**4.2 Neural population**

Stimulation delivered by a CI system will result in the depolarization of neural elements, resulting in action potentials being generated that propagate to the next stage of the auditory system: the cochlear nucleus. With reference to the schematic of **Figure 8**, there is a population of spiral ganglion neurones associated with each electrode and hence each frequency channel of the CI system. As mentioned above, a channel's stimulation current will need to recruit a certain population of neurones whose firing indicates to the brain the amount of activity in a particular frequency range. Ideally, there will be a sufficient local neural population such that progressively increasing stimulation current initiates an appropriate number of action potentials, so that the brain correctly perceives the amount of acoustic activity in the channel's frequency range.

Unfortunately a discrete neural population for each channel as shown in **Figure 10a** is not always available. In **Figure 10b** only a reduced neural population is available for each channel. Hence, when there is a lot of activity in one channel, requiring recruitment of a full population of neurones, these are not available locally. It is still possible to increase the stimulation current, spread the electrical field further away from the electrode and depolarize neurones that should really be associated with another channel. While this will satisfy the perception of loudness, it generates channel interaction so that we are no longer able to deliver frequency specific information to a discrete part of the cochlea. The perception will be of a blurred or fuzzy sound, particularly a problem when trying to listen to speech in the presence of competing noise.

An alternative situation is shown in **Figure 10c** where most electrodes have a sufficient local neural population but one electrode is located in a so-called dead region [32]. When electrode 2 is stimulated it can only recruit neurones from the population belonging to electrodes either side of it, delivering information about channel 2's frequency region to other parts of the cochlea, spreading stimulation widely and interfering with the otherwise discrete frequency information being delivered by the neighboring channels.

Unfortunately, it is not currently possible to determine what the number and distribution of neural elements is for any individual. The literature is not always helpful in this area. As shown in **Figure 1**, there is great variability in outcome for the identification of monosyllabic words. Since this task involves little top-down processing, much of the variability in outcome must come from the electro-neural interface. Beyond speech understanding, examining the ability to discriminate adjacent electrodes, or intra-electrode stimulation sites [33], also showed both

**95**

*Electrical Stimulation of the Auditory System DOI: http://dx.doi.org/10.5772/intechopen.85285*

contribution to variations in outcome.

*populations rightly belonging to adjacent channels.*

**Figure 10.**

threshold) is associated with better speech understanding.

their program, is another variable that we will examine next.

**4.3 Programming the cochlear implant**

great variability between subjects and across the electrode array of individual subjects. This task, having no confound with cognitive processes related to speech understanding, further confirms the presence of peripheral variability and its likely

*A schematic representation of three different neural populations: (a) a full population exists for each channel, (b) a depleted population results in channel interaction and (c) a dead region requires recruitment from the* 

It is unclear to what degree a loss of spiral ganglion cells (SGC) in humans will follow, even after years of severe to profound deafness. Histological studies of humans who had used a cochlear implant sometimes show a reasonable correlation between CNC word score and SGC count: for example R = 0.62 [34], R = 0.9 [35] but for a small group of only 6. However, the variability is such that the same SGC count can show variations of between 30% and 75% for CNC words, or the same CNC word score can be associated with 3000 or 18,000 SGCs. Examining the threshold current for detecting electrical stimulation in a group of 130 lateral wall electrode array users [36] showed significant differences between four groups: the increase in group mean threshold being associated with a reduction in monosyllabic word score. This works suggests that a higher SGC population (lower electrical

The literature listed above indicates that there is a relationship between the number of spiral ganglion cells and the ability to identify monosyllabic words when using a cochlear implant. Contributions to speech understanding may also come from a large number of additional factors, some of which include: the distribution of SGCs, angular insertion of the electrode array, distance of the electrode contacts from the modiolar wall, presence or absence of peripheral processes, fibrous sheath formation and intrusion of new bone into the cochlea. How well a given implant user has had the parameters of their sound processor set, commonly referred to as

As has been explained above, the small electrical dynamic range available to a CI user makes it necessary to carefully adjust the stimulation parameters to suit the *Electrical Stimulation of the Auditory System DOI: http://dx.doi.org/10.5772/intechopen.85285*

#### **Figure 10.**

*The Human Auditory System - Basic Features and Updates on Audiological Diagnosis and Therapy*

strategies name. Such a strategy supported only very modest levels of speech understanding, around 8% correct for monosyllabic words presented in quiet [30]. The information extracted was limited to begin with and further reduced through errors generated in real life listening situations where background noise, reverberation and intensity and frequency response variations led to the algorithm making mistakes in

It was eventually recognized that the brain was better at extracting information than the feature extraction algorithms and hence "whole-speech strategies" replaced feature extraction. Today's sound coding strategies simply average the energy in each channel's frequency range and generate levels of stimulation that represent this. In some cases a so-called n-or-m strategy will work out which subset (n) channels from the total (m) number available have the highest energy and then only stimulate this reduced set. Refinements to this may neglect adjacent channels on the basis that stimulating both will not add anything, so select a more distant

Stimulation delivered by a CI system will result in the depolarization of neural elements, resulting in action potentials being generated that propagate to the next stage of the auditory system: the cochlear nucleus. With reference to the schematic of **Figure 8**, there is a population of spiral ganglion neurones associated with each electrode and hence each frequency channel of the CI system. As mentioned above, a channel's stimulation current will need to recruit a certain population of neurones whose firing indicates to the brain the amount of activity in a particular frequency range. Ideally, there will be a sufficient local neural population such that progressively increasing stimulation current initiates an appropriate number of action potentials, so that the brain correctly

both the extraction of formant frequencies and in the estimation of Fo.

lower amplitude electrode to transmit more information [31].

perceives the amount of acoustic activity in the channel's frequency range.

Unfortunately a discrete neural population for each channel as shown in **Figure 10a** is not always available. In **Figure 10b** only a reduced neural population is available for each channel. Hence, when there is a lot of activity in one channel, requiring recruitment of a full population of neurones, these are not available locally. It is still possible to increase the stimulation current, spread the electrical field further away from the electrode and depolarize neurones that should really be associated with another channel. While this will satisfy the perception of loudness, it generates channel interaction so that we are no longer able to deliver frequency specific information to a discrete part of the cochlea. The perception will be of a blurred or fuzzy sound, particularly a problem when trying to listen to speech in the

An alternative situation is shown in **Figure 10c** where most electrodes have a sufficient local neural population but one electrode is located in a so-called dead region [32]. When electrode 2 is stimulated it can only recruit neurones from the population belonging to electrodes either side of it, delivering information about channel 2's frequency region to other parts of the cochlea, spreading stimulation widely and interfering with the otherwise discrete frequency information being

Unfortunately, it is not currently possible to determine what the number and distribution of neural elements is for any individual. The literature is not always helpful in this area. As shown in **Figure 1**, there is great variability in outcome for the identification of monosyllabic words. Since this task involves little top-down processing, much of the variability in outcome must come from the electro-neural interface. Beyond speech understanding, examining the ability to discriminate adjacent electrodes, or intra-electrode stimulation sites [33], also showed both

**4.2 Neural population**

presence of competing noise.

delivered by the neighboring channels.

**94**

*A schematic representation of three different neural populations: (a) a full population exists for each channel, (b) a depleted population results in channel interaction and (c) a dead region requires recruitment from the populations rightly belonging to adjacent channels.*

great variability between subjects and across the electrode array of individual subjects. This task, having no confound with cognitive processes related to speech understanding, further confirms the presence of peripheral variability and its likely contribution to variations in outcome.

It is unclear to what degree a loss of spiral ganglion cells (SGC) in humans will follow, even after years of severe to profound deafness. Histological studies of humans who had used a cochlear implant sometimes show a reasonable correlation between CNC word score and SGC count: for example R = 0.62 [34], R = 0.9 [35] but for a small group of only 6. However, the variability is such that the same SGC count can show variations of between 30% and 75% for CNC words, or the same CNC word score can be associated with 3000 or 18,000 SGCs. Examining the threshold current for detecting electrical stimulation in a group of 130 lateral wall electrode array users [36] showed significant differences between four groups: the increase in group mean threshold being associated with a reduction in monosyllabic word score. This works suggests that a higher SGC population (lower electrical threshold) is associated with better speech understanding.

The literature listed above indicates that there is a relationship between the number of spiral ganglion cells and the ability to identify monosyllabic words when using a cochlear implant. Contributions to speech understanding may also come from a large number of additional factors, some of which include: the distribution of SGCs, angular insertion of the electrode array, distance of the electrode contacts from the modiolar wall, presence or absence of peripheral processes, fibrous sheath formation and intrusion of new bone into the cochlea. How well a given implant user has had the parameters of their sound processor set, commonly referred to as their program, is another variable that we will examine next.

#### **4.3 Programming the cochlear implant**

As has been explained above, the small electrical dynamic range available to a CI user makes it necessary to carefully adjust the stimulation parameters to suit the requirements of each individual recipient. The most important adjustment is the amount of stimulation that will be delivered in response to acoustic activity. This must be done for each of the CI's separate channels. Each channel has two primary parameters that control its output. One will be typically called a most comfortable level, shortened to either M-Level or C-Level. The other is a threshold control, referred to as T-Level. The main CI manufacturers use these parameters slightly differently but to a good approximation T-Level sets the minimum stimulation level that the implant will deliver and M-Level will set the maximum stimulation level that can be delivered for an individual recipient. The sound processor will then arrange for the amount of acoustic range that it handles, somewhere between 40 and 80 dB depending on the user's setting and implant model, to be mapped to stimulation levels between T- and M-Level. In combination with the AGC of the system this will give the CI user access to their acoustic environment such that hearing levels of between 20 and 30 dB HL are achieved across the frequency range 250–8000 Hz. The combination of AGC and M-Level ensures that even high intensity sounds of 100 dB SPL do not produce uncomfortably loud sensations. Unlike acoustic hearing, it is generally possible to provide CI users with access to the full range of frequencies that are most important for speech understanding.

Which channels are activated is another important adjustment to make. Most audiologists are reluctant to deactivate channels, although sometimes a reduced set of channels can give a better outcome. In some cases an electrode array is not fully inserted into the cochlea, perhaps due to the cochlea being too small, or there being fibrosis tissue, or bone, that prevents a full insertion being obtained. Alternatively, electrode arrays can sometimes extrude from the cochlea [37, 38], either shortly after implantation or months to years later. In all these cases the more basal electrode contacts will need to be deactivated. Deleting electrodes from a program will lead to the frequency range being remapped across the remaining electrode contacts. There will be a coarser representation of frequency since fewer channels are now available. However, removing electrodes that are not inside the cochlea will produce a better outcome than simply leaving these electrodes active.

Beyond setting T- and M-Levels and defining an appropriate set of electrode contacts, there is sometimes adjustment made to the acoustic dynamic range mapped by the sound processor. This effectively controls the compression of acoustic sounds into the electrical dynamic range. It might seem logical to use as large an acoustic or input dynamic range (IDR) as possible, since this will maximize the range of sounds available to a CI user. However, it is the discrimination of different levels of sound in each channel that carries information. An excessively large IDR may squeeze these amplitude cues, reducing the ability of an implant recipient to understand speech. There are many parameters that can be adjusted in a CI system. However, it is common for the majority to remain at their default values. This may be through an inability to obtain user feedback, for example in young children, lack of time or knowledge on the part of the clinician, or a recommendation from the CI manufacturer.

How appropriate values are found for the T- and M-levels depends very much on the individual CI user. For a post-lingually deafened adult it is reasonably straightforward to find these. By presenting 200 ms bursts of stimulation and using a standard bracketing approach, the smallest detectable amount of stimulation for each channel can be found and this value set as the T-level. Similarly, progressively increasing the stimulation will allow an M-level to be found, the CI user often pointing to different categories on a loudness chart as the various levels of stimulation are presented. These measures can be made for each individual channel, channels can be programmed in groups of four, or only five or six channels across the electrode array measured with intermediate channels set to interpolated values.

**97**

to their acoustic environment.

*Electrical Stimulation of the Auditory System DOI: http://dx.doi.org/10.5772/intechopen.85285*

ered worth the additional effort needed for measurement.

adjusted to determine the M-levels that will be used in the program.

Other objective measures are used to assist with programming, although less often due to these requiring additional equipment to be used in collaboration with the CI fitting system. There is a reasonable correlation between an electrically elicited stapedius reflex threshold (eSRT) and M-level [43]. Unlike eCAPs here the same stimulation rate can be used to measure eSRT as will be used in the everyday program. This simplifies the setting of levels and is partly behind why there is such a good correlation with M-level. Less commonly the electrically elicited auditory brainstem response (eABR) is used [44]. Again, eABR will require a lower stimulation rate to be used, so that the characteristic waveforms can be seen in up to 5 or 6 ms following stimulation. This tends to produce an extrapolated threshold for eABR quite high in an individual's electrical dynamic range. As with the eCAP and eSRT measures, it is the relative levels across channels that are important, the profile then being globally

Less frequently, some statistically based approaches are used for programming. Simple so-called "flat maps" are used where the T- and M-level is the same on each channel. These are justified by the spread of monopolar stimulation recruiting neurones form a larger section of the cochlea than associated with an individual electrode contact thus tending to produce a spatial averaging. Other approaches might use a template based on the statistical average of levels previously measured for earlier CI recipients. Approaches such as FOX [45] extend this technique, recommending a sequence of programs with progressively increasing levels that are used from the very beginning. For many CI users these techniques can work quite well, although numbers of outliers will require individually tailored programs to realize their potential outcome with comfortable stimulation and reasonable access

Plasticity in the auditory system means that over time the M-levels will usually increase. The longer term M-levels might be typically double those that can be tolerated during the initial fitting. After the first 2 months of device use, neither T- or M-levels tend to change significantly over time [46]. Change in levels is highly individual requiring the initial program levels to be revised numbers of times during the first few months of implant use. Where a second (or third) fitting session is

For babies or young children and even for some adults, objective measures are often used to help set program levels. The most common measure used is the eCAP, the electrically elicited compound action potential [39]. The ability to record eCAPs is built into the fitting systems for all of today's major CI systems. Here maskerprobe or alternate-subtraction techniques [40, 41] are used to reduce the large stimulus artifact. The amplitude of the remaining physiological signal, arising from synchronized activity on the auditory nerve, is then graphed against the stimulation level. A regression line extrapolates to intersect the stimulation axis which would correspond to a zero amplitude of eCAP. The stimulation value for which this occurs is then used as a guide for setting programming levels. Avoiding stimulus artifact and allowing sufficient neural synchronization, means that much lower stimulation rates are used when measuring eCAPs than for actual everyday stimulation. The means that the absolute eCAP values can fall at various parts of an individual's electrical range. Fortunately, it is the profile of values across the electrode array that it is important to determine. Once this is estimated a global change in level can be made to obtain appropriate loudness. In many cases the T-levels are set to 10% of the M-level since this is almost certainly not going to leave them set too high. Typically T-levels are measured at something like 25% of M-level [42]. When they can be measured and hence individually set, T-levels will tend to improve access to low intensity sounds. Often in clinical practice T-levels are set at a percentage of M-level even where they could be individually set: the additional benefit not being consid-

#### *Electrical Stimulation of the Auditory System DOI: http://dx.doi.org/10.5772/intechopen.85285*

*The Human Auditory System - Basic Features and Updates on Audiological Diagnosis and Therapy*

requirements of each individual recipient. The most important adjustment is the amount of stimulation that will be delivered in response to acoustic activity. This must be done for each of the CI's separate channels. Each channel has two primary parameters that control its output. One will be typically called a most comfortable level, shortened to either M-Level or C-Level. The other is a threshold control, referred to as T-Level. The main CI manufacturers use these parameters slightly differently but to a good approximation T-Level sets the minimum stimulation level that the implant will deliver and M-Level will set the maximum stimulation level that can be delivered for an individual recipient. The sound processor will then arrange for the amount of acoustic range that it handles, somewhere between 40 and 80 dB depending on the user's setting and implant model, to be mapped to stimulation levels between T- and M-Level. In combination with the AGC of the system this will give the CI user access to their acoustic environment such that hearing levels of between 20 and 30 dB HL are achieved across the frequency range 250–8000 Hz. The combination of AGC and M-Level ensures that even high intensity sounds of 100 dB SPL do not produce uncomfortably loud sensations. Unlike acoustic hearing, it is generally possible to provide CI users with access to the full

range of frequencies that are most important for speech understanding.

produce a better outcome than simply leaving these electrodes active.

mendation from the CI manufacturer.

Which channels are activated is another important adjustment to make. Most audiologists are reluctant to deactivate channels, although sometimes a reduced set of channels can give a better outcome. In some cases an electrode array is not fully inserted into the cochlea, perhaps due to the cochlea being too small, or there being fibrosis tissue, or bone, that prevents a full insertion being obtained. Alternatively, electrode arrays can sometimes extrude from the cochlea [37, 38], either shortly after implantation or months to years later. In all these cases the more basal electrode contacts will need to be deactivated. Deleting electrodes from a program will lead to the frequency range being remapped across the remaining electrode contacts. There will be a coarser representation of frequency since fewer channels are now available. However, removing electrodes that are not inside the cochlea will

Beyond setting T- and M-Levels and defining an appropriate set of electrode contacts, there is sometimes adjustment made to the acoustic dynamic range mapped by the sound processor. This effectively controls the compression of acoustic sounds into the electrical dynamic range. It might seem logical to use as large an acoustic or input dynamic range (IDR) as possible, since this will maximize the range of sounds available to a CI user. However, it is the discrimination of different levels of sound in each channel that carries information. An excessively large IDR may squeeze these amplitude cues, reducing the ability of an implant recipient to understand speech. There are many parameters that can be adjusted in a CI system. However, it is common for the majority to remain at their default values. This may be through an inability to obtain user feedback, for example in young children, lack of time or knowledge on the part of the clinician, or a recom-

How appropriate values are found for the T- and M-levels depends very much

on the individual CI user. For a post-lingually deafened adult it is reasonably straightforward to find these. By presenting 200 ms bursts of stimulation and using a standard bracketing approach, the smallest detectable amount of stimulation for each channel can be found and this value set as the T-level. Similarly, progressively increasing the stimulation will allow an M-level to be found, the CI user often pointing to different categories on a loudness chart as the various levels of stimulation are presented. These measures can be made for each individual channel, channels can be programmed in groups of four, or only five or six channels across the electrode array measured with intermediate channels set to interpolated values.

**96**

For babies or young children and even for some adults, objective measures are often used to help set program levels. The most common measure used is the eCAP, the electrically elicited compound action potential [39]. The ability to record eCAPs is built into the fitting systems for all of today's major CI systems. Here maskerprobe or alternate-subtraction techniques [40, 41] are used to reduce the large stimulus artifact. The amplitude of the remaining physiological signal, arising from synchronized activity on the auditory nerve, is then graphed against the stimulation level. A regression line extrapolates to intersect the stimulation axis which would correspond to a zero amplitude of eCAP. The stimulation value for which this occurs is then used as a guide for setting programming levels. Avoiding stimulus artifact and allowing sufficient neural synchronization, means that much lower stimulation rates are used when measuring eCAPs than for actual everyday stimulation. The means that the absolute eCAP values can fall at various parts of an individual's electrical range. Fortunately, it is the profile of values across the electrode array that it is important to determine. Once this is estimated a global change in level can be made to obtain appropriate loudness. In many cases the T-levels are set to 10% of the M-level since this is almost certainly not going to leave them set too high. Typically T-levels are measured at something like 25% of M-level [42]. When they can be measured and hence individually set, T-levels will tend to improve access to low intensity sounds. Often in clinical practice T-levels are set at a percentage of M-level even where they could be individually set: the additional benefit not being considered worth the additional effort needed for measurement.

Other objective measures are used to assist with programming, although less often due to these requiring additional equipment to be used in collaboration with the CI fitting system. There is a reasonable correlation between an electrically elicited stapedius reflex threshold (eSRT) and M-level [43]. Unlike eCAPs here the same stimulation rate can be used to measure eSRT as will be used in the everyday program. This simplifies the setting of levels and is partly behind why there is such a good correlation with M-level. Less commonly the electrically elicited auditory brainstem response (eABR) is used [44]. Again, eABR will require a lower stimulation rate to be used, so that the characteristic waveforms can be seen in up to 5 or 6 ms following stimulation. This tends to produce an extrapolated threshold for eABR quite high in an individual's electrical dynamic range. As with the eCAP and eSRT measures, it is the relative levels across channels that are important, the profile then being globally adjusted to determine the M-levels that will be used in the program.

Less frequently, some statistically based approaches are used for programming. Simple so-called "flat maps" are used where the T- and M-level is the same on each channel. These are justified by the spread of monopolar stimulation recruiting neurones form a larger section of the cochlea than associated with an individual electrode contact thus tending to produce a spatial averaging. Other approaches might use a template based on the statistical average of levels previously measured for earlier CI recipients. Approaches such as FOX [45] extend this technique, recommending a sequence of programs with progressively increasing levels that are used from the very beginning. For many CI users these techniques can work quite well, although numbers of outliers will require individually tailored programs to realize their potential outcome with comfortable stimulation and reasonable access to their acoustic environment.

Plasticity in the auditory system means that over time the M-levels will usually increase. The longer term M-levels might be typically double those that can be tolerated during the initial fitting. After the first 2 months of device use, neither T- or M-levels tend to change significantly over time [46]. Change in levels is highly individual requiring the initial program levels to be revised numbers of times during the first few months of implant use. Where a second (or third) fitting session is

planned within around 2 weeks of the first fitting, most of the change can already be accommodated. Looking across large numbers of adult CI users, program levels will be stable by between 3 and 9 months following first fitting. Individual practice can result in pediatric levels being more slowly increased, leading to 6–12 months being needed to see stable levels.
