Educación Media Ciclo CLE
4.3 TALLER DE SENSIBILIZACIÓN UN RECURSO PARA LA INSTITUCIÓN Como parte del apoyo a los procesos de Educación Inclusiva que se realiza en la Institución,
4.3.2.5 MATERIAL VISUAL
An important potential source of information loss associated with the electrical/neural interface is “channel interaction”. The term refers to any effect that the stimulation of one electrode channel has on the activation of a spatially separated channel (Cohen et al., 2003). An important aspect of channel interaction is that simultaneous
presentation on a group of electrodes results in distorted perception because greater cross-channel electrical interaction occurs with simultaneous presentation compared to non-simultaneous presentation (Favre and Pelizzone, 1993). The majority of current CI processing strategies, including Nucleus 24 ACE or CIS, employ non- simultaneous pulse presentation to minimize channel interaction. However, it is also clear that channel interaction does occur despite the use of non-simultaneous pulse presentation as it has been measured in users of various strategies which use non- simulataneous pulse presentation.
Channel interaction has potential consequences for consonant recognition because of both spectral and temporal information. Related to this is the idea that channel interaction has a “spatial”, or spectral, aspect, in that stimulation of an individual electrode affects adjacent frequency channels and also a “temporal” aspect in that the neural response is affected for some time after stimulation (Chatterjee and Shannon, 1998; Throckmorton and Collins, 1999). The spatial aspect has been described by a space constant of exponential decay. Stimulation of different electrodes produces overlapping electrical fields and, as a consequence, the same neurons can be activated with stimulation of different electrodes. A number of studies have attempted to quantify the decay of electrical potential within the scala tympani beyond the site of the stimulation electrode as two decaying exponentials (e.g. one either side of the stimulation electrode). Wilson et al (1994) described a model of population responses of the auditory neurons by linking a description of the electrical field patterns in the cochlea with descriptions of individual neural responses derived from the large body of work on single-neurone responses to auditory stimulation. They suggested a space
constant (of exponential decay of neural excitation) for monopolar stimulation of 3.6 mm. This approximate space constant was supported by a modeling study by Kral et al. (1998)
Black and Clark (1980) developed a three-dimensional discrete resistance model of the cochlea which indicated that current spread from monopolar stimulation was 1dB/mm as measured in the scala tympani. The length constant λ was defined as the inverse of the natural logarithm of the voltage 1mm from the recording site, divided by the voltage at the site.
1 0 1 ln − ⎟⎟ ⎠ ⎞ ⎜⎜ ⎝ ⎛ = V V λ
Equation 2.1. Current decay in the scala tympani, according to Black and Clark, 1980.
This space constant of exponential decay was used in the AM study by Laneau et al. (2006), which is discussed further in 2.6.4. One of the aims of the present study was to determine if this model could be used to explain some of the variance in consonant recognition in CI users.
Channel interaction can also be described in its temporal characteristics, which have both a “physical” and “physiological” aspect. The “physical” aspect refers to the residual charge stored in neural tissue and membrane capacitances after pulse
presentation. This aspect of channel interaction is thought to be largely dealt with by use of biphasic pulses as the second phase of a stimulation pulse should remove most of the charge delivered in the first phase. However, some residual charge could still be present and therefore one recent line of work has evaluated the use of triphasic pulses (with zero net charge) to further reduce the possibility of residual charge (Bonnet et al., 2004). However, temporal channel interaction also has a more “physiological” aspect because of the refractory property of auditory neurons. Recent work has shown that much of the channel interaction, particularly the temporal aspect occurs at the neural level e.g. stimulation of one electrode does not produce as focused a neural response as might be expected given equivalent processing in the healthy cochlea (Boex et al., 2003a; Boex et al., 2003b; de Balthasar et al., 2003). However, the
distribution of current with a specific electrode will depend on a number of factors, many of which are highly variable between individuals.
A number of possible methods are available to measure channel interaction. Pitch ranking, pitch scaling and electrode discrimination all provide indirect psychophysical measures of spatial channel interaction (Busby et al., 1994; Busby and Clark, 1997; Zwolan et al., 1997). Gap detection and forward masking have been used as
psychophysical estimates of temporal channel interaction (Chatterjee and Shannon, 1998; Blamey and Dooley, 1993), although Throckmorton and Collins (1999) argued that these “temporal” measures also reflect spectral aspects of channel interaction as they are also affected by degree of neural population overlap. The most common method of measuring channel interaction is to measure masked thresholds in which the masker and probe electrodes vary in distance. A masking function obtained in this way will show the greatest masking effect when masker and probe coincide, but by increasing the distance between masker and probe electrodes, it is possible to determine the spread of excitation. Lim (1989) found that the spread of excitation decayed more gradually in the basal direction than the apical direction, and this finding has been supported in other studies, including Cohen et al. (2003), although the pattern, along with degree, vary quite markedly between individual CI users. The same approach to separating masker and probe electrodes has been used with the electrically evoked compound action potential; this can be measured in the Nucleus 24 system by using intracochlear electrodes as recording electrodes (Cohen et al., 2003; Cohen et al., 2004; Cohen et al. 2005). Interestingly, Cohen et al. (2004) found a good correlation between psychophysical measurements of forward masking and spatial spread of excitation as estimated using the electrically evoked compound action potential measurements. The convergence of these different types of measure suggests that the measurements of channel interaction are valid. An additional finding common to both psychophysical and electrical approaches to the masking paradigm is that channel interaction increases with current level (Abbas et al., 2004).
An important implication of recent research into channel interaction (Boex et al., 2003a; Boex et al., 2003b) is that the degree, direction, time course and spread of neural excitation may be a critical factor in explaining individual differences in CI
user ability, although the evidence base for this idea is not especially strong. Zwolan et al. (1997) evaluated speech recognition for Nucleus 22 users using two different electrode configurations. In one condition, the subjects used MAPs in which only discriminable electrodes were included; in the second condition, the same users used MAPs that included all possible active electrodes. They found an overall
improvement in speech perception with the first condition. Moreover, there were marked differences in electrode discriminability (presumably an indirect measure of channel interaction) across the CI users. This was given as indirect evidence that performance improves as channel interaction is reduced, although it is not in itself a direct measure of the correlation between channel interaction and speech perception. Loizou et al. (2003) found better recognition of consonants, in particular place and voicing transmission, in users of the Clarion device with users of pulsatile non- analogue strategies which were thought to produce less channel interaction, as compared with users of an analogue strategy which was thought to produce greater channel interaction. Stickney et al. (2006) measured channel interaction by measuring masked thresholds with varying probe to masker separations and then also measured vowel consonant and sentence recognition. The authors found a high degree of
correlation between speech recognition and channel interaction when a simulataneous pulse presentation strategy was used, but there was no correlation between speech perception and channel interaction for users of an interleaved pulsatile strategy.
It is not wholly clear from the literature to what extent individual differences in performance are related to channel interaction and, more specifically, how consonant recognition in quiet and noise relates to channel interaction. It has been hypothesized in a number of studies looking at channel number that the reason that CI user
performance does not increase beyond levels achieved with around 6-10 channels is due to spectral channel interaction (see 2.4.2). A related hypothesis is that
performance in “worse” CI users can be effectively modeled by AMs with smaller numbers of channels. That is, it is hypothesized that individual variations in channel interaction place an upper limit on the number of perceptually distinct channels available to that individual CI user and that, moreover, this is an important limiting factor determining speech perception abilities. This could be tested by comparing the channel number corresponding to performance asymptote with the degree of channel
interaction. Another way to approach this question, along with the more general hypothesis that variations in channel interaction determine variations in overall speech perception ability, would be to compare CI user performance with AMs that vary in terms of channel interaction characteristics. To date, no study of consonant
recognition has used this approach although Laneau et al. (2006) applied the principle to measures of F0 perception and Fu and Nogaki (2005) applied this approach to measures of sentence recognition. These and other AM studies relating to channel interaction are discussion in 2.5.