Click here for the PDF version

Contents

Speech perception and production, as well as language, must be considered in evaluating the benefits of cochlear implants for children. They are inter-related and their development is important for the child's education, social development, and career opportunities. Language is the medium for communicating a message, and can be broadly classified as expressive and receptive. The message has the following components: phonology, morphology, syntax, and pragmatics. Phonology refers to the basic perceptual and acoustic units in a word; morphology or semantics, the meanings of words and sentences; syntax or grammar, the rules governing the ordering of the words; and pragmatics, the skills to carry on a two-way conversation and is assessed by change interactions between child and family or therapist.

Children with a hearing loss and a hearing aid have language that falls predominantly outside the normal distribution of equivalent language age for children with good hearing and it is worse for those who are severely to profoundly deaf and need a cochlear implant.

Cochlear implants not only provide information in the low frequencies but also in the mid to high speech frequency ranges not available to hearing-impaired children. Thus, implants should improve the speech and language in severely-to-profoundly deaf children.

Before implanting children in 1985 at the University of Melbourne's Cochlear Implant Clinic at the Royal Victorian Eye and Ear Hospital biological safety was studied on experimental animals and human temporal bones. The efficacy of the speech processing strategies was also established on post-linguistically deaf adults before trialing in children.

The areas of biological safety examined were trauma, biotoxicity, electrical stimuli, and the risk of middle ear infection leading to labyrinthitis and meningitis. The most serious risk is middle ear infection spreading to the inner ear and leading to meningitis. This was a reasonable concern as in the 1960s and 1970s cases of fatal meningitis were reported in patients who had otitis media some time after a stapedectomy [1, 2, 3, 4, 5]. These reports showed that the stapes prosthesis could provide a potential pathway for bacteria to enter the inner ear and spread to the meninges. For example, in temporal bone sections from a report by Rutledge et al. [1], the polyethylene prosthesis acted as a dead space with pus in it, and the infection extended to the inner ear. There was purulent exudate in the cochlear aqueduct demonstrating this was a possible pathway for the infection to the meninges.

Because of these concerns, a series of animal experiments were commenced in the 1970s and 1980s. Firstly, it was found that with spontaneous middle ear infections in the cat, when the electrode entry was unsealed, there could be extensive infection throughout the cochlea [6, 7]. For this reason, from 1983 to 1984, the use of grafts of foreign material to seal around the electrode at the entry point was studied using experimentally induced otitis media in the cat with Streptococcus pyogenes [8]. It was found that dead spaces in the foreign material allowed bacteria to multiply as shown in Fig. 1. Furthermore, the foreign material and dead space assisted the spread of the infection to the inner ear and created a virulent infection, as illustrated in Fig. 2. Note the proximity of the infection to the cochlear aqueduct where the infection was extending to the meninges.

The development of a virulent infection due to a foreign body dead space was consistent with general studies on implanted devices. A study by Zimmerli et al. [9] showed that with the inoculation of Staphylococcus aureus into tissue cages, 100% would become infected for concentrations of only 10³ cfu, but with injection into the subcutaneous tissue at concentrations of 10⁸ cfu, there was no infection. The virulence was in part due to the fact that dead space reduced the bactericidal activity of neutrophils as shown in Fig. 3. The number of killed bacteria was compared with neutrophils taken from a tissue cage or from a sterile infection in the peritoneum or the blood. One explantation was a reduced level of opsonins and complement in the tissue cages [9], and this reduced the body's phagocytic activity. Furthermore, it has also been shown with dead space that the access for antibiotics is significantly reduced.

As a result, natural tissue from the same animal was investigated to create a seal. Fascia and crushed muscle grafts were placed around the electrode some weeks prior to the production of a S. pyogenes otitis media. It was found that the fascia graft facilitated the development of a sheath around the array, and that provided protection against the spread of infection as illustrated in Fig. 4. Crushed muscle was not desirable as it was a nidus for infection, nor was bone paste.

Further studies demonstrated that the sheath around a single-component array enabled three lines of defence to be used against the spread of infection [10, 11, 12]. The first line of defence is the surface activity of mucus secreting cells and their extension around the electrode. The second line of defence is the mobilization of phagocytes in and around the sheath. The third line of defence is the mobilization of type B lymphocytes, killer-type T cells to the sheath, and between the sheath and the electrode [11, 12, 13].

With the first line of defence against the spread of infection from otitis media, the surface cells around the electrode entry changed into mucus secreting cells and extended around and along the electrode array [10]. They produced mucus that is bacteriostatic, and the hairs of the mucous cells could beat to and fro to sweep the bacteria away. Their growth around the electrode is illustrated in Fig. 5. The second line of defence operates when the bacteria release toxins into the sheath. The blood vessels dilate and bring the phagocytes to the site so they can digest the bacteria (Fig. 6).

The third line of defence is the production of type B lymphocytes and type T lymphocytes in response to the bacterial surface antigen. The type B lymphocytes produce antibodies, and the type T lymphocytes are killer cells that pierce the bacteria. Note that in Fig. 7, the lymphocytes not only lie in the connective tissue around the sheath, but they come to lie between the sheath and the array. The defence mechanisms are illustrated in video 1.

The first and second formant/voicing (F0/F1/F2) and first and second formant, and high spectral frequency/voicing ("Multipeak") speech processing strategies were first shown to be effective on post-linguistically deaf adults before they were used on children from 1985 to 1994. With the F0/F1/F2 strategy the first (F1) and second (F2) formant frequencies were extracted with filters with wide bandwidths and presented as place of stimulation to the appropriate frequency site in the cochlea, the fundamental or voicing frequency (F0) was coded as rate of stimulation on each electrode, and the amplitude of F2 as the current level (A2). With "Multipeak," the outputs of fixed filters in three frequency bands (2000–2800 Hz; 2800–4000 Hz; and >4000 Hz) were coded as place of stimulation as well as the first and second formants.

The results of the F0/F1/F2 and "Multipeak" strategies on speech perception in adults for open-sets of word and sentence tests for electrical stimulation alone are shown in Fig. 8 in unselected patients at the University of Melbourne Clinic [14, 15, 16, 17, 18].

The Nucleus receiver-stimulator was modified for use in children, and made without a connector as our research showed that the array could be explanted and another inserted with ease and without trauma [19]. A magnet was incorporated for attaching the transmitting coil.

In 1985, the first child at the University of Melbourne, a 10-year-old, had the multiple-channel implant, and then a 5-year-old followed in 1986 [20, 21, 22]. In 1987, trials commenced in the US, Europe, and Australia for the U.S. Food and Drug Administration (FDA) with the F0/F1/F2, and later the "Multipeak" systems.

The trial showed that 60% of children born deaf were able to understand some open-set speech using electrical stimulation alone, and the majority had significant help in lipreading [23, 24, 25]. The Nucleus F0/F1/F2 and Multipeak speech processing strategies, as well as the receiver-stimulator and banded array, were approved as safe and effective by the U.S. Food and Drug Administration in 1990. The data from the FDA clinical study suggested that improved speech and language would result if the implant was carried out earlier than two years of age. However, before operating on infants and young children, safety studies were essential, as well as efficacy studies of speech strategies in adults.

Special biological safety risks in infants and young children had to be addressed. These were the risk of middle ear infection extending to the inner ear and leading to meningitis, the effect of the implant on head growth, and the effect of electrical stimulus parameters on a maturing nervous system.

Firstly, the effects of Streptococcus pneumoniae rather than Streptococcus pyogenes were studied, as it is a very common pathogen causing otitis media and meningitis in children. S. pneumoniae also has different characteristics, in particular a capsule that prevents it from being engulfed by white cells. It can shed the capsule to invade and release the toxin pneumolysin. The effects of sealing the electrode entry in the presence of otitis media due to S. pneumoniae were studied in cats for (1) unimplanted, (2) implanted but not sealed, and (3) implanted and sealed electrode entries.

It was found that S. pneumoniae produced labyrinthitis in 44% of unimplanted ears. This was consistent with the findings of Myerhoff et al. [26] who found this occurred in 56% of patients. The experimental data indicate that S. pneumoniae has a great propensity to invade the inner ear. In the case of the implanted but unsealed ear, the infection rate was 50%. The reason the incidence was similar to the unimplanted ear was presumed due to the development of a partial seal or sheath in some ears. In the case of the implanted and sealed ear, the infection rate was 6%. The fascia seal facilitated the development of an electrode sheath that helped prevent the entry of infection to the inner ear. A fibrous tissue sheath had formed around the nucleus single-component electrodes in the human temporal bones in the University of Melbourne's temporal bone collection. It should be stressed, however, that a child with the Mondini dysplasia or other abnormality has a higher risk of meningitis whether they are implanted or unimplanted. For example, Schuknecht and Gulya [27] show a large patency between the scala vestibuli and the cerebrospinal fluid (CSF) in the internal auditory meatus in a temporal bone. This type of defect would provide a route for organisms to inoculate the subarachnoid space that has limited defence against invasion. The risk is also seen in a temporal bone of a child with an implant who died from meningitis due to otitis media in the ear opposite to the implant as reported by Suzuki et al. [28]. In fact, it is possible that a sheath and fibrous tissue in the scala tympani due to an implant may possibly protect against meningitis.

The efficacy of the most recent strategies had to be assessed in adults before implanting young children. There were three speech-processing strategies to be evaluated in adults before use in young children. They were the Continuous Interleaved Sampler (CIS), Spectral Maxima Sound Processor (SPEAK), and Advanced Combination Encoder (ACE). CIS samples the outputs of six or more filters to stimulate six to eight electrodes on a place-coding basis non-simultaneously at a constant high rate of 800 pulses/s or above SPEAK extracts six spectral maxima from a bank of 20 filters, and the voltage outputs of the filters stimulate up to 20 appropriately selected electrodes non-simultaneously at a constant rate of 250 pulses/s on a place-coding basis. ACE is a flexible processor that allows the presentation of SPEAK at higher rates of 720 and 1800 pulses/s and for eight and 12 channels. When the electrode outputs of the CIS, SPEAK, and ACE strategies are examined, there is better spectral representation for speech with SPEAK than CIS, but there is greater temporal representation for CIS. On the other hand, ACE combines both the spectral and the temporal information.

Before evaluating the CIS, SPEAK, and ACE speech processing strategies on children, a comparative study was undertaken on adults. This too was preceded by safety studies to ensure that the stimulus parameters used with high rates were also safe [29, 30, 31, 32]. ACE was compared with SPEAK and CIS in a study by Arndt et al. [41]. The rate and number of channels were optimised for ACE and CIS, and were most frequently 720 and 1800 pulses/s for ACE, and 900 and 1800 pulses/s for CIS. Mean HINT sentence scores in quiet were 64.2% for SPEAK, 66.0% for CIS, and 72.3% for ACE. The ACE mean was significantly higher than the CIS mean (p<0.05), but not significantly different from SPEAK (Fig. 9). In a study by Skinner et al. [33], 58% of patients preferred ACE, 25% SPEAK, and 17% CIS. There was also a strong correlation between the preferred strategy and the performance of speech recognition (Fig. 9).

Future advances in speech processing are likely to require a better interface with the auditory nervous system so that more electrodes can be stimulated with a finer temporal and spatial pattern of responses. This could require placing the electrodes closer to the nerve fibers. Consequently, if children are to be able to take advantage of these advances, a peri-modiolar electrode has been designed to lie close to the modiolus, and this also allows the creation of a sheath that can be used by newer electrode designs in the years to come.

To produce a peri-modiolar electrode that would be safe and not cause trauma, a single-component array was developed and evaluated in the human temporal bone in the ARC Human Communication Research Centre. This array had a single component with an inbuilt curl and was held straight until it was inserted into the cochlea before being released. The single-component array lay close to the modiolus within the scala tympani and caused minimal trauma.

The single-component pre-curved array was further developed by the Cooperative Research Centre (CRC) for Cochlear Implant, Speech and Hearing Research and then the CRC for Cochlear Implant and Hearing Aid Innovation. It has a stylet to hold it straight prior to insertion. Once partially inserted along the basal turn of the cochlea, it is slid off the stylet and then passed inwards to curl close to the modiolus. It has been evaluated for safety in a number of temporal bones studies, using fluoroscopy, and histological sections (Fig. 11). It has half band electrodes with and area of approximately 0.3 mm² geometrical compared with the average area of 0.48 mm² geometrical for full bands in a standard electrode array. Due to lower thresholds, the charge density was similar and thus safe electrically.

In addition, explantation and re-implantation studies were undertaken in the experimental animal, and these showed the electrode track created after three months could be re-implanted, as illustrated in the section (Fig. 10) from the cat.

The advantage of the "Contour" electrode is its potential to provide new strategies for children and adults in the future, and also improve results with existing strategies. In order to evaluate advanced strategies on adults, the Nucleus 24 speech processor and 24R receiver-stimulator and "Contour" array have been developed. The Contour provides lower thresholds and comfortable listening levels. It also produced more localized stimulation as shown by the narrower masking curve for the Contour peri-modiolar array (Fig. 12).

Following the safety and efficacy studies on adults, both ACE and the Contour were used in a trial for the FDA on 256 children [34]. It was undertaken on profoundly deaf infants and children from 12 months and above; and on severely deaf children from 25 months and above, with open-set word recognition up to 30%. The results for age appropriate tests are shown in this figure (Fig. 13) and indicate significant improvement. The FDA approved it for use in these groups of children in November 2000.

Not only is it necessary to develop new speech-processing strategies, but it is also essential to know the other factors of importance for improving speech perception as they will not only show how to achieve the best performance with present strategies, but also how to develop improved systems. The general factors leading to good speech perception were analysed from the results on 102 children at the University of Melbourne and Bionic Ear Institute in 2000. The factors contributing to the variance in the study for phoneme scores from open-set monosyllabic words were: the processing strategy (SPEAK compared with MPEAK) accounted for 24% of variance; communication mode (exclusively oral compared with other modes) accounted for 14% of variance; age at implant/duration of deafness accounted for 10% of variance; experience with the implant accounts for 2% of variance; developmental delay accounted for 1% of variance; and late onset (>4 years), 1% of variance. A total of 52% of variance was accounted for in this study.

Other factors for which we have evidence of influence on outcomes are: family/educational support, residual hearing prior to implant, and degree of damage to spiral ganglion cells within the cochlea (inferred but not directly demonstrated).

The variations in performance may also be due to neural connectivity in early development and are reflected in the psychophysical responses in children. This was studied firstly by seeing whether the children's ability to discriminate electrode place of stimulation correlated with age at implantation. The results showed that the discrimination of electrodes was better the younger the child at operation [35, 36]. This suggests that there is a limited time over which the neural connectivity for place discrimination can occur, and this may be important for the development of speech perception.

To also see if there was a correlation between electrode place discrimination and speech perception, a comparison has been made using a closed-set speech test. The findings from the same group of children showed that the smaller the separation between electrodes that was detected, the better the speech perception. This supports the view that if "developmental plasticity" is responsible for creating the neural connections required for place coding, then speech perception will be enhanced as well.

However, when the ability of children to rank pitch tonotopically (i.e. according to place of stimulation) rather than simply discriminate electrode place was compared with their speech perception scores, it was found that not all children who could rank pitch had good speech perception results. For 75% of the 16 children in the study, a tonotopic ordering of pitch percepts was found. However, only 58% of these children with good ability to rank pitch had satisfactory speech perception of 30% or more. This suggests that the effect of "developmental plasticity" on the neural connectivity required for place discrimination is not the only factor for learning speech. At least another factor is required for speech perception, presumably language.

Language and other higher cognitive functions have been assessed to determine their role in speech perception. The close relation of language with speech perception is illustrated in Fig. 14 with data from both implant and hearing aid children over time. The relation is steep up to an equivalent language age of six years. The importance of language for speech perception is illustrated in this figure. In addition, three cochlear implant children were educated in the meaning of words they did not understand, and then tested to see the effects of understanding on speech perception. The mean results show there was a very significant improvement for the unknown words compared with the known words when meaning was provided. Furthermore, when the children were trained in the use of tense and other grammatical concepts, their mean BKB sentence scores increased after the use of tense [37].

In order to better understand the role of top down processing for speech perception and language, a group of children were compared for their cognitive skills using the neuropsychological measures of visual memory, attention, and executive functioning to see whether these skills could account for variance in speech perception, vocabulary, and language abilities. The study showed the children's visual memory skills (i.e. recognition memory, delayed recall, and paired associative learning memory) correlated with their language abilities. There was a positive correlation between visual pattern recognition and memory and language. Contrary to past research, attention skills and executive functioning did not relate to language. These data indicate the importance of the visual system in the presence of auditory deprivation. Deaf children become better able to use visual information in communicating.

To examine in more detail the influence of vision on audition, the McGurk task has been used. The McGurk effect occurs when a person is presented with incongruent visual and auditory stimuli. For example, an auditory /ba/ and a visual /ga/ create a percept of /da/ or /tha/ (voiceless). This is a fusion response to create a different percept. On the other hand, when presented with an auditory /ga/ and a visual /ba/, the sound /bga/ is typically reported. This is a combination response as the two sources have been added together. This is illustrated in video 2.

The response may be due to the importance of information from one modality relative to the ambiguity of the other information source. So, if an adult cannot perceive auditory speech sounds clearly, the sounds are ambiguous, and then information from visual speech cues becomes more important. Thus, a McGurk effect indicates the weighting of auditory and visual stimuli by the individual, relative to the degree of clarity contained within the signal. Fusion responses also suggest that the cortex is able to integrate auditory and visual stimuli, a skill that is possibly under-utilized prior to cochlear implantation.

The McGurk effect has been used by presenting children with an acoustic stimulus where the frequency transition is shifted progressively from low rising to high falling, and then combining this with a lipreading cue [38]. With the auditory cue, the varying second formant frequency shifts the percept from /ba/ to /da/ to /ga/. In the case of normal hearing children, when the rising transition becomes flat, the central brain probability processor is unsure whether the signal is a /ba/ or /da/ and then later as it rises whether it is a /da/ or /ga/ (Fig. 15).

At the bottom left of Fig. 15, the acoustic signals are accompanied by a visual /ba/. The visual /ba/ is a distinct signal, and so, the more indistinct differences in the acoustic cues for /ba/ and /da/ are weighted in favour of /ba/, and the stars show this as statistically significant. On the other hand, on the right with a visual /ga/, the auditory and visual cues are weighted in favour of /da/ as the visual signal for /da/ is distinct from /ba/, but more like the /ga/ that is presented. The stars indicate significantly more /da/ that responses were made by the children when looking at a visual /ga/ face and listening to /ba/. This is evidence of the McGurk fusion effect [38].

The same stimuli were presented to implanted children of the same age and non-verbal intelligence. The results are shown in Fig. 16. Acoustically, there is more confusion between the /da/ and /ga/ sounds, and this is probably due to reduced ability to process these higher frequency transitions. Nevertheless, when the visual /ba/ is given, there are more /ba/ responses because it is more distinct than the auditory signals for /da/ and /ga/. Implanted children's auditory perception of /ba/ improved when matched with the /ba/ lips, as marked by these stars for stimuli 2 and 3. The dominant visual effect appears to be more pronounced than for normal hearing children.

In addition, with the visual /ga/, the central processor had a stronger bias for the visual signal than for sound. With the visual /ga/ condition, there were significantly more /da/ responses to stimuli 1, 2, and 3 suggesting that the children were perceiving the /da/ fusion response, and therefore experienced the McGurk illusion. This suggests that with limited auditory experience, the cortex is able to appropriately integrate auditory and visual signals.

The results have important implications for habilitation. They show first that there is a visual bias in integrating information, and that a higher level function for fusion of auditory and visual information is present in deaf children. These findings suggest that either the visual signal be used to help children with ambiguities, or that speech processing strategies should be used to minimize key ambiguities.

The studies of the McGurk effect indicate the strong bias of the visual signal for implanted children. This can be used positively when the distinct and easy to read "ba" is combined with the auditory signal for "ba," and helpful in training. On the other hand, if the auditory signal for "ba" and "ga" is not strong and unambiguous, then the fusion with the lipreading will not produce an effective result. This means that implant speech processing should make the transitions between, for example, "da" and "ga" more distinct. A speech processing strategy has been developed to do that, and it emphasises the transitions. This transient emphasis speech processing strategy has been evaluated on six patients compared to the standard SPEAK and has shown there is a significant benefit [39, 40].

So in conclusion, the studies with children indicate the importance of emphasising the cues that are difficult to read with the lips, but more importantly producing speech processing strategies that overcome auditory ambiguities such as the transient emphasis speech processor. By clarifying the ambiguities this should alter the effects of visual dominance of auditory areas. It is also consistent with an auditory/verbal method of education.