The Journal of Laryngology and Otology July 1979. Vol. 93. pp. 679-695.
A preliminary report on a multiple-channel cochlear implant operation By Y. C. TONG, R. C. BLACK, G. M. CLARK, I. C. FORSTER, J. B. MILLAR, B. J. O'LOUGHLIN and J. F. PATRICK (Melbourne)
Introduction INTRA-COCHLEAR single-channel electrical stimulation has recently been attempted by Michelson (1971) and by House and Urban (1973). Douek et al. (1977) have described experiments with a single-channel promontory electrode system. It is generally accepted that a single-channel system is useful in conveying crude auditory information such as the presence of sounds and some prosodic features of speech (Bilger et al., 1977; Douek et al, 1977). Bilger et al. (1977) have concluded that single-channel electrical stimuli appear to be analogous to acoustic stimuli varying along a single auditory dimension as subjects seem able to identify members of a small set of five to nine stimuli. This small message set is not adequate for the purpose of speech comprehension. Another approach is to use multiplechannel electrical stimulation systems. The rationale is to stimulate selectively small groups of residual auditory nerve fibres with a number of electrodes so that more information about the input acoustic signals may be conveyed to the higher levels of the auditory system. Furthermore, with appropriate placement of electrodes, the tonotopic organization of the auditory nerve might be utilized, and frequency components of the input acoustical signal could be coded in terms of the place principle of hearing. Multiple-channel electrical stimulation using modiolar electrodes was used by Simmons (1966). Both monopolar and bipolar stimulation was attempted with six implanted modiolar electrodes. The results showed that the pitch sensations produced depended both on the electrode that was activated (described by Simmons as 'characteristic pitch') and the pulse rate of stimulation (described as 'rate pitch'). The subject, however, was unable to identify individual speech sounds. The disadvantage of the modiolar approach is that the electrode array cannot be located with the precision necessary to allow consistent tonotopic stimulation in different patients. An alternative approach is to make use of the tonotopic organization of the auditory nerve fibres in the cochlea by inserting an electrode array into the scala tympani. Multiple-channel intra-cochlear electrode systems were implanted and studied by Mladejovsky et al. (1975) and by Chouard and Macleod (1976). Both groups used monopolar electrodes implanted in the scala tympani, 679
680 TONG, BLACK, CLARK, FORSTER, MILLAR, O'LOUGHLIN AND PATRICK
with the single ground electrode outside the cochlea. The results of the study of Mladejovsky et ah were similar to those of Simmons (1966)— pitch was found to be dependent on electrode placement and rate of stimulation. The usefulness of their system for speech communication has yet to be described. Chouard et ah have reported some improvement in lip-reading and articulation for subjects using their electrode and speech processing systems. However, the results of their rehabilitation programme have not yet been reported. In this paper, we present the results of a series of preliminary experiments conducted with a patient having a multiple-electrode array implanted into the scala tympani (Clark et ah, 1979b). The major difference between the present system and previous intra-cochlear monopolar systems (Mladejovsky et ah, 1975; Chouard and Macleod, 1976) is that an electrode array with inter-leaving ground and active electrodes is used. Electrical stimulation using such an array reduces the spread of electric current in the scala tympani (Black and Clark, 1977; Black, 1978; Black and Clark, 1978), and therefore reduces the size of the overlapping fibre population stimulated by adjacent electrodes. Furthermore, the electrodes are driven by an implanted receiver-stimulator designed to allow experimentation of different stimulus-coding strategies. Patient selection and clinical management
The patient, R.S., who received our first multiple-channel cochlear implant on 1 August 1978 was a 46-year-old male who experienced sudden and total hearing loss on 15 January 1977 following a head injury. He received a severe blow to the back of the head in a car accident, and this was complicated by a left occipital extradural haematoma requiring urgent surgery. Post-operatively he manifested bulbar and facial nerve palsies which gradually improved. Bilateral total hearing loss, tinnitus and loss of balance have persisted since that time. On 17 April 1978 the patient was first referred for evaluation for a multiple-channel cochlear implant and this was carried out according to the guidelines previously discussed (Clark et ah, 1977b). Clinical assessment revealed no previous history of ear disease, and the ears, nose and throat were normal on examination. Apart from essential hypertension there were no other relevant medical findings. Pure-tone and speech audiometry showed no hearing in either ear, and consequently other audiological and psycho-acoustic tests could not be performed. An impedance audiogram demonstrated normal middle ear and Eustachian tube function. The x-rays confirmed the absence of middle ear or mastoid disease, but demonstrated a small crack fracture entering the left superior semicircular canal. Electronystagmography showed no vestibular function on either side. Electrical stimulation of the promontory was performed on both ears
REPORT ON A COCHLEAR IMPLANT OPERATION
681
under local anaesthetic. Stimulation was carried out with trains of monophasic rectangular current pulses with a pulse width of 200 jxs and a duration of 300 ms. These pulse trains were presented at a rate of 1-2/s. Ascending and descending threshold currents were determined for both ears. The patient reported on the left side that he had tonal sensations from 50 pps to 200 pps. The perception of pitch as high or low did not show a significant trend with increasingly higher rates of stimulation. On the right side the patient did not initially describe the electrical stimuli as tonal, but commented on them as a 'dash—dash', 'click—click' sensation. With an increase in the rate of stimulation, however, there was an increase in the pitch perceived. In addition the patient also reported that his tinnitus was masked by electrical stimulation of the left ear but not the right. As the x-rays did not show a fracture in the region of the internal auditory meati but a small one entering the left superior semicircular canal, and as electrical stimulation of the promontory revealed that the patient could experience hearing sensations, it was decided that the injury had not caused transection of the auditory nerves but concussion of the cochlea. For this reason a multiple-channel cochlear implant was recommended, and the left ear was selected as electrical stimulation of the promontory had produced more tonal sensations on this side. A further advantage in selecting this patient was his musical appreciation and previous experience in singing in choirs. After the risks and expectations of a multiple-channel cochlear implant were explained, the patient was assessed by an Ethics Committee to ensure that all the risks had been adequately explained and no coercion used. This Ethics Committee consisted of one member representing the implant team, an independent otolaryngologist, a clinical psychologist, a practising lawyer and a mother of a deaf child. On 1 August 1978 a multiple-channel cochlear implant operation was performed on the patient. A ten-channel electrode array was inserted through an opening in the round window membrane for a distance of 25 mm around the scala tympani. The receiver-stimulator unit was positioned in a bed created in the mastoid bone and held in place by a silastic mould, and superficial and deep fascial flaps (Clark et ah, 1979b). The patient made a good recovery and was ambulant in three days, and discharged from hospital in two weeks. Wound healing was satisfactory, and the first electrical stimulation test was performed three weeks after surgery. A subsequent x-ray of the temporal bone, to evaluate the electrode placement, snowed that the proximal or most basal electrode was opposite the round window region. This would indicate that the electrode array had shifted about 10 mm, and emphasizes the need for adequate extra-cochlear fixation. Electrode array
The multiple-channel cochlear prosthesis requires an electrode system which produces a stimulus current spatially localized to each electrode. A
682 TONG, BLACK, CLARK, FORSTER, MILLAR, O'LOUGHLIN AND PATRICK
bipolar electrode pair placed in the scala tympani produces significantly less spread of current than with monopolar stimulation of the scala tympani in which the earth electrode is outside the cochlea (Schindler et ah, 1977; Black and Clark, 1979). The electrode array used in the present implant does not consist of discrete bipolar pairs, but uses an extended earth system which is common to all stimulus channels. It consists of twenty individual platinum foil bands 0-3 mm wide and spaced 0-75 mm apart on a silicon rubber tube of outside diameter 0-4 mm at the tip, increasing to 0-64 mm diameter at the point 8 mm from the tip. Each band is connected to the receiver-stimulator by a 0025 mm teflon-coated wire running inside the silicon rubber tube. Ten alternate bands are connected to form the extended earth system, and the remaining bands serve as the ten active electrodes. Experiments have been performed with such electrodes implanted into the scala tympani of cats. Measurements of the ground current in each earth band electrode during stimulation of the active electrodes showed that even for such an extended ground system, the current attenuation was in the range 2-4-5 dB/mm. This compares with an attenuation in the range of 0-5-10 dB/mm for monopolar stimulation of the scala tympani determined from nerve excitability studies (Schindler et ah, 1977; Black and Clark, 1979). System hardware
The prosthesis system acts as a multiple-channel transcutaneous stimulation system (Forster, 1978; Forster and Patrick, 1978), and may be considered to comprise three components as shown in block diagram form in Fig. 1. The stimulus processor and transmitter stages are both located external to the patient, while the receiver-stimulator was located in a surgically-created bed in the mastoid bone. The stimulus pattern to be delivered to the intracochlear electrode array described above, is specified by an HP2100 computer system. The External
components
Implanted components /
STIMULUS PROCESSOR
RECEIVER
TRANSMITTER
jC t
(HP2100)
I Input
f /, /e/, /I/) as reported in single-electrode experiments using 200 ms pulse trains, the hearing sensations evoked by several of the two-electrode combinations were identified as /A/ (as in hwt). One of the members of these electrode-pairs was invariably electrode 9 which possessed the /b/ colour when tested as a single electrode. This result suggested that simultaneous activation of two electrodes could produce a percept which was different from those observed when the electrodes were activated in isolation. Furthermore, an 'averaging' process as discussed earlier could again be adopted by the patient. Further experimental work is required to determine the extent to which this result could be applied in stimulus coding strategies. (ii) Stimulation using different pulse rates Two electrodes (7 and 9) which have similar loudness growth characteristics and thresholds, when tested as single electrodes, were activated with similar currents but at different rates. Electrodes 7 and 9 were activated at rates of 150 pps and 140 pps respectively, and stimulus currents were adjusted to provide a subjective judgment of medium loudness. The patient's response was that it sounded like a repeated consonantvowel (CV) sequence /da, da, da/ (/a/ as in about). Three repetitions were reported for a 300 msec stimulus, two repetitions for a 200 msec stimulus, one repetition for a 100 msec stimulus and still one repetition for a 50 msec stimulus. The continuous stimulus may be conceptually divided into 50 msec segments which are alternately characterized by the two pulse trains moving out of phase, into phase, out of phase, etc. The full cycle of moving out of phase and into phase (100 msec) produced a response of a single CV (/da/), and multiples of such a cycle produced a corresponding number of CV percepts to the total stimulus. When both parts of the phase cycle were presented in isolation as 50 msec segments the same percept was reported, but reduced loudness was reported for the 'moving into phase' half cycle. Equivalent results were obtained when electrodes 7 and 9 were activated at rates of 100 pps and 90 pps respectively. However, when the rate differential was changed to 5 pps, a 200 msec stimulus produced a single CV response instead of two CV responses as obtained with a 10 pps rate differential. The most significant result from this study was that repeated syllables could be cued by differential rates of stimulation applied to two electrodes, the perceived repetition rate being related to the difference between stimulus rates. Further work is required to elucidate the mechanism responsible for the CV percepts, but the present evidence suggests that it could be due to loudness variations. , , , Discussion ,.s n. , (A) Pitch ana spectral colour The results have shown that the pitch estimated in the pitch scaling procedure and the spectral colour determined in the identification experi-
6 9 2 TONG, BLACK, CLARK, FORSTER, MILLAR, O'LOUGHLIN AND PATRICK
ments are functions of pulse rate and electrode position among other physical parameters. Furthermore, pitch and spectral colour are somehow related. For example, the variation of pitch estimates at 50 pps (Fig. 4) follows the same trend as the variation of spectral colours identified with 75 pps pulse trains. The 'low frequency' /b/, /:>/ and /E/ colours at electrodes 1, 2, 8 and 9 correspond to relatively low pitch estimates, while higher estimates were obtained with electrodes 3 to 7 which possessed the 'high frequency' /I/ and /i/ colour. Secondly, the 'lower frequency' /D/ colour at electrode 9 could become a 'higher frequency' /a/ colour with an increase in pitch following an increase in pulse rate. In short, these two examples demonstrated that pitch appears to vary monotonically with spectral colour. The results of the pitch-scaling experiment are likely to be influenced by the patient's concept of pitch, and the unnaturalness of the electrically evoked hearing sensations. As pointed out by Evans (1978), 'pitch' is not an easily definable concept in psychoacoustic experiment and theory. A listener can hear the pitches of the individual components of an acoustical signal (analytical mode), or hear a unified pitch of a complex signal as a single 'gestalt' (synthetic mode). In addition, the reported pitch percept depends on the quality of the acoustical signal. In the pitch scaling experiments, it was assumed that the synthetic mode was in operation, and that for an electrical stimulus the patient could hear a unified 'pitch' and could assign a number indicating the position of this 'pitch' on a monotonic scale. However, because of the unnaturalness of the hearing sensations, the analytical mode might also be involved. The hearing sensation could very well be interpreted as having two pitch components: one related to the pulse rate, and the other related to the placement of the electrodes. The pitch was then estimated as a weighted sum of the two component pitches, based on the patient's concept of pitch. In the same vein, one might speculate that an analysis-synthesis process in terms of a weighted sum of two component colours could be involved in the identification of spectral colours. Another possible interpretation of these results is that pulse rate is responsible for a pitch percept similar to the 'low pitch' percept reported in psychoacoustical experiments (Plomp, 1976), while electrode location is responsible for the perception of the spectral colours. In this case the variation of spectral colours with pulse rate and the variation of pitch with electrode location would be simply a result of semantic confusion— confusion between pitch and spectral colour by the patient. (B) Physiological considerations The spectral colours associated with single electrodes 1 to 7 are consistent with the tonotopical organization of the cochlea. A 'lower frequency' /e/ colour was identified at the two most apical electrodes (1 and 2), while 'higher frequency' /I/ and /i/ colours were identified at the more basal
REPORT ON A COCHLEAR IMPLANT OPERATION
693
electrodes (3-7). As described earlier, a single-formant synthetic vowel with its formant frequency at 2,160 Hz was identified as having a vowel sound similar to /e/. This frequency of 2,160 Hz is in good agreement with the characteristic frequency of the cochlear place where electrode 1 is located. For an average human cochlea, the characteristic frequency of the cochlear place, 15 mm from the round window, is approximately 2,200 Hz (Bekesy, 1960). Furthermore, the characteristic frequencies of the cochlear places for electrodes 3 to 7 are above 3,000 Hz, and Delattre et al. (1952) have shown that single-formant synthetic vowels with formant frequency above 3,000 Hz were always identified as /i/. For electrodes 8 and 9 which are in the proximity to the round window, the electric current distribution produced by these two single electrodes is likely to be quite complex, and extra-cochlear electric current paths could be involved. The /b/, joj and /e/ colours associated with these two electrodes could be produced by stimulating nerve fibres innervating areas of the cochlea with characteristic frequencies of approximately 2,160 Hz and 720 Hz, or nerve fibres with these characteristic frequencies located in the internal auditory meatus. An interpretation of the variation of pitch and spectral colour with pulse rate is more difficult at a neuronal level. This is because synchronous firing of nerve fibres at predetermined intervals in electrical stimulation is not normally observed in acoustical stimulation. For example, in a population of neurons innervating a small region of the cochlear partition vibrating at low to medium frequencies ( < 5 kHz), at any instant a random subset of neurons out of the population is activated. The size of the subset is a function of the instantaneous displacement of the cochlear partition in the direction of the scala vestibuli. In other words, neural discharges are temporally dispersed over the 'vestibular half period' of vibration of the partition. For electrical stimulation, a whole population of neurons is synchronously activated in response to a supra-threshold current pulse. The temporally dispersed discharge pattern in response to a 100 Hz pure tone is therefore not reproduced by a 100 pps electrical pulse train. Another difficulty involves the range of pulse rates employed in these preliminary experiments. The discharge rate of an auditory neurone under acoustical stimulation normally saturates at approximately 150 discharges per second (Evans, 1975), although it is capable of discharging at up to 900 discharges per second under electrical stimulation (Moxon, 1971). Any interpretation of the effect of pulse rates on the hearing sensations produced by electrical stimuli at pulse rates greater than 150 pps should be treated with caution. For these reasons, the mechanism that is responsible for the effects of pulse rate on the electrically evoked sensations described in this paper remains obscure. An elucidation of the process involved awaits further understanding in psychoacoustics, normal hearing physiology and electrical stimulation in humans and animals. An excellent review of some of the
694 TONG, BLACK, CLARK, FORSTER, MILLAR, o'LOUGHLIN AND PATRICK
research currently being undertaken can be found in Ballantyne et al. (1978). (C) Speech processing considerations The results obtained in these preliminary experiments are both informative and encouraging. The reported spectral colours confirm some of the advantages of multiple-channel over single channel systems, and these colours are useful in conveying speech information. Our observations have indicated that although electrode placement is the major factor governing the perception of these colours, pulse rate and current level also influence these percepts. The results from the pitch-scaling experiment have shown that a significant range of pitch percepts can be evoked by different stimulus configurations. The pitch percepts obtained with a fixed current level principally depend on pulse rate, but are also influenced by electrode placement. Another percept which is important to speech perception is loudness. Our observations have indicated that loudness depends on both current level and pulse rate. Furthermore, simultaneous activation of two electrodes with the same pulse rate can produce a spectral colour corresponding to a 'weighted average' of the spectral colours reported when the electrodes were activated in isolation. In the light of these preliminary results, a possible speech coding scheme may be proposed. In this scheme, speech signal parameters such as fundamental frequency, formant frequencies and the amplitude envelope of the overall speech signal, would be measured by a pre-processing unit following the same principles as a formant-vocoder (Flanagan, 1972). These speech signal parameters would be then transformed to electrical stimulus parameters. As an initial attempt, formant frequencies could be transformed to a pattern of activated electrodes, fundamental frequency to pulse rate (the same for all activated electrodes), and amplitude envelope to current level. With this scheme, it is assumed that spectral colours can be cued by the electrodes that are activated, pitch by stimulus pulse rate and loudness by current level. This proposal takes account of the major factors (electrode placement and pulse rate) influencing the perception of spectral colour and pitch, and from our observation the effect of current level on spectral colour and pitch appears to be secondary. Finally, it should be noted that the proposed scheme is based on results obtained with periodic pulse trains, and therefore strictly speaking is only applicable to quasi-periodic hearing sensations such as vowel sounds. Further experimentation is required to determine whether this proposal is adequate to convey usable information on the transitional characteristics of speech, and on other categories of speech sound such as the unvoiced fricatives. Acknowledgements
We would like to acknowledge the financial support obtained from the National Health and Medical Research Council of Australia, the Channel
REPORT ON A COCHLEAR IMPLANT OPERATION
695
O Nerve Deafness Appeal and the Clive and Vera Ramaciotti Foundation. We would like to thank Mr. R. J. Walkerden and Mr. L. Cole for technical support and Miss J. Maher for the typing. REFERENCES AINSWORTH, W. A. (1976) Mechanisms of Speech Recognition. Pergamon Press, Oxford, p. 42. BALLANTYNE, J. C , EVANS, E. F., and MORRISON, A. W. (1978) Journal of Laryngology and
Otology, Supplement, 1. BEKESY, G. VON (1960) Experiments in Hearing, McGraw-Hill, New York. p. 442. BERNARD, J. R. L. (1970) Zeitschrift fur Phonetik, 23, 113. BILGER, R. C , BLACK, F. O., MYERS, G. K , HOPKINSON, N. T., VEGA, A., and WOLF, R. V.
(1977) Annals of Otology, Rhinology and Laryngology, 86, Supplement, 38, 1. BLACK, R. C. (1978) Ph.D. Thesis, University of Melbourne. BLACK, R. C , and CLARK, G. M. (1977) University of Melbourne Research Report, 140. BLACK, R. C , and CLARK, G. M. (1978) Proceedings of the Australian Physiological and Pharmacological Society, 9, 7IP. BLACK, R. C , and CLARK, G. M. (1979) Submitted for publication. CHOUARD, C. H., and MACLEOD, P. (1976) Laryngoscope, 86, 1743. CLARK, G. M., BLACK, R., DEWHURST, D. J., FORSTER, I. C , PATRICK, J. F., and TONG, Y. C.
(1977a) Medical Progress through Technology, 5, 127. CLARK, G. M., O'LOUGHLIN, B. J., RICKARDS, F. W., TONG, Y. C , and WILLIAMS, A. J. (1977b)
Journal of Laryngology and Otology, 91, 697. CLARK, G. M., TONG, Y. C , BLACK, R., FORSTER, I. C , PATRICK, J. F., and DEWHURST, D. J.
(1977c) Journal of Laryngology and Otology, 91, 935. CLARK, G. M., BLACK, R., FORSTER, I. C , PATRICK, J. F., and TONG, Y. C. (1978) Journal of
the Acoustical Society of America, 63, 631. CLARK, G. M., PATRICK, J. F., and BAILEY, Q. (1979a) Journal ofLaryngology and Otology, 93,
107. CLARK, G. M., PYMAN, B. C , and BAILEY, Q. R. (1979b) Journal of Laryngology and Otology,
93,215. DELATTRE, P. C , LIBERMAN, A. M., COOPER, F. S., and GENSTMAN, L. J. (1952) Word, 8, 195. DOUEK, E. E., FOURCIN, A. J., MOORE, B. C. J., and CLARKE, G. P. (1977) Proceedings of the
Royal Society of Medicine, 70, 379. EVANS, E. F. (1975) Cochlear Nerve and Cochlear Nucleus. In: Handbook of Sensory Physiology. V/2. Eds. W. D. Keidel and W. D. Neff, pp. 1-108. Springer-Verlag, Berlin. EVANS, E. F. (1978) Audiology, 17, 369. FLANAGAN, J. L. (1972) Speech Analysis, Synthesis and Perception, Springer, Berlin. FORSTER, I. C. (1978) Ph.D. Thesis. University of Melbourne. FORSTER, I. C , and PATRICK, J. F. (1978) 18th Australasian Conference on Physical Sciences and Engineering in Medicine and Biology. HOUSE, W. F., and URBAN, J. (1973) Annals of Otology, Rhinology and Laryngology, 82, 504. MICHELSON, R. P. (1971) Archives of Otolaryngology, 93, 317. MLADEJOVSKY, M. G., EDDINGTON, D. K., DOBELLE, W. H., and BRACKMAN, D. E. (1975)
Transactions of the American Society for Artificial Internal Organs, 21, 1. MOXON, E. C. (1971) Ph.D. Thesis. Massachusetts Institute of Technology. PLOMP, R. (1976) Aspects of Tone Sensation. Academic Press, London, p. 109. ScinNDLER, R. A., MERZENICH, M. M., WHITE, M. W., and BJORKROTH, B. (1977) Archives of
Otolaryngology, 103, 691-699. SIMMONS, F. B. (1966) Archives of Otolaryngology, 84, 24. STEVENS, S. S. (1975) Psychophysics. John Wiley & Sons, New York. TEGHTSOONIAN, R. (1971) Psychological Review, 78, 71.
A preliminary report on a multiple-channel cochlear implant operation By Y. C. TONG, R. C. BLACK, G. M. CLARK, I. C. FORSTER, J. B. MILLAR, B. J. O'LOUGHLIN and J. F. PATRICK (Melbourne)
Introduction INTRA-COCHLEAR single-channel electrical stimulation has recently been attempted by Michelson (1971) and by House and Urban (1973). Douek et al. (1977) have described experiments with a single-channel promontory electrode system. It is generally accepted that a single-channel system is useful in conveying crude auditory information such as the presence of sounds and some prosodic features of speech (Bilger et al., 1977; Douek et al, 1977). Bilger et al. (1977) have concluded that single-channel electrical stimuli appear to be analogous to acoustic stimuli varying along a single auditory dimension as subjects seem able to identify members of a small set of five to nine stimuli. This small message set is not adequate for the purpose of speech comprehension. Another approach is to use multiplechannel electrical stimulation systems. The rationale is to stimulate selectively small groups of residual auditory nerve fibres with a number of electrodes so that more information about the input acoustic signals may be conveyed to the higher levels of the auditory system. Furthermore, with appropriate placement of electrodes, the tonotopic organization of the auditory nerve might be utilized, and frequency components of the input acoustical signal could be coded in terms of the place principle of hearing. Multiple-channel electrical stimulation using modiolar electrodes was used by Simmons (1966). Both monopolar and bipolar stimulation was attempted with six implanted modiolar electrodes. The results showed that the pitch sensations produced depended both on the electrode that was activated (described by Simmons as 'characteristic pitch') and the pulse rate of stimulation (described as 'rate pitch'). The subject, however, was unable to identify individual speech sounds. The disadvantage of the modiolar approach is that the electrode array cannot be located with the precision necessary to allow consistent tonotopic stimulation in different patients. An alternative approach is to make use of the tonotopic organization of the auditory nerve fibres in the cochlea by inserting an electrode array into the scala tympani. Multiple-channel intra-cochlear electrode systems were implanted and studied by Mladejovsky et al. (1975) and by Chouard and Macleod (1976). Both groups used monopolar electrodes implanted in the scala tympani, 679
680 TONG, BLACK, CLARK, FORSTER, MILLAR, O'LOUGHLIN AND PATRICK
with the single ground electrode outside the cochlea. The results of the study of Mladejovsky et ah were similar to those of Simmons (1966)— pitch was found to be dependent on electrode placement and rate of stimulation. The usefulness of their system for speech communication has yet to be described. Chouard et ah have reported some improvement in lip-reading and articulation for subjects using their electrode and speech processing systems. However, the results of their rehabilitation programme have not yet been reported. In this paper, we present the results of a series of preliminary experiments conducted with a patient having a multiple-electrode array implanted into the scala tympani (Clark et ah, 1979b). The major difference between the present system and previous intra-cochlear monopolar systems (Mladejovsky et ah, 1975; Chouard and Macleod, 1976) is that an electrode array with inter-leaving ground and active electrodes is used. Electrical stimulation using such an array reduces the spread of electric current in the scala tympani (Black and Clark, 1977; Black, 1978; Black and Clark, 1978), and therefore reduces the size of the overlapping fibre population stimulated by adjacent electrodes. Furthermore, the electrodes are driven by an implanted receiver-stimulator designed to allow experimentation of different stimulus-coding strategies. Patient selection and clinical management
The patient, R.S., who received our first multiple-channel cochlear implant on 1 August 1978 was a 46-year-old male who experienced sudden and total hearing loss on 15 January 1977 following a head injury. He received a severe blow to the back of the head in a car accident, and this was complicated by a left occipital extradural haematoma requiring urgent surgery. Post-operatively he manifested bulbar and facial nerve palsies which gradually improved. Bilateral total hearing loss, tinnitus and loss of balance have persisted since that time. On 17 April 1978 the patient was first referred for evaluation for a multiple-channel cochlear implant and this was carried out according to the guidelines previously discussed (Clark et ah, 1977b). Clinical assessment revealed no previous history of ear disease, and the ears, nose and throat were normal on examination. Apart from essential hypertension there were no other relevant medical findings. Pure-tone and speech audiometry showed no hearing in either ear, and consequently other audiological and psycho-acoustic tests could not be performed. An impedance audiogram demonstrated normal middle ear and Eustachian tube function. The x-rays confirmed the absence of middle ear or mastoid disease, but demonstrated a small crack fracture entering the left superior semicircular canal. Electronystagmography showed no vestibular function on either side. Electrical stimulation of the promontory was performed on both ears
REPORT ON A COCHLEAR IMPLANT OPERATION
681
under local anaesthetic. Stimulation was carried out with trains of monophasic rectangular current pulses with a pulse width of 200 jxs and a duration of 300 ms. These pulse trains were presented at a rate of 1-2/s. Ascending and descending threshold currents were determined for both ears. The patient reported on the left side that he had tonal sensations from 50 pps to 200 pps. The perception of pitch as high or low did not show a significant trend with increasingly higher rates of stimulation. On the right side the patient did not initially describe the electrical stimuli as tonal, but commented on them as a 'dash—dash', 'click—click' sensation. With an increase in the rate of stimulation, however, there was an increase in the pitch perceived. In addition the patient also reported that his tinnitus was masked by electrical stimulation of the left ear but not the right. As the x-rays did not show a fracture in the region of the internal auditory meati but a small one entering the left superior semicircular canal, and as electrical stimulation of the promontory revealed that the patient could experience hearing sensations, it was decided that the injury had not caused transection of the auditory nerves but concussion of the cochlea. For this reason a multiple-channel cochlear implant was recommended, and the left ear was selected as electrical stimulation of the promontory had produced more tonal sensations on this side. A further advantage in selecting this patient was his musical appreciation and previous experience in singing in choirs. After the risks and expectations of a multiple-channel cochlear implant were explained, the patient was assessed by an Ethics Committee to ensure that all the risks had been adequately explained and no coercion used. This Ethics Committee consisted of one member representing the implant team, an independent otolaryngologist, a clinical psychologist, a practising lawyer and a mother of a deaf child. On 1 August 1978 a multiple-channel cochlear implant operation was performed on the patient. A ten-channel electrode array was inserted through an opening in the round window membrane for a distance of 25 mm around the scala tympani. The receiver-stimulator unit was positioned in a bed created in the mastoid bone and held in place by a silastic mould, and superficial and deep fascial flaps (Clark et ah, 1979b). The patient made a good recovery and was ambulant in three days, and discharged from hospital in two weeks. Wound healing was satisfactory, and the first electrical stimulation test was performed three weeks after surgery. A subsequent x-ray of the temporal bone, to evaluate the electrode placement, snowed that the proximal or most basal electrode was opposite the round window region. This would indicate that the electrode array had shifted about 10 mm, and emphasizes the need for adequate extra-cochlear fixation. Electrode array
The multiple-channel cochlear prosthesis requires an electrode system which produces a stimulus current spatially localized to each electrode. A
682 TONG, BLACK, CLARK, FORSTER, MILLAR, O'LOUGHLIN AND PATRICK
bipolar electrode pair placed in the scala tympani produces significantly less spread of current than with monopolar stimulation of the scala tympani in which the earth electrode is outside the cochlea (Schindler et ah, 1977; Black and Clark, 1979). The electrode array used in the present implant does not consist of discrete bipolar pairs, but uses an extended earth system which is common to all stimulus channels. It consists of twenty individual platinum foil bands 0-3 mm wide and spaced 0-75 mm apart on a silicon rubber tube of outside diameter 0-4 mm at the tip, increasing to 0-64 mm diameter at the point 8 mm from the tip. Each band is connected to the receiver-stimulator by a 0025 mm teflon-coated wire running inside the silicon rubber tube. Ten alternate bands are connected to form the extended earth system, and the remaining bands serve as the ten active electrodes. Experiments have been performed with such electrodes implanted into the scala tympani of cats. Measurements of the ground current in each earth band electrode during stimulation of the active electrodes showed that even for such an extended ground system, the current attenuation was in the range 2-4-5 dB/mm. This compares with an attenuation in the range of 0-5-10 dB/mm for monopolar stimulation of the scala tympani determined from nerve excitability studies (Schindler et ah, 1977; Black and Clark, 1979). System hardware
The prosthesis system acts as a multiple-channel transcutaneous stimulation system (Forster, 1978; Forster and Patrick, 1978), and may be considered to comprise three components as shown in block diagram form in Fig. 1. The stimulus processor and transmitter stages are both located external to the patient, while the receiver-stimulator was located in a surgically-created bed in the mastoid bone. The stimulus pattern to be delivered to the intracochlear electrode array described above, is specified by an HP2100 computer system. The External
components
Implanted components /
STIMULUS PROCESSOR
RECEIVER
TRANSMITTER
jC t
(HP2100)
I Input
f /, /e/, /I/) as reported in single-electrode experiments using 200 ms pulse trains, the hearing sensations evoked by several of the two-electrode combinations were identified as /A/ (as in hwt). One of the members of these electrode-pairs was invariably electrode 9 which possessed the /b/ colour when tested as a single electrode. This result suggested that simultaneous activation of two electrodes could produce a percept which was different from those observed when the electrodes were activated in isolation. Furthermore, an 'averaging' process as discussed earlier could again be adopted by the patient. Further experimental work is required to determine the extent to which this result could be applied in stimulus coding strategies. (ii) Stimulation using different pulse rates Two electrodes (7 and 9) which have similar loudness growth characteristics and thresholds, when tested as single electrodes, were activated with similar currents but at different rates. Electrodes 7 and 9 were activated at rates of 150 pps and 140 pps respectively, and stimulus currents were adjusted to provide a subjective judgment of medium loudness. The patient's response was that it sounded like a repeated consonantvowel (CV) sequence /da, da, da/ (/a/ as in about). Three repetitions were reported for a 300 msec stimulus, two repetitions for a 200 msec stimulus, one repetition for a 100 msec stimulus and still one repetition for a 50 msec stimulus. The continuous stimulus may be conceptually divided into 50 msec segments which are alternately characterized by the two pulse trains moving out of phase, into phase, out of phase, etc. The full cycle of moving out of phase and into phase (100 msec) produced a response of a single CV (/da/), and multiples of such a cycle produced a corresponding number of CV percepts to the total stimulus. When both parts of the phase cycle were presented in isolation as 50 msec segments the same percept was reported, but reduced loudness was reported for the 'moving into phase' half cycle. Equivalent results were obtained when electrodes 7 and 9 were activated at rates of 100 pps and 90 pps respectively. However, when the rate differential was changed to 5 pps, a 200 msec stimulus produced a single CV response instead of two CV responses as obtained with a 10 pps rate differential. The most significant result from this study was that repeated syllables could be cued by differential rates of stimulation applied to two electrodes, the perceived repetition rate being related to the difference between stimulus rates. Further work is required to elucidate the mechanism responsible for the CV percepts, but the present evidence suggests that it could be due to loudness variations. , , , Discussion ,.s n. , (A) Pitch ana spectral colour The results have shown that the pitch estimated in the pitch scaling procedure and the spectral colour determined in the identification experi-
6 9 2 TONG, BLACK, CLARK, FORSTER, MILLAR, O'LOUGHLIN AND PATRICK
ments are functions of pulse rate and electrode position among other physical parameters. Furthermore, pitch and spectral colour are somehow related. For example, the variation of pitch estimates at 50 pps (Fig. 4) follows the same trend as the variation of spectral colours identified with 75 pps pulse trains. The 'low frequency' /b/, /:>/ and /E/ colours at electrodes 1, 2, 8 and 9 correspond to relatively low pitch estimates, while higher estimates were obtained with electrodes 3 to 7 which possessed the 'high frequency' /I/ and /i/ colour. Secondly, the 'lower frequency' /D/ colour at electrode 9 could become a 'higher frequency' /a/ colour with an increase in pitch following an increase in pulse rate. In short, these two examples demonstrated that pitch appears to vary monotonically with spectral colour. The results of the pitch-scaling experiment are likely to be influenced by the patient's concept of pitch, and the unnaturalness of the electrically evoked hearing sensations. As pointed out by Evans (1978), 'pitch' is not an easily definable concept in psychoacoustic experiment and theory. A listener can hear the pitches of the individual components of an acoustical signal (analytical mode), or hear a unified pitch of a complex signal as a single 'gestalt' (synthetic mode). In addition, the reported pitch percept depends on the quality of the acoustical signal. In the pitch scaling experiments, it was assumed that the synthetic mode was in operation, and that for an electrical stimulus the patient could hear a unified 'pitch' and could assign a number indicating the position of this 'pitch' on a monotonic scale. However, because of the unnaturalness of the hearing sensations, the analytical mode might also be involved. The hearing sensation could very well be interpreted as having two pitch components: one related to the pulse rate, and the other related to the placement of the electrodes. The pitch was then estimated as a weighted sum of the two component pitches, based on the patient's concept of pitch. In the same vein, one might speculate that an analysis-synthesis process in terms of a weighted sum of two component colours could be involved in the identification of spectral colours. Another possible interpretation of these results is that pulse rate is responsible for a pitch percept similar to the 'low pitch' percept reported in psychoacoustical experiments (Plomp, 1976), while electrode location is responsible for the perception of the spectral colours. In this case the variation of spectral colours with pulse rate and the variation of pitch with electrode location would be simply a result of semantic confusion— confusion between pitch and spectral colour by the patient. (B) Physiological considerations The spectral colours associated with single electrodes 1 to 7 are consistent with the tonotopical organization of the cochlea. A 'lower frequency' /e/ colour was identified at the two most apical electrodes (1 and 2), while 'higher frequency' /I/ and /i/ colours were identified at the more basal
REPORT ON A COCHLEAR IMPLANT OPERATION
693
electrodes (3-7). As described earlier, a single-formant synthetic vowel with its formant frequency at 2,160 Hz was identified as having a vowel sound similar to /e/. This frequency of 2,160 Hz is in good agreement with the characteristic frequency of the cochlear place where electrode 1 is located. For an average human cochlea, the characteristic frequency of the cochlear place, 15 mm from the round window, is approximately 2,200 Hz (Bekesy, 1960). Furthermore, the characteristic frequencies of the cochlear places for electrodes 3 to 7 are above 3,000 Hz, and Delattre et al. (1952) have shown that single-formant synthetic vowels with formant frequency above 3,000 Hz were always identified as /i/. For electrodes 8 and 9 which are in the proximity to the round window, the electric current distribution produced by these two single electrodes is likely to be quite complex, and extra-cochlear electric current paths could be involved. The /b/, joj and /e/ colours associated with these two electrodes could be produced by stimulating nerve fibres innervating areas of the cochlea with characteristic frequencies of approximately 2,160 Hz and 720 Hz, or nerve fibres with these characteristic frequencies located in the internal auditory meatus. An interpretation of the variation of pitch and spectral colour with pulse rate is more difficult at a neuronal level. This is because synchronous firing of nerve fibres at predetermined intervals in electrical stimulation is not normally observed in acoustical stimulation. For example, in a population of neurons innervating a small region of the cochlear partition vibrating at low to medium frequencies ( < 5 kHz), at any instant a random subset of neurons out of the population is activated. The size of the subset is a function of the instantaneous displacement of the cochlear partition in the direction of the scala vestibuli. In other words, neural discharges are temporally dispersed over the 'vestibular half period' of vibration of the partition. For electrical stimulation, a whole population of neurons is synchronously activated in response to a supra-threshold current pulse. The temporally dispersed discharge pattern in response to a 100 Hz pure tone is therefore not reproduced by a 100 pps electrical pulse train. Another difficulty involves the range of pulse rates employed in these preliminary experiments. The discharge rate of an auditory neurone under acoustical stimulation normally saturates at approximately 150 discharges per second (Evans, 1975), although it is capable of discharging at up to 900 discharges per second under electrical stimulation (Moxon, 1971). Any interpretation of the effect of pulse rates on the hearing sensations produced by electrical stimuli at pulse rates greater than 150 pps should be treated with caution. For these reasons, the mechanism that is responsible for the effects of pulse rate on the electrically evoked sensations described in this paper remains obscure. An elucidation of the process involved awaits further understanding in psychoacoustics, normal hearing physiology and electrical stimulation in humans and animals. An excellent review of some of the
694 TONG, BLACK, CLARK, FORSTER, MILLAR, o'LOUGHLIN AND PATRICK
research currently being undertaken can be found in Ballantyne et al. (1978). (C) Speech processing considerations The results obtained in these preliminary experiments are both informative and encouraging. The reported spectral colours confirm some of the advantages of multiple-channel over single channel systems, and these colours are useful in conveying speech information. Our observations have indicated that although electrode placement is the major factor governing the perception of these colours, pulse rate and current level also influence these percepts. The results from the pitch-scaling experiment have shown that a significant range of pitch percepts can be evoked by different stimulus configurations. The pitch percepts obtained with a fixed current level principally depend on pulse rate, but are also influenced by electrode placement. Another percept which is important to speech perception is loudness. Our observations have indicated that loudness depends on both current level and pulse rate. Furthermore, simultaneous activation of two electrodes with the same pulse rate can produce a spectral colour corresponding to a 'weighted average' of the spectral colours reported when the electrodes were activated in isolation. In the light of these preliminary results, a possible speech coding scheme may be proposed. In this scheme, speech signal parameters such as fundamental frequency, formant frequencies and the amplitude envelope of the overall speech signal, would be measured by a pre-processing unit following the same principles as a formant-vocoder (Flanagan, 1972). These speech signal parameters would be then transformed to electrical stimulus parameters. As an initial attempt, formant frequencies could be transformed to a pattern of activated electrodes, fundamental frequency to pulse rate (the same for all activated electrodes), and amplitude envelope to current level. With this scheme, it is assumed that spectral colours can be cued by the electrodes that are activated, pitch by stimulus pulse rate and loudness by current level. This proposal takes account of the major factors (electrode placement and pulse rate) influencing the perception of spectral colour and pitch, and from our observation the effect of current level on spectral colour and pitch appears to be secondary. Finally, it should be noted that the proposed scheme is based on results obtained with periodic pulse trains, and therefore strictly speaking is only applicable to quasi-periodic hearing sensations such as vowel sounds. Further experimentation is required to determine whether this proposal is adequate to convey usable information on the transitional characteristics of speech, and on other categories of speech sound such as the unvoiced fricatives. Acknowledgements
We would like to acknowledge the financial support obtained from the National Health and Medical Research Council of Australia, the Channel
REPORT ON A COCHLEAR IMPLANT OPERATION
695
O Nerve Deafness Appeal and the Clive and Vera Ramaciotti Foundation. We would like to thank Mr. R. J. Walkerden and Mr. L. Cole for technical support and Miss J. Maher for the typing. REFERENCES AINSWORTH, W. A. (1976) Mechanisms of Speech Recognition. Pergamon Press, Oxford, p. 42. BALLANTYNE, J. C , EVANS, E. F., and MORRISON, A. W. (1978) Journal of Laryngology and
Otology, Supplement, 1. BEKESY, G. VON (1960) Experiments in Hearing, McGraw-Hill, New York. p. 442. BERNARD, J. R. L. (1970) Zeitschrift fur Phonetik, 23, 113. BILGER, R. C , BLACK, F. O., MYERS, G. K , HOPKINSON, N. T., VEGA, A., and WOLF, R. V.
(1977) Annals of Otology, Rhinology and Laryngology, 86, Supplement, 38, 1. BLACK, R. C. (1978) Ph.D. Thesis, University of Melbourne. BLACK, R. C , and CLARK, G. M. (1977) University of Melbourne Research Report, 140. BLACK, R. C , and CLARK, G. M. (1978) Proceedings of the Australian Physiological and Pharmacological Society, 9, 7IP. BLACK, R. C , and CLARK, G. M. (1979) Submitted for publication. CHOUARD, C. H., and MACLEOD, P. (1976) Laryngoscope, 86, 1743. CLARK, G. M., BLACK, R., DEWHURST, D. J., FORSTER, I. C , PATRICK, J. F., and TONG, Y. C.
(1977a) Medical Progress through Technology, 5, 127. CLARK, G. M., O'LOUGHLIN, B. J., RICKARDS, F. W., TONG, Y. C , and WILLIAMS, A. J. (1977b)
Journal of Laryngology and Otology, 91, 697. CLARK, G. M., TONG, Y. C , BLACK, R., FORSTER, I. C , PATRICK, J. F., and DEWHURST, D. J.
(1977c) Journal of Laryngology and Otology, 91, 935. CLARK, G. M., BLACK, R., FORSTER, I. C , PATRICK, J. F., and TONG, Y. C. (1978) Journal of
the Acoustical Society of America, 63, 631. CLARK, G. M., PATRICK, J. F., and BAILEY, Q. (1979a) Journal ofLaryngology and Otology, 93,
107. CLARK, G. M., PYMAN, B. C , and BAILEY, Q. R. (1979b) Journal of Laryngology and Otology,
93,215. DELATTRE, P. C , LIBERMAN, A. M., COOPER, F. S., and GENSTMAN, L. J. (1952) Word, 8, 195. DOUEK, E. E., FOURCIN, A. J., MOORE, B. C. J., and CLARKE, G. P. (1977) Proceedings of the
Royal Society of Medicine, 70, 379. EVANS, E. F. (1975) Cochlear Nerve and Cochlear Nucleus. In: Handbook of Sensory Physiology. V/2. Eds. W. D. Keidel and W. D. Neff, pp. 1-108. Springer-Verlag, Berlin. EVANS, E. F. (1978) Audiology, 17, 369. FLANAGAN, J. L. (1972) Speech Analysis, Synthesis and Perception, Springer, Berlin. FORSTER, I. C. (1978) Ph.D. Thesis. University of Melbourne. FORSTER, I. C , and PATRICK, J. F. (1978) 18th Australasian Conference on Physical Sciences and Engineering in Medicine and Biology. HOUSE, W. F., and URBAN, J. (1973) Annals of Otology, Rhinology and Laryngology, 82, 504. MICHELSON, R. P. (1971) Archives of Otolaryngology, 93, 317. MLADEJOVSKY, M. G., EDDINGTON, D. K., DOBELLE, W. H., and BRACKMAN, D. E. (1975)
Transactions of the American Society for Artificial Internal Organs, 21, 1. MOXON, E. C. (1971) Ph.D. Thesis. Massachusetts Institute of Technology. PLOMP, R. (1976) Aspects of Tone Sensation. Academic Press, London, p. 109. ScinNDLER, R. A., MERZENICH, M. M., WHITE, M. W., and BJORKROTH, B. (1977) Archives of
Otolaryngology, 103, 691-699. SIMMONS, F. B. (1966) Archives of Otolaryngology, 84, 24. STEVENS, S. S. (1975) Psychophysics. John Wiley & Sons, New York. TEGHTSOONIAN, R. (1971) Psychological Review, 78, 71.
