
tomatica 1993; 29:15231527.
21. U. Forssell and L. Ljung, A projection method for closed-loop COCHLEAR IMPLANTS
identification. IEEE Trans. Auto. Control 2000; 45:2101
2106. HUGH J. MCDERMOTT
22. G. Baselli, S. Cerutti, S. Civardi, A. Malliani, and M. Pagani, The University of Melbourne
Cardiovascular variability signals: towards the identification East Melbourne, Australia
of a closed-loop model of the neural control mechanisms.
IEEE Trans. Biomed. Eng. 1988; 35(12):10331046.
23. M. C. Khoo, F. Yang, J. J. Shin, and P. R. Westbrook, Estima-
tion of dynamic chemoresponsiveness in wakefulness and 1. HEARING AND DEAFNESS
non-rapid-eye-movement sleep. J. Appl. Physiol. 1995;
78(3):10521064.
Figure 1 is a sectional diagram of the human auditory
24. B. Widrow, Adaptive model control applied to real-time blood anatomy. (The figure includes components of a typical co-
pressure regulation. Pattern Recognition Machine Learning
chlear implant system, which will be described later.) The
1971; 310.
pinna, at the left, directs airborne sound via the external
25. C. W. Frei, E. Bullinger, A. Gentilini, A. H. Glattfelder, T. Si-
ear canal to the tympanic membrane, or eardrum. Sound
eber, and A. M. Zbinden, Artifact-tolerant controllers for au-
pressure waves produce vibration of the eardrum, which is
tomatic drug delivery in anesthesia. Crit. Rev. Biomed. Eng.
2000; 28(12):187192. coupled to the cochlea, or inner ear, by means of three ar-
ticulated bones, known as ossicles. The cochlea is a spiral-
26. E. A. Woodruff, J. F. Martin, and M. Omens, A model for the
design and evaluation of algorithms for closed-loop cardio- shaped structure containing fluid in several chambers
vascular therapy. IEEE Trans. Biomed. Eng. 1997; 44(8): separated by membranes. Acoustic vibrations generate
694705. oscillatory motion of structures inside the cochlea. The
27. S. Vozeh and J. L. Steimer, Feedback control methods for drug physical properties of the internal cochlear structures re-
dosage optimisation. Concepts, classification and clinical ap- sult in a mechanical filtering characteristic that distrib-
plication. Clin. Pharmacokinet. 1985; 10(6):457476. utes this motion spatially in accordance with the
28. F. Cantraine, Computer driven i.v. injection systems. State of frequency content of incoming acoustic signals. High-
the art, future developments. Acta Anaesthesiol. Belg. 1988; frequency sounds produce most movement near the base
39(4):257266. of the cochlea, where vibrations are introduced from the
886 COCHLEAR IMPLANTS

modulate neural activity in the auditory cortex of the


brain, where hearing sensations are perceived.
In sensorineural hearing impairment, a partial or total
loss of hair cells usually occurs. Damage to, or loss of, hair
cells causes a loss of hearing sensitivity because the me-
chanical vibrations created in the cochlea by sound waves
are not represented in neural excitation patterns. The hair
cells tend to be more susceptible to permanent damage
from various causes than most other structures in the ear
that are essential for hearing. Some common causes of
hair-cell damage include prolonged exposure to excessive
levels of noise, diseases such as meningitis and otosclero-
sis, head trauma, and treatment with certain antibiotic
drugs. Once destroyed, hair cells in humans do not regen-
erate (although hair-cell regeneration is observed to occur
in some other species). However, many of the primary au-
ditory neurons survive even in cases of total hair-cell loss.
Therefore, it is possible to create hearing sensations even
in an ear that is profoundly or totally deaf as a conse-
quence of extensive loss of hair cells by stimulating audi-
tory neurons directly with electric currents. The purpose
Figure 1. Cutaway diagram showing a right human ear and a of a cochlear implant (CI) is to generate electric stimuli
cochlear implant. Behind the pinna (external ear) at the left is the
that are appropriately controlled to convey useful infor-
sound processor. The transmitting coil, above the pinna, is
mation about sounds, and that do not cause major unde-
aligned with the receiving coil of the device implanted under
the skin. The electrode array lies inside the spiral-shaped cochlea sirable side effects even with chronic application.
shown at the lower right. A separate lead connects the receiver-
stimulator to an extracochlear electrode that has been placed im-
mediately above the receiving coil. Also shown are the external 2. DESIGN OF COCHLEAR IMPLANT SYSTEMS
ear canal leading from the pinna to the eardrum, and the ossicles
that couple sound vibrations from the eardrum to the inner ear. Although many different designs of CIs have been de-
(This figure is available in full color at http://www.mrw.intersci- scribed over at least the past three decades, all designs
ence.wiley.com/ebe.) have general features in common. The components of a
typical CI system are shown in Fig. 2. All implant systems
pick up sound signals with a microphone, which is usually
ossicular chain, whereas lower frequencies produce max- packaged in a small enclosure that is worn on the users
imal activity at places closer to the cochlear apex. pinna, as with conventional hearing aids of the behind-
The motion of the structures inside the cochlea is trans- the-ear type. An electric signal representing the variation
duced into neural signals by approximately 15,000 hair of air pressure associated with airborne sound waves is
cells that are distributed regularly around the spiral. conveyed from the microphone to an electronic signal pro-
Sound vibrations apply shearing forces to projections, cessor. The processor is designed to convert selected fea-
known as stereocilia, on the hair cells. Deflections of the tures of acoustic signals into a pattern of electric nerve
stereocilia initiate a sequence of electrochemical events stimuli that will provide useful hearing sensations to the
that result in the production of action potentials in audi- implant user. Considerable flexibility is available to de-
tory neurons. These action potentials are conveyed from signers of sound processing circuits and algorithms, which
the cochlea by fibers of the auditory nerve to neural pro- has led to the development and evaluation of a number of
cessing centers in the brainstem. Ultimately, these signals distinct processing schemes. Functional details of the

User-specific
stimulus parameters

Microphone
Signal Stimulus
Estimation of
amplification and parameter encoding
signal spectrum
conditioning and transmission

Data decoding
Stimulus
and power
production
generation
Electrode array Inductive
Figure 2. The main functional blocks of a
typical cochlear implant system. link
COCHLEAR IMPLANTS 887

most important schemes that are currently in use are pre- other than the sound processor, which is one advantage
sented later. of implanting a relatively complex receiver-stimulator
In most cases, the output of the sound processor con- rather than connecting the signal received from the im-
sists of a digital code specifying the parameters of the planted coil directly to the stimulating electrodes. How-
electric stimuli to be delivered to the implanted electrode ever, the latter technique has been employed in some
array. The code is usually conveyed to the implanted de- simple CI systems, and some of its characteristics are dis-
vice via an inductive link (see Fig. 2). The link operates on cussed further later.
the same principle as a transformer, and comprises two In many recent implant systems, the receiver-stimula-
coils of wire separated by the skin overlying the implant. tor has the additional capability of transmitting informa-
An electric current flowing in the external coil magneti- tion outward for external analysis and monitoring. Such
cally induces a secondary current in the implanted coil. information might include measurements of the electrical
The currents are modulated by the transmitter circuit of impedances of the electrodes, or an indication of the state
the sound processor to represent the encoded information of the internal power supply. This information can be used
specifying the required electric stimuli. An integrated cir- to diagnose faults in the implant, or to minimize the
cuit in the implant, often called the receiver-stimulator, amount of power that the sound processor needs to trans-
demodulates the signal obtained from the subcutaneous mit for a given user. The power requirements depend on
coil. In addition, the inductive link serves to provide elec- various individual factors including the thickness of the
tric power to the implanted electronics and hence to the skin separating the transmitting coil from the receiving
electrodes. The power is extracted by rectifying and coil, which affects the coupling efficiency of the inductive
smoothing the modulated alternating current induced in link. When measurements of the electrical state of the
the implanted coil. This technique obviates the need for a implanted components are available externally, it is pos-
power source, such as a battery, in the implant. The ab- sible for the transmitter of the sound processor to be ad-
sence of an implanted battery helps to ensure that, after justed automatically to provide the minimum required
the device has been implanted, no further surgical inter- power, and thus to maximize the lifetime of the battery.
vention should be necessary throughout the lifetime of the The output of the receiver-stimulator is usually a pre-
implant user. cisely controlled electric current that is directed to one or
more of the implanted electrodes. In the majority of exist-
ing implant systems, the stimulus waveform is usually a
2.1. Implanted Components symmetric, biphasic, rectangular pulse (Fig. 4, top). This
A typical implantable device, including a receiver-stimu- waveform comprises two intervals, or phases, of equal du-
lator and electrode array, is illustrated in Fig. 3. The re- ration, during each of which current flows through the
ceiver-stimulator decodes the information transmitted by selected electrodes. The phases may be separated by a
the external sound processor, which specifies parameters short time during which no current flows. The insertion of
such as the onset time and magnitude of the stimulus to be such an inter-phase gap can reduce the current required
generated, and the particular electrodes that are to con- to produce an audible sensation, at the expense of a small
duct the stimulus current. To prevent unintended stimu- increase in the overall stimulus duration. Current flows in
lation, the receiver-stimulator contains circuits to check opposite directions during the two phases, but with equal
that the received code is free from errors, and that ade- magnitudes. Therefore, the total charge transferred
quate power is available from the inductive link to gener-
ate the required stimulus. Thus, little chance exists that
the implant will be activated incorrectly or unexpectedly CURRENT
as a result of poor coupling between the coils, or the re-
ception of electromagnetic fields generated by devices
0

CURRENT

TIME
Figure 4. Two types of stimulating waveform generated by co-
chlear implants. At the top is a series of two biphasic current
Figure 3. Photograph of a cochlear implant, showing the re- pulses. Each pulse comprises two intervals, or phases, during
ceiving coil with magnet at center, the electronics package, the which a constant current flows between the active electrodes.
intracochlear electrode array, and the separate extracochlear These phases are separated by a brief zero-current interval. The
electrode. (This figure is available in full color at http://www.mrw. direction of current flow is reversed between the two phases. At
interscience.wiley.com/ebe.) the bottom is an example of an analog stimulus.
888 COCHLEAR IMPLANTS

through the electrodes during the first phase (i.e., the instance, and the effects of many of these factors cannot
product of current and phase duration) is equal to the be predicted accurately before the device is implanted, re-
charge transferred in the reverse direction during the sec- ceiver-stimulators must be designed with enough spare
ond phase. This cancellation of the charge delivered by the capacity to function correctly in any potential user. In
electrodes within each stimulating period has been shown most CI users, the range over which currents (or pulse
to minimize damage to neurons and other tissues in the widths) vary on each electrode is typically about 520 dB.
vicinity of the electrodes, and to prevent corrosion of the In the design of receiver-stimulators, the electrical im-
electrode materials. Two widely used methods of ensuring pedance of the electrodes is an important consideration.
that the charge is balanced accurately between the phases The impedance of a pair of metal electrodes immersed in a
are to couple the electrodes to the stimulator via capaci- conductive fluid can be represented approximately by the
tors, which prevent accumulation of charge on the elec- equivalent circuit shown in Fig. 5. The circuit includes
trodes as a consequence of any nonzero net current flow, capacitive and resistive elements whose values depend on
and to connect the electrodes together whenever the stim- various parameters, such as the surface area and material
ulator is not delivering current (i.e., at all times between of the electrodes and the electrical properties of the fluid.
successive pulses). The latter technique helps to dissipate It is difficult to predict the precise impedance of electrodes
any residual charge remaining in the electrode circuit af- after implantation, and it is also known that the imped-
ter the completion of a biphasic pulse, and is often applied ance can vary over time. For stimulation with constant
in implanted stimulators regardless of whether coupling currents, changes in the electrode impedance cause corre-
capacitors are also used. However, some stimulators that sponding changes in the voltage. Fortunately, neurons are
are designed to drive a large number of electrodes may not much more sensitive to the local current intensity than to
incorporate coupling capacitors, because such components the voltage at the electrodes, so moderate impedance
tend to be bulky and would require a relatively large im- changes generally have little effect on the sensations per-
plant package. ceived by the implant user. However, the maximum cur-
The safety of the electric stimulus for chronic applica- rent and pulse width that can be applied on a particular
tion is also promoted by using the smallest possible cur- electrode are affected by that electrodes impedance. As
rents and pulse durations. Both the current and the pulse can be inferred from the circuit of Fig. 5, the onset of a
duration (as well as other parameters of the stimulation) current pulse (see Fig. 4, top) produces an abrupt increase
affect the loudness of electrically induced sensations. Gen- in the voltage across the electrode pair. The voltage
erally, increases in either of these parameters increase the increases smoothly during the first phase of the pulse
number of auditory neurons that are activated or the rate while the capacitance is charged, and reaches a maximum
at which neural action potentials are produced. At least at the end of the first phase. For the current generator in
two additional ways of increasing the susceptibility of the receiver-stimulator to maintain the intended constant
neurons to the stimulation exist. First, the electrodes current throughout the pulse, it is essential that this
can be placed closer to the surviving auditory neurons.
Second, the configuration of the electrodes can be selected
to increase the spatial distribution of the stimulus cur-
rents and thereby to excite a larger number of neurons.
These aspects of electrode design are discussed further
below.
In general, however, the current levels that can be gen-
erated by existing implants range from a few microamps E1
up to a few milliamps, and the pulse durations range from
tens to hundreds of microseconds per phase. The inter-
phase gap (see Fig. 4, top) is usually less than 50 ms long. from
In some receiver-stimulator designs, the stimulus cur- receiver-
rents are related by a nonlinear function to the digital stimulator
code that is transmitted from the sound processor. For ex-
ample, each step in the current-level code might result in a
proportional (rather than absolute) change in the magni-
E2
tude of the electrode current. This design derived from an
early finding that the loudness of sensations produced by
an implant was related to the current magnitude approx-
imately by a mathematical power function; thus, loudness
can be controlled more regularly when currents are chan-
ged by a constant ratio rather than in fixed absolute steps.
For a particular electrode in a given implant user, the
Figure 5. Simplified equivalent circuit of an intracochlear elec-
range of currents (or pulse durations) that encompasses trode pair. The nodes labeled E1 and E2 at the left are driven by
the widest useful range of loudness is always much the current generators of the stimulator. Each parallel R/C com-
smaller than the full range that the receiver-stimulator bination represents the metal-fluid interface for each electrode,
can generate. As numerous individual factors affect the and the resistance on the right represents the current path within
range of currents that may be required in a specific the cochlear tissues.
COCHLEAR IMPLANTS 889

maximum voltage be less than a limit known as the com- normal cochlea, an orderly relationship exists between the
pliance voltage, which is slightly less than the power sup- frequencies of sounds and the location of maximal excita-
ply voltage of the receiver-stimulator. The supply voltage tion of auditory neurons. The essence of this relationship
varies among implant devices, and can also be affected by is that high frequencies produce most activity in neurons
factors such as the thickness of the skin separating the that innervate hair cells near the base of the cochlea,
sound processors transmitting coil from the receiving coil, whereas lower frequencies activate neurons that inner-
but is usually around 12 V. vate hair cells located at more apical positions. This tono-
It is possible with some recent implant systems to topic organization applies not only to hair cells, but also to
check postoperatively that all electrodes are operating the cell bodies and dendrites of auditory neurons. There-
within their limits of compliance. If it is found that some fore, even in cases of profound sensorineural deafness, in
stimuli are causing compliance to be exceeded, the prob- which there may be few or no surviving hair cells, it is still
lem can be eliminated in most cases by reducing the stim- possible for a CI to take advantage of the tonotopic orga-
ulus current, which will usually require an increase in nization of auditory neurons with a suitably designed ar-
the pulse width or a change of the electrode configuration ray of electrodes. Generally such arrays comprise many
(e.g., from bipolar to monopolar). Characteristics of these discrete electrodes mounted on a carrier that can be in-
and other types of electrode configurations are described serted into the fluid-filled interior of the cochlea through
below. the round window, or through an artificial opening near
A final requirement for the safe long-term use of coch- the round window (see Fig. 1). When an array is inserted
lear implants is that the materials from which the devices deeply into the cochlea in this way, electrodes near the tip
are constructed be biocompatible and structurally and preferentially stimulate neurons that would normally
chemically stable. Studies investigating the properties of have responded best to low-frequency sounds, whereas
candidate materials in the biological environment have electrodes nearer the cochlear base stimulate neurons
identified a number of suitable metals, ceramics, and that would normally have been responsive to higher-fre-
polymers. Electrodes and connecting wires are usually quency sounds.
fabricated from platinum or an alloy having similar char- For this principle to be effective in practice, it is impor-
acteristics. Electrode carriers and the protective jacket tant that the current delivered by each electrode can stim-
surrounding the implanted package are made from med- ulate exclusively a nearby subpopulation of the available
ical-grade silicone rubber compounds. The package itself neurons. If excessive spatial overlap exists among the
is often constructed using titanium or ceramic materials groups of neurons that respond best to each member of a
and is hermetically sealed to prevent fluids contacting set of closely spaced electrodes, then it is unlikely that the
the electronic components inside. The use of appropriate auditory percepts produced by activation of each electrode
materials, combined with the intensive testing of implan- will be distinct. Such an electrode array would provide
table devices during and after manufacture, have gener- little advantage over an array having fewer electrodes,
ally succeeded in maintaining postoperative failure rates particularly from a perceptual point of view. The question
at acceptably low levels. of how many electrodes should be provided in a CI is a
complex one, and some of the issues involved are beyond
the scope of this article. However, if benefit is to be ob-
2.2. Stimulating Electrodes
tained from implanting many, rather than few, electrodes,
In early CIs, the primary objective was to deliver currents it is essential that the spatial distribution of the current
to the entire surviving population of auditory neurons in delivered by each electrode be restricted to distinct groups
people whose degree of deafness was so severe that of the available neurons, which is the rationale for the
they could not perceive sound no matter how much am- trend in recent CI developments for electrode arrays to be
plification was attempted using conventional hearing positioned closer to the modiolus at the center of the co-
aids. Often a single electrode was placed near to the neu- chlea, and therefore closer to the cell bodies of the auditory
ral population, with the electric circuit being completed neurons.
through a second electrode that was located at a remote Regardless of the location of the electrodes, different
site. In one device, which was implanted in a relatively ways exist in which multiple electrodes can be configured
large number of people mainly in the 1970s and 1980s, the to deliver stimulating currents to the auditory neurons.
single active electrode was inserted into the basal region The three main configurations available with existing de-
of the cochlea through an opening created surgically at or vices are illustrated in Fig. 6. The monopolar electrode
near the round window (1). An upgraded version of this configuration comprises an active electrode that is located
device, manufactured by AllHear Inc. (2), is the only de- close to the target auditory neurons and a second electrode
vice presently available from a commercial manufacturer (sometimes referred to as the indifferent or ground
that does not have the capability of stimulating on multi- electrode) that is located relatively far from the neurons
ple electrodes. (Fig. 6, left). The indifferent electrode typically has a
Although the use of a single active electrode may be larger surface area than the active electrode, and may
attractive, mainly because of the relative simplicity of serve as the current return path for many discrete active
both the surgery and the stimulator electronics, great electrodes. In multiple-electrode implants employing
advantages exist in being able to stimulate different sec- monopolar stimulation, it is important that the active
tors of the population of surviving auditory neurons with electrodes be located close to the target neurons so that
some degree of independence. As mentioned earlier, in the each electrode may stimulate a perceptually distinct
890 COCHLEAR IMPLANTS

Unfortunately, such short circuits occur occasionally, and


can occur intermittently as well as permanently. Their oc-
currence is associated with the need, in multielectrode
arrays, for many, very thin wires to be packed closely to-
Monopolar Bipolar Common ground gether in passing between each active electrode and the
implanted receiver-stimulator. When short circuits occur
Figure 6. Three electrical configurations of the stimulating elec-
with a bipolar electrode configuration, the threshold cur-
trodes: monopolar (left), in which current flows between an in-
rents can be much lower than expected, because, in effect,
tracochlear electrode and a remote electrode; bipolar (center), in
which current flows between two intracochlear electrodes; and the spatial extent of the electrode pair is larger than in-
common ground (right), in which current flows between one in- tended. Use of the common ground configuration circum-
tracochlear electrode and all other electrodes. vents this problem, and therefore it is often selected in
situations where there are known short-circuits between
some electrodes, or where an intermittent short circuit is
subpopulation of neurons. The electric current distribu- suspected.
tions associated with each electrode become more over- A variation on the common ground electrode configu-
lapping at increasing distances from the active electrodes, ration is used in the COMBI 40 implant, manufactured
and would be completely indistinguishable at locations by the company MED-EL (4). The standard intracochlear
close to the indifferent electrode. electrode array comprises 12 pairs of electrodes. However,
To improve the spatial separation of stimulating cur- rather than employing the bipolar configuration, the
rents in multiple-electrode implants, bipolar stimulation MED-EL device treats 12 of the electrodes (one from
is often used. In the bipolar configuration, currents are each pair) as a common current return path for each of
passed between two electrodes, both of which are located the remaining 12 electrodes. Like other implant manufac-
relatively close to the auditory neurons (Fig. 6, center). turers, MED-EL provides modified electrode arrays for
The decrease in current density with distance from a bi- special situations, such as cases where surgical access to
polar electrode pair is much steeper than the decrease as- the cochlea is restricted.
sociated with an electrode used in the monopolar In some CI designs, the way in which electrodes are
configuration. Theoretically, therefore, the spatial selec- configured to deliver stimulation can be selected at any
tivity of bipolar electrodes should be better than that of time by transmitting appropriate data to the receiver-
monopolar electrodes. However, psychophysical studies stimulator. For example, in the CI24 series of devices,
have shown that the perceptual distinctiveness of elec- manufactured by Cochlear Limited (5), there are 22 in-
trodes is not related exclusively to whether bipolar or tracochlear electrodes mounted longitudinally on a flexi-
monopolar stimulation is applied (3). ble carrier that is inserted into the scala tympani of the
Several variations exist on the bipolar electrode config- cochlea, and two additional electrodes that are located
uration that can provide practical benefits in some cir- outside the cochlea (see Fig. 3). The 22 intracochlear elec-
cumstances. In one variation, the separation between the trodes can be configured in a common ground or bipolar
two active electrodes (called the spatial extent) can be arrangement without the use of the extracochlear elec-
increased, which results in a reduction of the stimulus trodes. In the bipolar configuration, any two electrodes
current required to produce an audible sensation because can be selected to form a stimulating pair, so that all spa-
more auditory neurons are in the path of the current. (For tial extents possible within the limits imposed by the ge-
the same reason, threshold currents are generally lower ometry of the electrode array are available. In addition,
with monopolar stimulation than with bipolar stimula- monopolar configurations can be selected using any of the
tion.) Another variation on the bipolar configuration in- intracochlear electrodes in combination with either or
volves mounting the electrodes on their carrier in an both of the extracochlear electrodes. In principle, a differ-
arrangement that directs the current along approximately ent configuration could be selected for each pulse of stim-
a radial, rather than longitudinal, path. Currents that are ulation to be generated, although in practice the use of the
aligned radially with respect to the cochlear spiral are same configuration for all active electrodes is more usual.
more efficient at producing action potentials in nerve cells
than currents flowing in other directions, and thus radial
2.3. Stimulating Waveforms
bipolar electrode pairs should, at least in theory, provide
better selectivity and lower thresholds than alternative Previously, the constant-current, biphasic pulse was de-
electrode configurations. scribed briefly, because it is the most widely used stimu-
A third type of electrode configuration is known as the lating waveform in existing CI devices. As illustrated in
common ground stimulation mode (Fig. 6, right). In this Fig. 4 (top), this waveform can be defined completely using
configuration, one active electrode is selected, and all of only four parameters: the current level, the pulse onset
the remaining intracochlear electrodes are used together time, the duration of the two phases, and the time interval
as the return path for the stimulating current. In some separating the phases. To generate each pulse of stimula-
functional respects, the common ground arrangement is tion, data specifying these parameters, and the particular
intermediate between the bipolar and monopolar configu- electrodes to be activated, must be transmitted to the im-
rations. One advantage of the common ground config- planted receiver-stimulator. In part because this small
uration is that its use can minimize the adverse effects number of data parameters can be represented in digital
of electric short circuits within the electrode array. form relatively compactly, the complexity and the electrical
COCHLEAR IMPLANTS 891

power consumption of both the external sound processor However, the large physical volume required for both the
and the implant can be kept to manageable levels, even external transmitting coils and the implanted receiving
when stimulation at high pulse rates is required. Existing coils renders the multiple-link technique impractical for
receiver-stimulators have maximum pulse-rate capabili- large numbers of electrodes.
ties of about 15100 kHz. However, research has shown Whether stimulating with analog waveforms confers
that different waveshapes can have different effects when perceptual benefits over stimulating with simple rectan-
neural tissue is stimulated electrically. Some CI devices gular pulses is a research question that is presently unre-
have the capability for delivering waveforms other than solved. Sound-processing schemes using each technique
rectangular biphasic current pulses to the electrodes. have provided many users of CIs with the ability to un-
These waveforms are generally referred to as analog derstand speech, as discussed further later. In normal
waveforms, because in some CI systems, the stimulus hearing, the mechanical stimulation of hair cells by vibra-
waveform represents a continuously varying waveform de- tions resulting from the variations in pressure associated
rived from the microphone signal, usually after some pro- with acoustic signals produce action potentials in auditory
cessing such as filtering (see Fig. 4, bottom). However, it is neurons. As action potentials are all-or-nothing events,
also possible to generate nonrectangular waveforms by good reason exists to suggest that the use of brief current
specifying the current to be delivered to the active elec- pulses to stimulate auditory neurons directly can convey as
trodes at each of a number of discrete time instants. With much information about sound signals as can nonrectan-
this technique, employed in certain existing receiver-stim- gular waveforms. On the other hand, it is well known that,
ulators, it is feasible to construct an approximation to any in normal hearing, auditory neurons that innervate the
desired nonrectangular waveform while retaining the cochlea at sites distributed along the length of the basilar
practical advantages of digital encoding of the data trans- membrane convey information about incoming acoustic
mitted to the implant. signals concurrently. Based on this fact, it seems reason-
One way of avoiding nearly all restrictions on stimulus able that CIs that can stimulate several locations in the
waveshape is to provide a direct connection between the cochlea simultaneously might provide perceptual advanta-
external sound processor and the implanted electrodes, ges over devices that are restricted by design to present
rather than an implanted receiver-stimulator driven via a stimulation on multiple electrodes nonsimultaneously.
transcutaneous inductive link. Such a percutaneous con-
nector was used with the Ineraid CI, which is now obso-
2.4. Simultaneous Stimulation
lete (6). This system relied on a plug and socket to connect
the body-worn electronics with each of 8 (6 intracochlear, 2 As mentioned above, CIs that employ a percutaneous con-
extracochlear) electrodes. Although an arrangement of nector, or several independent inductive links, to convey
this type is extremely flexible in terms of stimulus wave- stimuli to multiple intracochlear electrodes are inherently
forms and electrode configurations, it has three serious capable of providing stimulation with analog waveforms
disadvantages. First, the permanent protrusion of the simultaneously to several subpopulations of auditory neu-
connector through the skin increases the risk of infection rons. Some implant designs have achieved similar capa-
and local tissue damage. Second, it is difficult to fabricate bilities using a single inductive link and digital encoding
a reliable connector of acceptable size for use with a large of stimulus parameters. In general, these devices contain
number of implanted electrodes. Third, the visibility of the a complex receiver-stimulator that converts data specify-
connector would be undesirable to many potential users of ing the required pattern of stimulation into a set of stim-
CIs. Nevertheless, percutaneous connectors are valuable ulating currents that are delivered to the selected active
for short-term use in research investigating the percep- electrodes independently. For example, the HiRes CI sys-
tual characteristics of stimuli with parameters outside the tem, manufactured by Advanced Bionics Corporation (9),
range that can be generated by existing CIs. is capable of receiving data that specifies the current to be
For small numbers of implanted electrodes, it is possible delivered by up to 16 independent electrodes simulta-
to provide a separate inductive link for each intracochlear neously. The rate at which these data can be processed
electrode pair. This technique was used in previous single- is high enough to permit the currents on the active elec-
channel (1,7) and 4-channel (8) devices, and is used in the trodes to be modified approximately every 12 ms. Alterna-
present AllHear single-channel system (2). Generally, the tively, data can be transmitted to the implant that
inductive links in these devices convey analog waveforms specifies nonsimultaneous stimulation. Bipolar and mono-
to the electrodes by modulating a radio-frequency carrier. polar electrode configurations are available.
The process of demodulating the signal received by the Numerous technical problems still need to be overcome
implant is much less complex than decoding digital data, to enable simultaneous stimulation to be produced on
and therefore the implanted electronic circuitry is simple multiple electrodes without undesirable side effects. One
and consumes very little power. The AllHear system is dis- important problem derives from the summation of cur-
tinctive in that it employs a relatively low carrier fre- rents within the cochlea. As mentioned above, for the use
quency (16 kHz) and delivers the amplitude-modulated of multiple electrodes to be beneficial perceptually, it is
carrier waveform directly to the stimulating electrode essential that the subpopulations of auditory neurons that
without demodulation. An advantage of using multiple in- are excited by each electrode be at least partially non-
ductive links in the devices with more than one active elec- overlapping. With a monopolar electrode configuration,
trode is that independent, analog stimulating waveforms currents produced simultaneously by several electrodes
can be delivered to the active electrodes simultaneously. combine in the conductive fluids and tissues surrounding
892 COCHLEAR IMPLANTS

the electrode array and the target population of neurons.


The occurrence of such current summation means that the
effects of the stimulation produced by each electrode are
dependent in a complex manner on the currents produced
by all other electrodes that are active at overlapping
times. In the Clarion device (10), some degree of indepen-
dence among the simultaneously active electrodes is
achieved by delivering currents to the electrodes while
they are configured as bipolar pairs, with the members of
each pair closely spaced. In addition, the stimulator cir-
cuitry is designed to prevent currents destined for a par-
ticular electrode pair from flowing to any other electrodes
Figure 7. Photograph of a complete cochlear implant system,
that are active at the same time.
including the external sound processor (left), the implantable de-
In addition to the potential perceptual benefits deriving
vice (center), and a programming interface unit (right). (This fig-
from the simultaneous activation of subpopulations of au- ure is available in full color at http://www.mrw.interscience.
ditory neurons at distinct sites, a further reason for pro- wiley.com/ebe.)
viding simultaneous stimulation capabilities in CIs is to
exploit some of the effects of current summation. For ex- improvements in the design of sound processors (which at
ample, it is possible in theory to create stimulating cur- present must be worn externally by implant users). One
rent distributions that are more restricted spatially than recent CI system is illustrated in Fig. 7, showing three
those resulting from a bipolar electrode configuration by main components: a sound processor, designed to be worn
employing three intracochlear electrodes simultaneously. behind the ear; the implantable receiver-stimulator and
In this tripolar configuration (sometimes referred to as electrode array; and a programming interface that enables
quadrupolar), a central electrode is used to deliver current the sound processor to be connected to a computer. A large
that is returned in equal parts by the two adjacent elec- number of distinct sound processing schemes have been
trodes. In practice, the tripolar configuration can be cre- developed and evaluated for CIs so far (11). For various
ated by activating simultaneously two bipolar electrode reasons, it is not realistic to rank implant systems or
pairs that share one central electrode. Experimental stud- sound processor techniques in a strict order of perfor-
ies suggest that tripolar stimulation may result in en- mance. For example, different designs have been shown to
hanced spatial selectivity in the target subpopulation of provide comparable levels of benefit to individual implant
auditory neurons by comparison with bipolar (or monopo- users, and furthermore, considerable variability exists in
lar) stimulation. The same principle can, of course, be ap- the benefit users may obtain from the same devices and
plied to any larger number of electrodes to produce a processing schemes. Nevertheless, three broad categories
particular current density distribution along the length of of sound-processing schemes (often referred to as strate-
the cochlea. Whether this would provide perceptual ben- gies) can be identified. Most of the successful strategies
efits to implant users in comparison with stimulating at that have been developed for a particular implant device
only one site at a time has not yet been determined. can also be used by people implanted with a different de-
To summarize, the development of CI devices and the vice, at least in principle, by changing or reprogramming
evaluation of their performance over time has enabled the sound processor. In spite of the commercial rivalries
some general principles of good implant design to be iden- among device manufacturers, many fundamental princi-
tified. First, better perceptual performance is likely to be ples of sound-processing schemes are similar across sys-
obtained if multiple electrodes are used rather than a sin- tems in each category. In the following, certain superseded
gle electrode. Secondly, it is preferable for the stimulating techniques, including those employed in single-channel CI
electrodes to be located relatively close to the target audi- systems, are not described in detail.
tory neurons, and in an arrangement that enables the to-
notopic organization of the cochlea to be exploited. Third, it
3.1. Feature-Extracting Strategies
is possible to provide substantial perceptual benefit to im-
plant users by stimulating with rectangular current pulses One early and distinctive approach to the design of sound
delivered sequentially to multiple electrodes, although the processors was taken by researchers at the University of
use of analog stimulus waveforms or simultaneous stimu- Melbourne, Australia, and the CI manufacturer Cochlear
lation may also be worthwhile. Research has also shown Limited (formerly Nucleus Limited) (5). Initially, this ap-
that the degree of benefit obtained by deaf individuals from proach focused on providing implant users with specific
CIs depends very heavily on the methods used to process information about speech sounds. A limited set of acoustic
sound signals and to generate stimulation patterns that features of the speech signal that were known to be im-
represent those sounds. This topic is discussed below. portant for speech recognition was identified. Stimuli rep-
resenting some parameters of those features were
generated by delivering trains of nonoverlapping pulses
3. SOUND PROCESSING FOR COCHLEAR IMPLANTS to a small number of electrodes selected from an intra-
cochlear array of 22 electrodes. In an evolving series of
The evolution of design of the implantable components of designs, the number of features represented was increased
hearing prostheses has been accompanied by continual progressively, resulting in improvements in the average
COCHLEAR IMPLANTS 893

Stimulus
BPF encoder Output
4 - 6kHz data to
cochlear
implant
BPF
2.8 - 4kHz
Acoustic to electric
level conversion
BPF
2 - 2.8kHz
Input
signal
F2 F2 frequency to
estimator electrode conversion

F1 F1 frequency to
estimator electrode conversion

F0 F0 frequency to pulse
estimator rate conversion

Voicing Active electrode


Figure 8. Functional block diagram of the
detector selection
Multipeak (MPEAK) speech-processing scheme.

speech perception of implant users. The ultimate in this stimulate at a constant pulse rate, the stimulation period
series of sound processors was the Multipeak (or in MPEAK depended on whether voicing (i.e., vibration of
MPEAK) strategy (12), a block diagram of which appears the speakers larynx) was detected in the input signal. If
in Fig. 8. so, the fundamental frequency (F0) of the speech signal
For vowels and other voiced speech sounds, the fre- was estimated, and the stimulus period was set equal to
quency spectrum of the acoustic signal is characterized by the inverse of F0. The four pulses presented within the
a number of broad peaks that are associated with reso- period represented F1, F2, and the lower two of the three
nances in the vocal tract. Two of these peaks correspond to high-frequency bands. If voicing was not detected, a stim-
the first two formants (known as F1 and F2). As the lower ulation rate of about 250 Hz was used, and the four pulses
formants carry information that enables many speech presented in each period represented F2 and all of the
sounds to be recognized, parameters describing them three high-frequency bands.
were estimated in the MPEAK strategy. The frequency Although research studies demonstrated that feature-
ranges of the acoustic signal associated with F1 and F2 estimating schemes including MPEAK could provide
were extracted by bandpass filters, and the corresponding many implant users with information sufficient to enable
amplitudes were estimated by means of envelope detec- the recognition of most speech sounds, feature estimation
tors. The center frequencies of the formant peaks were has several inherent disadvantages. A major problem is
estimated by measuring the periods of the waveforms that it is technically difficult to obtain accurate estimates
passed by the filters using zero-crossing detectors. These of the relevant parameters of speech signals in a real-time
frequency estimates were converted to the locations of the processor, especially in unfavorable conditions. Situations
active electrodes, such that an apical set of electrodes was with high levels of background noise, multiple speakers,
assigned to F1 (which has a range of about 300 to 800 Hz) or signals with parameters outside the expected ranges
and a basal set of electrodes was assigned to F2 (which has (such as music and environmental noises) can result in
a range of about 800 to 2800 Hz). Three additional band- unreliable performance. These considerations led to the
pass filters and envelope detectors estimated the ampli- development of alternative processing schemes (described
tude of the incoming signal in three higher frequency later), which were also designed to present information
regions (2.02.8, 2.84.0, and 4.06.0 kHz). These filters about prominent spectral features of sounds, but did not
were assigned to three fixed electrodes near the basal end rely on the assumption that those spectral features were
of the array. Each of the five amplitude estimates was always associated with speech.
converted to a stimulation level using an instantaneous
nonlinear function, as described further later. However,
3.2. Simultaneous Analog Schemes
only four of the above feature estimates were used to con-
trol stimulation on four active electrodes in each stimulus Analog sound-processing schemes have been used success-
period, and the positions of only the electrodes represent- fully in several earlier CI systems (1,7,8). At present, a
ing the frequencies of F1 and F2 were variable. Unlike scheme of this type, called Simultaneous Analog Stimula-
most other modern sound-processing schemes, which tion (10), is available as a programming option in CI
894 COCHLEAR IMPLANTS

BPF Level
information about speech is present. Typical center fre-
4 control 4 quencies are 0.5, 1.0, 2.0, and 4.0 kHz. The signal passed
Input by each filter is amplified by an adjustable amount, and
signal finally delivered to the corresponding electrode. When the
(after BPF Level Output sound processor is fitted to an implant user, the gain in
AGC) 3 control 3 signals to each of the final stages of the processor is adjusted to pro-
cochlear duce appropriate loudness when speech (and other
implant
BPF Level sounds) are picked up by the microphone. One important
2 control 2 electrodes
characteristic of processing schemes of this type is that the
waveforms of the signals in each of the bandpass filters
BPF Level (rather than the amplitudes of those signals envelopes)
1 control 1 are used as the basis of the electric stimuli produced by
the electrodes. As mentioned previously, the simultaneous
Figure 9. Functional block diagram of the Compressed Analog
delivery of multiple analog current waveforms by an array
(CA) sound-processing scheme.
of intracochlear electrodes can lead to undesirable inter-
actions that adversely affect the auditory perception of
some implant users. Part of the rationale for the develop-
systems manufactured by Advanced Bionics Corporation ment of nonsimultaneous pulsatile schemes was to pre-
(9). A generic block diagram of a sound processor employ- vent such interactions.
ing simultaneous analog stimulation with four channels is
shown in Fig. 9. Sounds picked up by the microphone are
3.3. Nonsimultaneous Pulsatile Schemes
converted to electric signals that are amplified and com-
pressed using an Automatic Gain Control (AGC) circuit. A block diagram of a widely used sound-processing strat-
The compression reduces the dynamic range of electrical egy known as Continuous Interleaved Sampling (CIS) is
levels in an attempt to match the relatively narrow range shown in Fig. 10. As the name implies, the CIS technique
of stimulus levels that is acceptable for the implant user. presents current pulses cyclically to a number of elec-
This range is delimited by the minimum level at which trodes such that only one electrode (at most) is conducting
hearing sensations are just audible, and by the maximum current at any time instant. In a CIS system, much of the
level that produces a comfortable loudness. As a result of processing of microphone signals is similar to that em-
this reliance on reducing the dynamic range of signals be- ployed in a generic simultaneous analog scheme as de-
fore delivery to the electrodes, an early version of this type scribed above. An initial version of a CIS processor was
of processing strategy was known as the compressed an- developed for use with the Ineraid CI, and provided one
alog scheme (6). The compressed signal is divided into four bandpass filter for each of the six intracochlear electrodes
partially overlapping frequency bands by a bank of band- available in that device (13). A CI system developed pre-
pass filters. The filters have center frequencies that are viously in Paris, France, also used a CIS-like processing
usually spaced approximately uniformly on a logarithmic scheme (14). In that system, the levels of signals in each
scale encompassing the frequency spectrum in which most of 12 bandpass filters modulated the levels of pulses

BPF
6

BPF
5
Acoustic to electric
level conversion
BPF
Input
4
signal

BPF
3 Stimulus
encoder
Output
BPF data to
2 cochlear
implant

Figure 10. Functional block diagram of the Constant rate


BPF pulse generator
Continuous Interleaved Sampling (CIS) 1
sound-processing scheme.
COCHLEAR IMPLANTS 895

presented in a nonoverlapping cyclical sequence to an in- CIS, several variations of this type of processing have been
tracochlear array of 12 electrodes. At present, CI systems developed, although the main functional principles are
based in part on this relatively early work are manufac- similar in all existing practical implementations.
tured by Laboratoires MXM in France (15). Versions of A block diagram of the ACE scheme appears in Fig. 12.
CIS processors have been developed for implants with dif- The bandpass filters and envelope detectors are function-
ferent numbers of active electrodes and corresponding ally similar to those used in other sound-processing
bandpass filters, and with several variations to the de- schemes, such as CIS (Fig. 10). However, a relatively large
tails of the signal processing. However, the main func- number of filter channels is provided, with each channel
tional blocks shown in Fig. 10 are common to all CIS assigned tonotopically to each intracochlear electrode
implementations. available in the implant. The filters have partially over-
In the figure, a six-channel processor is shown. Al- lapping frequency responses. Typically, 20 filters encom-
though the bank of bandpass filters is similar to that em- pass a frequency range of about 150 Hz to 10 kHz. In each
ployed in the analog scheme (Fig. 9), CIS has additional stimulation period, the outputs of the envelope detectors
processing in each frequency channel, and introduces a are compared, and the channels with the highest short-
pulse generator to control the stimulation sequence. Each term amplitudes are identified. The number of channels
of the bandpass-filtered signals is processed by a detector selected is limited, usually to approximately eight chan-
circuit, such as a rectifier and low-pass filter, to extract the nels. The amplitudes of these channels are converted to
envelope of the waveform. These envelope signals are appropriate levels of stimulation using an instantaneous
sampled at regular times determined by the pulse gener- conversion similar to that used in other nonsimultaneous
ator, and their amplitudes are converted to stimulation pulsatile schemes (described further below). The period of
levels by means of an instantaneous nonlinear function. stimulation, within which the eight selected electrodes are
As described further below, this amplitude conversion activated, varies among implementations of this type of
function ensures that a suitable range of acoustic signal scheme. Stimulation rates as low as 250 Hz per electrode
levels is perceived by the implant user to have appropriate have been found to provide satisfactory performance for
loudness. An example of an amplitude-modulated stim- many implant users (17), although evidence exists that
ulus pulse train, similar to the waveform that a CIS higher rates (up to at least 2.4 kHz per electrode) are often
scheme would present on each electrode, is shown in beneficial (18). Other parameters, such as the number and
Fig. 11. The rate at which stimulus pulses are delivered design of the bandpass filters, and the number of channels
to the electrodes varies among implementations of the CIS selected in each stimulation period, can also be adjusted.
scheme, and is limited in any case by the capabilities of In general, if the number of electrodes available for
the implanted receiver-stimulator. In typical CI systems, stimulation is small, and therefore the number of corre-
pulse rates of at least 800 Hz per electrode are applied, sponding bandpass filters is also small, then there may be
although in some instances either much higher or slightly no advantage in selecting channels dynamically. For ex-
lower rates have been used successfully. Studies investi- ample, if only six electrodes are available, use of the CIS
gating the perceptual effects of different pulse rates with scheme, in which the outputs of all six filters are repre-
CIS schemes (and with other nonsimultaneous pulsatile sented in each period, might be most appropriate. How-
strategies) have found that the use of higher pulse rates ever, if the number of available intracochlear electrodes is
does not always result in better performance. large (up to 22 in some existing implant devices), present-
Another sound-processing scheme that is widely used ing stimulation on a periodically changing subset of elec-
at present is sometimes described generically as the n- trodes may be advantageous. In particular, because the
of-m scheme, which is because n outputs of m bandpass overall stimulation rate possible for any nonsimultaneous
filters are represented by electric pulses in each stimula- pulsatile scheme is limited by the capabilities of the im-
tion period, with nom. An initial version of this scheme planted receiver-stimulator, it may be effective to distrib-
was called the Spectral Maxima Sound Processor (SMSP) ute the rate among a relatively small number of
(16). Modified versions of this strategy have been devel- electrodes, rather than activating all electrodes cyclically,
oped for use in commercial CI systems, including the which may improve the representation of relatively fast
SPEAK and ACE schemes of the Nucleus CI system amplitude modulations in the signals passed by each
manufactured by Cochlear Limited, Australia. As with bandpass filter. Furthermore, selecting only the filters

CURRENT

Figure 11. Example of a stimulating current


waveform comprising a train of biphasic
pulses with amplitudes modulated to repre-
sent a time-varying analog signal (dotted
TIME line).
896 COCHLEAR IMPLANTS

BPF
n

Selection of bands
with highest levels
Input BPF
signal 4

Acoustic to electric
BPF level conversion
3

BPF Stimulus encoder


2
Output
data to
cochlear
Figure 12. Functional block diagram of the BPF Constant rate implant
Advanced Combination Encoder (ACE or 1 pulse generator
SPEAK) sound-processing scheme.

that contain the highest short-term amplitudes can reduce The overall range of currents allowed on any electrode
the amount of background noise presented to the implant in each implant user is determined when the sound pro-
user in some conditions. For example, in the common sit- cessor is fitted. The lowest current, at which a sound is just
uation where high-level speech is mixed with a spectrally audible, is generally called the threshold or T-level. Usu-
broad, relatively low-level noise, the filters with the high- ally a higher current, at which the sensation is loud but not
est amplitudes may contain primarily the spectral fea- uncomfortable, is also estimated, and is often called the C-
tures that carry most speech information (such as the level. The values for the T- and C- levels vary as a function
peaks corresponding to the first few formants), rather of pulse width, pulse rate, electrode configuration, and
than components of the noise. other parameters. They may vary widely across electrodes
in a given implant user, and even more widely among im-
plant users. However, the ratio between the current at the
3.4. Stimulation Level Control
C-level and the current at the T-level is much smaller than
All sound-processing schemes that generate nonsimulta- the overall range of T- and C- levels observed among a
neous pulsatile stimulation incorporate a conversion func- large population of implant users. That ratio, known as the
tion to relate the amplitude of each stimulus pulse to a electrodes dynamic range, is often less than about 10 dB
corresponding level derived from the incoming acoustic (i.e., a ratio of about 3:1).
signal. Psychophysical studies of the way loudness varies In contrast, the dynamic range of acoustic intensities
as a function of stimulus amplitude have suggested that, for normal hearing is generally at least 90 dB. In most ex-
for a steady pulse train with other parameters held con- isting sound processors for CIs, it is not practical to pro-
stant, a change in loudness by a given ratio occurs when vide a dynamic range as large as that at the microphone
the electric current changes by a ratio rather than by an input, which is because electrical noise in the microphone,
absolute amount (19). This relationship is comparable preamplifier, and related circuits usually prevents sounds
with the well-established corresponding relationship for with levels near the threshold for normal hearing (ap-
normal hearing, in which a doubling (or halving) of loud- proximately 0 dB SPL) from being detected reliably. At the
ness occurs when acoustic intensity is increased (or de- other extreme, acoustic levels above about 90 dB SPL are
creased) by approximately 10 dB (20). To produce a similar considered uncomfortably loud by many people with nor-
relationship between loudness and acoustic intensity for mal hearing. In many CI sound processors, high input
CI users, a nonlinear function is applied that converts levels (e.g., above about 70 dB SPL) are held to a constant
acoustic intensity into the current amplitude of the stim- level by a fast-acting compression limiter. This limiter
ulus pulses. Although the shape of this function varies helps to prevent overloading and consequent distortion in
among sound-processor implementations, and in some the filters and other sound-processing stages of the sys-
cases can be altered during the fitting of a sound proces- tem. Thus, the effective dynamic range of signals at the
sor to suit the CI user, the general form of the function is microphone input of CI systems may be restricted to only
typically logarithmic, which means that a ratio change in approximately 30 dB in some sound-processor designs, al-
acoustic intensity (i.e., a change by a constant number of though the dynamic range can be much wider (e.g., 60 dB)
decibels) produces a ratio change of stimulus current. in other existing systems.
COCHLEAR IMPLANTS 897

The compression of the dynamic range at the audio in- the implant, and minimizes the power consumption of the
put of the sound processor to be compatible with the much receiver-stimulator. However, it can happen that C-levels
smaller dynamic range needed for the electric stimuli pro- are unattainable with the minimum pulse width avail-
duced at the implanted electrodes can be performed in two able, either because the highest possible current does not
stages. The first stage is wide dynamic range compression, produce sensations that are loud enough, or because of
which has been used successfully in hearing aids and insufficient voltage compliance at the output of the re-
other amplification systems for many years. This type of ceiver-stimulator (as explained previously). In these cases,
compression employs an amplifier with automatic gain the pulse width can be increased, and the T- and C- levels
control (AGC). The AGC acts to reduce the variation in redetermined. In certain implant systems, the appropriate
output level when a change in input level occurs. For ex- setting of T- and C- levels is simplified by the use of a sin-
ample, with a compression ratio of 2:1, a change in input gle level parameter that varies both the current and pulse
level of 10 dB would result in a change in output level of width together. For example, in some of the CI systems
5 dB. The speed at which gain changes can occur in re- manufactured by Cochlear Limited, this parameter is re-
sponse to changes in level is defined in terms of an attack ferred to as stimulus level. By this means, a stimulus-level
time and a release time. In a typical compressor, the at- parameter having a relatively small number of discrete
tack time (defined as the time for the AGC to respond to a steps can vary the total charge per phase of the stimulus
level increase of 25 dB) is about 25 ms and the release pulses (i.e., the product of current amplitude and pulse
time (response time for a level decrease of 25 dB) is about width) over a wider range than is possible when either
50200 ms. Longer release times, up to several seconds, current or pulse width is varied separately. In a sound
are used in some systems. Generally, AGC circuits are de- processor programmed to use the stimulus-level rather
signed to reduce overall variations in the average level of than the current-level parameter, the output of the am-
signals while preserving the size of relatively rapid level plitude conversion function is modified to produce stim-
changes, mainly because more rapid level changes convey ulus levels, but the effective shape of the function is
information important for the recognition of speech generally not changed.
sounds. If it is preferable to keep the pulse width fixed (ideally
The second stage of compression is provided by the to a relatively small value that maximizes the possible
amplitude conversion function that is applied to signals stimulation rate), an alternative way of reducing the cur-
after they have been filtered and their envelopes have rent required to elicit appropriate sensations at the T- and
been estimated. Although this conversion is often de- C- levels is to change the configuration of the active elec-
scribed as being instantaneous, time constants are inher- trodes. For example, changing from a bipolar to a mono-
ent in the envelope-estimation process. Those time polar configuration, or increasing the spatial extent with a
constants are usually much shorter than the release bipolar configuration, will reduce the current require-
time of typical AGC circuits. For example, if the envelope ments. Furthermore, the current requirements are lower
detectors comprise a rectifier followed by a low-pass filter, for electrodes that are located closer to the auditory neu-
time constants of a few milliseconds are typical. rons. In general, T- and C- levels can be obtained with
The approximately logarithmic relationship between shorter pulse widths using monopolar electrodes designed
loudness and the current amplitude of pulsatile stimula- to lie close to the modiolus of the cochlea than when using
tion has led to the use of a nonlinear scale of currents in bipolar electrodes on an array that is inserted more loosely
some implantable receiver-stimulators. In these devices, into the scala tympani.
the digital amplitude parameter transmitted to the re- The detailed design of both the AGC that compresses
ceiver-stimulator results in changes of current by a con- input signals and the amplitude conversion function that
stant ratio when the parameter changes by a constant controls stimulation levels differ among implant systems,
number of steps. The parameter is referred to as current and several research studies have investigated the effects
level, and causes the stimulus current magnitude to vary of modifying them (21,22). To maximize the perceptual
over a wide range across the range of possible values of the performance of a sound processor, and in particular to
level parameter. For example, in the CI24 implant man- maximize the intelligibility of speech, two requirements
ufactured by Cochlear Limited, the current-level param- are important. First, the minimum acoustic level that is
eter has 8 bits, and therefore 256 discrete steps. The audible (i.e., the input level just producing stimulation at
stimulus current in this device ranges from approximately the T-level) must be low enough to ensure that all infor-
10 mA to 1.75 mA, with each step corresponding to a mation-carrying components of the speech signal can be
change in the current of 2%. perceived by the implant user. Second, changes in level
As the overall range of currents required to ensure that above the minimum input level must be perceptible, at
T- and C- levels can be obtained on all electrodes in every least over the range of acoustic levels present in speech
implant user is much wider than the current range needed signals. Whether these requirements are met fully in a
on any individual electrode, it may be necessary to select sound processor depends not only on the operation of the
an appropriate stimulus pulse width when the users AGC and the amplitude conversion function, but also on
sound processor is fitted. The width of the pulses affects the sensitivity of the microphone. Usually, the sensitivity
the loudness of sensations in much the same way as the can be adjusted by the user with a manual control. Some
current amplitude. Generally, it is desirable to use the types of AGC, particularly those with large compression
minimum possible pulse width, because this both maxi- ratios or long release times, may reduce the need for man-
mizes the overall rate at which pulses can be delivered by ual sensitivity adjustments to compensate for variations
898 COCHLEAR IMPLANTS

in the overall level of input signals. When the input signal However, even for the most successful CI users, under-
is speech, and its average level is constant, the relatively standing speech in the presence of high levels of back-
rapid fluctuations in intensity at each frequency must be ground noise is much more difficult than for listeners with
perceived by the implant user to maximize intelligibility. normal hearing. The use of a directional microphone can
These fluctuations encompass at least 30 dB (23). There- provide substantial benefit, provided that some spatial
fore, even in the absence of wide dynamic range compres- separation exists between the signal of interest and the
sion at the input of a sound processor, the range of levels unwanted noise. The simultaneous use of implants in both
applied to the amplitude conversion function is at least ears can also be advantageous in noisy situations (24). In
30 dB. As the average level of speech and other sounds can general, CI recipients obtain inadequate performance
also vary widely, it is advantageous for the sound proces- from currently available devices when listening to music
sor to be able to control the level of stimulation as a func- or environmental sounds. Research has shown that iden-
tion of acoustic intensity over a range greater than 30 dB. tification of melodies, musical instruments, and common
However, increasing the amount of compression applied nonspeech noises is usually unsatisfactory, and many CI
by the instantaneous amplitude conversion function may users report experiencing minimal enjoyment from listen-
be less effective than increasing the compression applied ing to music (25).
by the slower-acting AGC at the input of the sound pro- A recent development that promises to alleviate many
cessor. This is because excessive instantaneous compres- of the above problems, at least for people with certain
sion may reduce the amount of information available to types of hearing impairment, is the implantation of ears
implant users in the form of rapid changes in stimulation that have some usable acoustic hearing (26). Typically,
level, which are known to convey important speech cues. hearing loss affects sensitivity to sounds at high frequen-
In contrast, increasing the compression in the AGC circuit cies more than at low frequencies. In some cases, the high-
while maintaining a relatively long release time constant frequency impairment is so severe that conventional
has the advantage of reducing the range of average levels acoustic aids provide no benefit, although sufficient low-
to be processed by the instantaneous conversion function frequency hearing remains that the person can perceive
without affecting the size of these rapid level changes. many sounds with appropriate acoustic amplification.
Such people may elect to receive a cochlear implant to ob-
tain information about high-frequency signals via electric
4. THE FUTURE OF COCHLEAR IMPLANTS stimulation. If the low-frequency acoustic hearing can be
substantially preserved in the implanted ear (or the op-
This article has focused on aspects of the design and func- posite ear), it is possible for combined acoustic and electric
tion of sound processors and implantable stimulators that stimulation to provide more auditory information than
are specific to existing cochlear implant hearing prosthe- either a hearing aid or a cochlear implant could provide
ses. Other practical aspects are common to alternative when used separately. In some clinics, a special short elec-
technologies that have been developed to assist people trode may be used in these cases to minimize damage to
with a hearing impairment, particularly acoustic hearing the apical region of the cochlea where low-frequency
aids. In both hearing aids and CIs, there has been a trend sounds are transduced (27). Although research with com-
toward smaller devices, accompanied by a progression bined acoustic and electric stimulation is relatively new,
away from analog electronic circuitry toward the use of experimental results suggest that it can be highly benefi-
digital signal processing. These changes have succeeded in cial for certain people with hearing impairment, particu-
reducing the size of devices and in improving their per- larly for understanding speech in noise and for listening
formance. With CIs, pocket-sized sound processors have to music (28). In the future it is likely that increasing
been superseded by smaller and more cosmetically accept- numbers of people with partial, rather than total, hearing
able devices that are packaged in enclosures that can be loss in both ears will become successful users of cochlear
worn comfortably on the pinna, as with many high-pow- implants.
ered acoustic HAs. This trend is likely to continue until it
is practical for all the components of a hearing prosthesis Acknowledgment
to be implanted in the body, obviating the need for any
visible external devices in normal use. This work was supported financially by the Garnett Passe
The ability of CI users to understand speech without and Rodney Williams Memorial Foundation. The images
any visual cues varies widely. Research has found that in Figs. 1, 3, and 7 were provided by courtesy of Cochlear
adults who have acquired a hearing loss after develop- Limited, NSW, Australia. The remaining figures, except
ment of spoken language skills are more likely to obtain for 5 and 9, were first published in Ref. 25, and are repro-
satisfactory performance from a CI than adults whose duced by permission of Westminster Publications r 2004.
hearing impairment occurred pre-linguistically. Some
evidence also exist that deaf children may obtain greater BIBLIOGRAPHY
benefits if they receive a CI at the earliest feasible age. At
present, the majority of CI recipients can converse easily 1. W. F. House and J. Urban, Long term results of electrode im-
in quiet conditions, especially when some contextual in- plantation and electronic stimulation of the cochlea in man.
formation is available. In such situations, many users of Ann. Otol. Rhinol. Laryngol. 1973; 82:504517.
the latest systems can communicate face-to-face or via the 2. AllHear. (2003, March 24). Home page. (online). Available:
telephone almost as well as people with normal hearing. http://www.allhear.com/.

S-ar putea să vă placă și