- Email: [email protected]
The cochlear implant electrode–pitch function
Hearing Research 213 (2006) 34–42 www.elsevier.com/locate/heares
Research paper
The cochlear implant electrode–pitch function Uwe Baumann a
a,*
, Andrea Nobbe
b
Department of Otorhinolaryngology, University of Munich, Marchioninistr. 15, 81377 Mu¨nchen, Germany b Med-El GmbH, Fu¨rstenweg 77A, 6020 Innsbruck, Austria Received 27 May 2005; received in revised form 12 December 2005; accepted 12 December 2005 Available online 25 January 2006
Abstract The cochlear frequency-place function in normal hearing ears has been found to be an exponential relationship in a wide variety of species [D.D. Greenwood, J. Acoust. Soc. Am. 87 (1990) 2592–2605]. Although it seems reasonable to assume a similar function for electrical stimulation by means of an intra-cochlear electrode array, the exact frequency-place function for this special type of stimulation needs to be investigated. Six users of the MED-EL COMBI 40+ cochlear implant device with moderate to profound hearing loss between 125 and 1000 Hz in the non-implanted ear took part in a binaural pitch adjustment experiment. The COMBI 40+ electrode array provides a deep insertion into the scala tympani and a wide spatial separation between the stimulating electrodes. Insertion depth was controlled by Stenver’s view plain radiographs and the insertion angle was estimated. The task of the subjects was to adjust the frequency of a sinusoid presented in the non-implanted ear by means of an adjusting knob until they perceived the same pitch as was elicited by a reference stimulus in the implanted ear. The results show adjustments corresponding to electrode positions along the cochlea, with the exception of the two most apical electrodes for most of the subjects. Pitch increased in an orderly fashion with an average of 98 Hz per electrode separation (40 Hz/mm). In contrast to the exponential predictions according to [D.D. Greenwood, J. Acoust. Soc. Am. 87 (1990) 2592–2605] for normal hearing, the average electrode–pitch function shows a linear relationship. Ó 2006 Elsevier B.V. All rights reserved. Keywords: Cochlear implant; Electric hearing; Frequency-place function; Pitch sensation
1. Introduction The internal parts of current multi-channel cochlear implant systems consist of a stimulator housing and an electrode array which is at one end connected to the stimulator and at the other end inserted into the scala tympani of the cochlea. Depending on the design of the implant a number of electrodes are distributed on a silicone electrode array. Damaged non-functional hair cells are bypassed by an electrical field emitted from these electrodes in order to stimulate neural structures of the distal part of the auditory nerve. The success of multi-channel cochlear implants is based on the fact that the stimulation of electrodes on
different locations inside the cochlea evokes different pitch percepts. Electrodes located near the round window elicit the perception of high pitch, whereas electrodes at the apical end of the cochlea transmit low pitch sensations. In normal hearing, the oscillation pattern of the traveling wave defines the mapping of an acoustic pure tone stimulus to a certain place of strongest deflection of the basilar membrane and the subsequent movement of inner and outer hair cells. The cochlear frequency-place function in normal hearing ears has been investigated for human cadavers and a wide variety of other species (Be´ke´sy, 1960; Otte et al., 1978), and an exponential relationship F ¼ Að10ax kÞ
*
Abbreviations: pps – Pulses per second Corresponding author. Tel.: +49 89 7095 3878; fax: +49 89 7095 6869. E-mail address: [email protected] (U. Baumann).
0378-5955/$ - see front matter Ó 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.heares.2005.12.010
ð1Þ
has been proposed (Greenwood, 1961, 1990). According to Greenwood, suitable constants (for human ears) are
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
A = 165.4 (to yield frequency in Hz), a = 0.06 (x expressed in millimeters distance from apex) and k = 0.88 to yield the conventional lower frequency limit. Although it seems reasonable to assume a similar function for electrical stimulation by means of an intra-cochlear electrode array, the exact frequency-place function for this special type of stimulation needs to be investigated. An electric frequency-place function—or better termed electrode–pitch function—is of high importance for the design of the signal processing strategies implemented in cochlear implant speech processor devices. In current signal processing strategies the signal is filtered by a band pass filter bank and the information of each band pass filter is then transmitted to an electrode location inside the cochlea. Usually, the arrangement of band pass filter center frequencies is based on a logarithmic function in order to simulate the frequency-place function for normal hearing. Due to the facts that the individual insertion depth of the electrode array shows a large variation and that the total length of the electrode array depends on the type of the implant, large between-subject variations of the electrode–pitch functions are expected. An exact allocation of spectral information in the signal to an electrode with corresponding pitch perception might contribute to a better acceptance of the sound of a cochlear implant and might enhance the representation of spectral information. One way to examine the exact electrode–pitch function is to gain data from implanted subjects with considerable residual acoustic hearing in the non-implanted ear. Most patients of this group show residual hearing in the lower frequency range up to 1 kHz. Regarding Greenwood’s formula (1), the apical electrodes of cochlear implants with long electrode arrays (c.f. MED-EL COMBI 40+) might evoke pitch percepts which correspond to the range of residual acoustic hearing at the non-implanted ear. The most apical electrode of the MED-EL COMBI 40+ array is located at a 30.3-mm distance from the round window. If the correspondence of frequency to place derived from normal hearing is also valid for electrical stimulation this electrode might evoke a pitch percept in electric hearing which corresponds to about a 170 Hz pure tone in normal hearing (according to formula (1) and assuming a total average length of the human cochlea of 35 mm). In the same way, the next more basally located electrode (27.9 mm) might evoke a pitch percept corresponding to a 296 Hz pure tone and the sixth electrode (16.7 mm) a pitch percept corresponding to a 1.54 kHz pure tone. The most basal electrode (3.9 mm) of this implant type might evoke a pitch percept corresponding to a 12.6 kHz pure tone. If Greenwood’s formula is applicable to electrical stimulation, the same place of excitation inside the cochlea will elicit the same pitch perception regardless whether acoustical or electrical stimulation takes place. In order to test this hypothesis, an experiment with selected cochlear implant subjects with considerable residual hearing at the non-implanted ear was conducted. Subjects had to compare the pitch height of acoustical and electrical stimula-
35
tion by matching the frequency of a pure tone delivered to the non-implanted ear with the pitch percept elicited by electrical stimulation of a certain apical electrode with a fixed stimulation rate. The main problem resulting from the hearing loss in the acoustic hearing ear is that the validity of the acoustic pitch as a reference in these subjects is questionable. In particular, it is not possible to gauge whether the acoustic stimuli are truly appropriate as standards for pitch. The fact that subjects can discriminate nearby acoustic frequencies with reasonable accuracy does not necessarily mean that they perceive pitch in the same way as a normal hearing listener would: for instance, their discrimination may have been based on other sensations (timbre, loudness, or other quality differences). These restrictions have to be considered in the interpretation of the data. 2. Materials and methods 2.1. Subjects The six participating subjects (S1, S2, S3, S7, S13 and S15) had residual hearing at the non-implanted ear with hearing loss between 45 and 100 dB HL at 125 Hz and between 70 and 105 dB HL at 1000 Hz. Four subjects had a sloping hearing loss characteristic with residual hearing mainly in the lower frequency region. One subject (S4) of this group had only limited residual hearing up to only 400 Hz. S1 and S15 had a flat profound hearing loss with very limited residual hearing. The individual pure tone audiograms are shown in Fig. 1. Three subjects were regularly using a hearing aid at the contralateral ear (S2, S13, S15). Table 1 shows the demographic data of the subjects, including age, aetiology, duration of implant use, duration of deafness of the implanted ear, score for a sentence test in noise (SNR 10 dB) and the insertion angle of the most apical electrode estimated with a procedure according to Cohen et al. (1996). The subjects were all postlingually deafened with varying aetiologies. Subject S13 lost his hearing on the left side at the age of 3 and developed a sensorineural hearing loss on the right ear with the age of 30. The subjects were selected on the basis of their availability for psychophysical testing and had all participated in previous psychophysical and speech perception studies. Their age ranged from 31 to 77 years with an average of 52 years. The subjects had been using their cochlear implants between 3 and 50 months. All subjects were implanted at the ENT clinic of the University of Munich (surgeon Prof. G. Rasp) with the MED-EL COMBI 40+ implant (Zierhofer et al., 1995). This device is equipped with an electrode array comprising 12 channels spaced 2.4 mm apart which is inserted into the scala tympani of the cochlea, and a separate lead with a reference electrode at the end which is placed under the temporalis muscle. Electrical stimulation is carried out in monopolar mode. The electrodes are numbered E1– E12 in apical to basal direction (c.f. figure of array in Baumann and Nobbe, 2004a). The subjects received an allowance for participation in the study and gave informed consent. The design of the study was approved by the local ethical committee. The electrode array was successfully placed into the cochlea for all subjects, which was confirmed by a modified Stenver’s view plain radiograph of the implanted ear (‘cochlear view’, Cohen et al., 1996). Fig. 2 displays a summary of the individual radiographs with electrode positions highlighted in order to enhance visibility. Although precise measurements of insertion depth are difficult with plain radiographs, insertion angle markers are given in Fig. 2 to roughly divide the subjects into groups with deeper or shallower electrode insertion. Four subjects show an insertion angle of more than 630° and even the shallower inserted electrodes of S2 and S15 show insertion angles of more than 540°. Compared to insertion angle measurements of Nucleus CI22M or CI24 implants given by Yukawa
36
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
250
500
1000
2000
4000
-1 0 0 10 20 30 40 50 60 70 80 90 100 110 120 130
S7
125
250
500
1000
2000
4000
500
1000
2000
4000
125
250
500
1000
2000
Frequency / [Hz]
8000
4000
0 10
20
20
20
30
30
30
40
40
40
[dB HL]
0 10
[dB HL]
0
70
125
250
500
1000
2000
4000
8000
500
1000
2000
4000
8000
-10
10
60
250
Frequency / [Hz]
8000
-10
50
125
-10 0 10 20 30 40 50 60 70 80 90 100 110 120 130
S15
Frequency / [Hz]
8000
-10
[dB HL]
250
-10 0 10 20 30 40 50 60 70 80 90 100 110 120 130
S13
Frequency / [Hz] 125
S4
Frequency / [Hz]
8000
[dB HL]
[dB HL]
S2
Frequency / [Hz] 125
[dB HL]
S1
50 60 70
50 60 70
80
80
90
90
90
100
100
100
80
110
110
110
120
120
120
130
130
130
Fig. 1. Pure tone audiograms showing the individual amount of residual hearing at the acoustically stimulated ear.
Table 1 Subject data Index
Sex
Age at testing [years]
Cause of deafness
Duration of implant use [months]
Duration of deafness on implanted ear [months]
Sentence test with SNR 10 dB [%]
Insertion angle of most apical electrode [°]
S1 S2 S4 S7 S13 S15
M M F F M M
62 31 77 37 54 50
Progr. degen. Trauma, progr. degen. Progr. degen. Progr. degen. Progr. degen. Progr. degen.
50 39 41 24 33 3
3 60 8 33 552 36
89 98 100 83 99 99
650 576 680 663 643 600
et al. (2004), the radiographs confirm that the electrode tip of the MEDEL COMBI 40+ device is on average inserted more than a half turn deeper into the cochlea.
2.2. Procedure The pitch-matching experiment was controlled by a IBM PC compatible computer via ‘Matlab’Ò software and customized routines were applied to deliver instructions to an interface for provision of the electrical stimulation by the implant (Research Interface Box, RIB, Univ. of Innsbruck). The task was performed at different apical electrodes and with varying starting frequencies of the acoustic stimulus. The stimulated electrode was chosen randomly. The stimuli for the implanted and the residual hearing ear were presented alternately between both sides. The subject had to turn an adjusting knob to change the frequency of the acoustic stimulus. The matching task was terminated when the subject pressed a key to indicate that the acoustic stimulus was perceived with the same pitch as the electric reference stimulus in the implanted ear. The subject was allowed to listen to the alternating stimuli without a time limit. For five of the subjects (S1, S2, S7, S13, S15), ten fixed starting frequencies were chosen ran-
domly between 125 and 1000 Hz. For each electrode and each starting frequency two adjustments were collected. The average adjustment for each reference electrode was calculated as the median of 20 adjustments. Due to the limited range of residual hearing of subject S4, a reduced set of seven starting frequencies was chosen randomly between 75 and 300 Hz resulting in 14 estimations for each electrode. The electric stimuli consisted of biphasic current pulses with pulse duration of 26.7 ls per phase. Zeng (2002) showed that electrical stimulation with rates above 300–500 pps evoked mostly place pitch perceptions. A previous study with some of the subjects had also shown that with pulse rates above roughly 300 pps no further increase in pitch height was perceived (Baumann and Nobbe, 2004b). Therefore, the stimulation rate was fixed to 800 pps. Depending on the residual hearing in the non implanted ear of the subject, three (subject S4: upper limit of hearing 400 Hz) to six apical electrodes (E1 to E3–E6) were stimulated. Due to an unwanted co-stimulation of the facial nerve by electrodes E5 and E6, Subject S13 performed the test only for the four most apical electrodes. The current amplitude was adjusted to the perception of comfortable loudness for each stimulated electrode with an ascending-descending technique described elsewhere (Baumann and Nobbe, 2004a).
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
37
Fig. 2. Modified Stenver’s view radiographic images with highlighted electrodes. Estimation of insertion angle indicated. Subjects S2, S4, S7 with deep electrode insertion.
The acoustic stimuli consisted of pure tones with a 25-ms rise/fall time and were digitally generated. The pure tones were delivered via D/A converter (24-bit, 44.1 kHz sample rate, ADI 8 DS, RME, Germany) and amplifier and presented over audiometric headphones (HDA 200, Sennheiser). The frequency of the sinusoids was adjustable between 125 and 1000 Hz (75–400 Hz for S4) in 1-Hz steps. The level of the acoustic stimuli was determined via ‘Matlab’Ò software as follows: prior to testing the level to achieve comfortable loudness was determined at 125, 250, 500, 750 and 1000 Hz (additionally 75 Hz for S4). Depending on the adjusted frequency, the actual acoustical presentation level during the experimental run was then calculated based on a linear interpolation between the predetermined comfortable loudness levels. To achieve comfortable loudness level, the sound pressure level was in the range of 80–110 dB SPL (B&K coupler 4152), depending on frequency and individual hearing loss. All stimuli had a duration of 500 ms. The inter-stimulus gap between electric and acoustic stimulus was 500 ms.
3. Results The individual results of the pitch matching task are shown in Fig. 3. The median and the twenty adjustments are plotted for each electrode and subject. Although the adjustments for each electrode scatter within a wide range, the average adjusted frequency of the acoustic stimulus is increasing with increasing electrode number in each subject. Most of the subjects matched the frequency of the pure tone for two most apical electrodes E1 and E2 as equal in pitch. Two subjects do not show differences in the adjustments between more basal electrodes: S1 between E4 and E5, S15 between E3 and E4. The average adjustment corresponding to E1 was between 150 Hz (S4) and 380 Hz (S13). The adjusted frequency corresponding to E6 varies between 520 Hz (S1) and 780 Hz (S15). The results of the subjects S1, S2, S7 and S15 (S13 too, but only three significantly different test electrodes) show a similar
slope for those electrodes which elicit a significantly different pitch perception. Fig. 4a and Table 2 gives a comparison of the individual average data (medians). For the purpose of comparison, the Greenwood formula (1), with parameters set for human ears,1 is also included in Fig. 4a (dashed line). It is obvious that the individual data is poorly characterized by the Greenwood function. Every individual average electrode frequency adjustment for the most apical test electrode is higher, and the adjustments derived from the two most basal test electrodes are lower than the Greenwood function. In addition, the slope of the individual electrode pitch function seems to appear shallower than the Greenwood function. The individual slope of the electrode pitch function for each subject is reflected by parameter a in Table 2 which is the result of a linear regression model calculated with SigmaPlotÒ Version 8 for a set of test electrodes indicated in Table 2. The data used as input into the regression analysis were all electrodes in basal direction after the first apical electrode pair with significantly different pitch perception including this pair. Despite the relatively large between subject variation in terms of the individual frequency adjustments for each electrode the slope parameter a is similar for all subjects with available data (average a = 37.0 ± 1.9). The coefficient of determination R2, the most common measure of how well a regression model describes the data, is high for subjects S1, S2, S7 and S15. Due to the extremely sloping characteristic of
1 Since the data is relative to electrode number and not to an absolute place, for the application of the Greenwood formula an assumption of 2.5 mm distance of E1 to the helicotrema was assumed to gain the absolute locations in mm.
38
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
* t -test, p < 0.05
Adjusted frequency in Hz
1000 700 500
*
*
300
*
*
*
*
200
*
100
S1 E1
E2
E3
E4
E5
S2
E6
E1
E2
E3
E4
E5
S4 E1
E6
E2
E3
E4
E5
Adjusted frequency in Hz
1000 700
*
*
500
* *
*
300
*
*
200
*
S7
100 E1
E2
*
*
*
E6
E3
E4
E5
S13
E6
E1
Electrode number
E2
E3
E4
E5
S15
E6
E1
Electrode number
E2
E3
E4
E5
E6
Electrode number
Fig. 3. Individual adjusted frequencies of pure tones matching in pitch height perception with electrical stimulation plotted as a function of electrode number. Average depicted with continuous line. Significant differences between two neighbored electrodes are indicated with a star. Intracochlear electrode position schematically depicted for each subject.
500
700 500 300 S2 S7 S13 S15 S3 S1 Greenwood
200
100 E1 a
E2
E3 E4 E5 Electrode Number
Frequency difference (Ref. E1)
Adjusted frequency in Hz
1000
400 300 200 100 0
E6
E1 b
E2 E3 E4 E5 Electrode number
E6
Fig. 4. (a) Average of the individual pitch adjustments. Dashed line: Greenwood function formula (1), parameters set to human ears, E1 distance set to 2.5 mm from helicotrema. (b) Average adjustments of subjects with complete data sets (S1, S2, S7, S15, error bars standard deviation) in relation to the average adjustment of E1. Dashed line: linear regression between E2 and E6 (R2 = 0.98).
Table 2 Average adjusted frequencies and results of linear regression Subject
fE1 [Hz]
fE2 [Hz]
fE3 [Hz]
fE4 [Hz]
fE5 [Hz]
fE6 [Hz]
E in linear regression
R2
S1 S2 S4 S7 S13 S15
202 331 150 223 380 331
190 325 134 220 408 496
260 392 254 241 454 608
372 568
423 565
520 686
320 863 587
534
629
738
773
E2–E6 E3–E6 E2–E3 E2–E6 E1–E4 E1–E6
0.99 105.2 34.28 0.90 221.3 38.94 Only 2 data points available 0.84 97.8 37.59 0.54 342.9 38.32 0.93 367.9 36.00
Y0
Slope a
SE 0.04 0.10 0.19 0.25 0.09
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
the hearing loss of subject S13 only 4 data points were available for the linear regression model which maybe explains the low R2 coefficient for this subject. Subject S4 was excluded from the regression model due to the fact that only two data points were available. It is reasonable to assume, that the large between subject variations in terms of the individual adjusted frequency for each test electrode are related to the individual insertion depth of the electrode array. Fig. 4b shows the results of an attempt to compensate individual differences, whereby the averaged adjustments (median) are depicted in relation to the adjustment of E1 for subjects with complete data sets (S1, S2, S7, S15). The average data shows no statistical significant difference (p < 0.05) in terms of frequency adjustment between the two most apical electrodes. The average adjustment starts to increase for electrodes more basal than E2. A linear regression analysis of the adjustments between E2 and E6 shows a high correlation (R2 = 0.98). The slope of the regression is 98 Hz per electrode, which corresponds to an increase of 40 Hz per mm. 4. Discussion One of the most important questions regarding the liability of the results presented here is whether the acoustic frequencies presented to the profoundly hearing impaired ear can serve as a standard in the comparison procedure between electrical and acoustical stimulation. Turner et al. (1983) investigated binaural pitch matches in a unilateral hearing impaired subject and reported that the pitch adjustments were within the limits of matching variability and normal diplacusis effects. The matching reliability in the order of one semitone (6%) was significantly higher than in normal hearing subjects. The large variance of the individual estimations (standard deviation in average 20%) in the present study indicates that the subjects’ task to adjust the acoustic frequency was difficult. During the experimental runs subjects often reported that the perception of the acoustic and electric stimuli differed to such a large extent that the comparison of pitch height was distracted. Whereas the electrical stimulation elicited a clear and pleasant pitch sensation, the acoustic stimulus was perceived by subjects with profound hearing loss as buzzy and was sometimes accompanied by a feeling of uncomfortable non-auditory vibrotactile or pressure sensations. Especially in subjects with profound sloping hearing loss (S4, S7, S13), regions of nonfunctional inner hair cells in the basal cochlea (dead regions) might be present (Moore et al., 2000; Moore and Alcantara, 2001). To address the question whether the residual acoustic hearing is accurate enough to serve as a reference, subjects S1, S7, and S15 repeated the experiment with a modified signal presentation. In order to estimate pitch discrimination accuracy, two consecutive acoustic signals were presented on the ear with residual hearing with the first interval serving as reference stimulus. The task of the sub-
39
ject was to adjust the frequency of the second interval stimulus according to the pitch perception elicited by the reference stimulus. The frequency of the acoustic reference stimulus was individually set to the previously determined average adjustment derived with electrical reference stimulus, the starting frequency of the second stimulus was set randomly out of ten fixed frequencies in the range of 125 Hz to 1000 Hz as described earlier for the electrical/ acoustical test procedure. Fig. 7 allows a comparison of the median and interquartile ranges between adjustments with acoustical or electrical reference. In regard of the adjustments obtained with acoustical reference (indicated by numbers displaying the frequency of the reference stimulus) it is obvious that the interquartile ranges are narrower than for the electric reference. Therefore it can be concluded that the pitch perception elicited by acoustical stimulation in the profoundly hearing impaired ear of the subjects is reasonably accurate, although the profound hearing loss clearly deteriorates the individual frequency resolution of the subject. This conclusion is further supported by the JND Df data from S2 obtained from an additional 2 interval AFC frequency discrimination test in his acoustic hearing ear. For reference frequencies set to 331 Hz, 392 Hz and 564 Hz the JND Df was 22.5 Hz, 16.1 Hz and 34.2 Hz respectively (stimulus presentation level 95–100 dB SPL). This means that the acoustic frequency discrimination accuracy of S2 is in the order of one semitone. This result is in accordance with the results for acoustic reference for S1 and S15 displayed in Fig. 7. Our conclusion is, that although the frequency discrimination of the hearing impaired ear is limited to some extent, the accuracy of the acoustic reference is reasonably high to support the general observations. Four out of six subjects (S1, S2, S4, S7) estimated the pitch of the two most apical electrodes E1 and E2 as equal. Subject S13 showed a small, but significant pitch estimation difference between E1 and E2 (paired t-test T = 2.16, p = 0.044, df = 18). Interestingly, with the exception of S2, these subjects show a comparably deep insertion of the electrode. It is likely that this observation is related to the neurophysiology of the most apical part of the cochlea where no spiral ganglion cells are present (Hochmair et al., 2003; Spoendlin and Schrott, 1989). The most apical half turn of the cochlea is innervated through Rosental’s canal with dendrites, which are connected to narrowly neighbored spiral ganglion cells located more centrally in the modiolus. Due to the spread of the electrical field emitted by the most apical electrodes2 it can be assumed that the same set of dendrites is stimulated and the same spiral ganglion cells will be activated. Consequently, no difference in terms of pitch height will be perceivable. Some important observations will be discussed in more detail in the following: 2
Rattay et al. (2001a,b) have presented a model for the spread of the electrical field with calculations for an electrode located in the first turn of the cochlea.
40
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
best frequency in normal hearing at this place of the cochlea. All other more basal located electrodes are matched to a lower pitch than predicted by formula (1). This mismatch was confirmed also recently by Boe¨x et al. (2006) in a study with six clarion implant users with residual hearing in the non-implanted ear. (2) The frequency adjustments show a large within as well as between subject variability. The most apical electrode was estimated between 150 and 380 Hz. It seems reasonable to assume, that the individual differences can be attributed to differences in insertion depth of the electrode array. Subjects with shallower insertion (S2, S13 and presumably S15) show higher adjustments than subjects with deeper insertion (S1, S4, S7). An attempt for the compensation of major
(1) A dependency of pitch height on electrode location was observed. Between E2 and E6 the pitch is linearly increasing with a slope of about 40 Hz per mm from apical to basal. Although there is a lack of sufficient data to confirm the observation completely, the average frequency adjustments seem not to follow the exponential frequency-place function of normal hearing proposed by Greenwood (1961, 1990). The pitch of the most apical electrode E1 (average 277 Hz) is matched with a higher acoustic frequency than was expected (30.4 mm distance to the round window corresponds to a best frequency of about 170 Hz in normal hearing). Only the adjusted acoustic frequency of reference electrode E2 (average 272 Hz) as well as E3 (average 326 Hz) corresponds roughly to the assumed
1000
1000
Adjusted frequency in Hz
700
800 500 S2
300
600
S7 S13
200
400
S15 S3
200
S1
100
0 E1
a
E2
E3
E4
E5
E6 E7* E8*
E1
b
E2
E3
E4
E5
E6 E7* E8*
Fig. 5. (a) Same as Fig. 4, but pitch adjustment data from subjects S2, S13, S15 shifted 2 more basal electrode places to consider the shallower insertion depth. The asterisk at E7, E8 indicates, that these electrode were not tested. (b) Median and interquartile range of the collapsed data (all subjects included). Linear regression for pitch adjustments of electrodes E2–E8* plotted with dashed line.
2000
Electric vs. acoustic reference: Median and interquartile ranges 900
1000
E3
700
757 E6
500
Adjusted frequency [Hz]
Frequency [Hz]
E6
E6
700
300 200 100 50
Greenwood Cochlear Implant (lin. regression)
30
500
5
10
15
20
E1
530 368
300
611
757
E4
400
370 E1
330
E1
200
213 200
20 0
E4
S1
25
Distance from apex Distance from E1 [mm] Fig. 6. Comparison of the cochlear frequency-place function for human ears according to Greenwood (solid) and cochlear implant electrode–pitch function (dashed, replotted from Fig. 5b. Assumption of 2.5 mm minimal distance of the tip of the electrode array from the apex.
100
Gray: electrode number
S7
S15
Black: frequency of acoustic reference [Hz]
Fig. 7. Comparison of reproducibility between bimodal (electrical– acoustical alternating between the ears) and pure acoustical stimulation (presentation of two consecutive stimuli on the ear with residual hearing). Median and interquartile ranges. Test electrodes E1, E4 (E3 S15) and E6, acoustic reference frequency given as label.
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
insertion depth differences is to apply an X-axis shift to the data derived in the pitch matching task. Fig. 5a displays data from Fig. 4a with the exception that data derived from subjects with an insertion angle below 650° (S2, S13, S15) was shifted 2 electrodes more basal. Applying the shift, the results of subjects with shallower insertion seem to fit better into the whole data set. The collapsed data for all subjects derived with the described basal electrode shift is displayed in Fig. 5b. The extension to ‘‘virtual’’ electrodes E7* and E8* (electrodes E5 and E6 shifted) is still in agreement with the assumption of a linear place–pitch relationship as already shown by linear regression for E2–E6 in Fig. 4b. The slope of the linear regression for E2–E8* is with 95.4 Hz/electrode nearly equal to that of the slope of the linear regression for data from E2 to E6 (98 Hz/electrode) and the coefficient of determination is R2 = 99.3. (3) In normal hearing, the best frequency increases by 70 Hz per mm in the apical region (Zwicker and Fastl, 1999). The slope of the electrode–pitch function differs from the cochlear frequency-place function in normal hearing. This finding is demonstrated inFig. 6, where formula (1), with parameters set for human ears, is compared with the results of the linear regression shown in Fig. 5a. The observation of a linear electrode–pitch function was also made in a previous study where a comparable pitch matching task with one Inneraid cochlear implant subject was performed (Dorman et al., 1994). (4) Exclusion of effects contributed to octave confusions: the difference between electric/acoustic frequency matches of the present study and the frequency-place function of normal hearing might be caused by octave confusions in their adjustments of the acoustic frequencies. The range of remaining hearing and therefore the range of adjustable frequencies for the acoustic stimulus were limited. As a consequence the subjects had to find matching frequencies within the limited range of their residual hearing. This effect is therefore more likely to occur at basal electrodes because the acoustic range of the apical electrodes corresponds better to the assumed frequency-place allocation. Octave confusion would be characterized by bimodal frequency adjustment distributions. An evaluation of histograms derived for each electrode and subject gave no evidence for the occurrence of octave confusions (data not shown).
5. Conclusion (1) Electrical stimulation on different places in the apical last turn of the cochlea does not elicit the perception of pitch differences.
41
(2) Successively stimulated electrodes located in more basal cochlear regions convey pitch increments in an orderly linear fashion. (3) The lowest pitch perception is determined by the insertion depth of the most apical electrode. (4) The application of the exponential frequency-position function derived by Greenwood (1961), Greenwood (1990) for normal hearing seems to be questionable for intracochlear electrical stimulation.
Acknowledgements We express our thanks and appreciation to the subjects for their time and effort. Prof. G. Rasp provided us with the Stenver’s radiographs. We are indebted to the Department of Applied Physics at the University of Innsbruck (Prof. E. Hochmair) for provision of the research interface and programming support. We gratefully acknowledge comments from Ernst Terhardt, Bob Carlyon, Fan-Gang Zeng and another anonymous reviewer on earlier versions of this paper. This research was conducted while author Andrea Nobbe was working at the ENT department of the Ludwig-Maximilians-University in Munich. It was supported by DFG Grant No. FO 306/2-3 (Research group ‘‘Auditory objects’’). References Baumann, U., Nobbe, A., 2004a. Pitch ranking with deeply inserted electrode arrays. Ear Hear. 25, 275–283. Baumann, U., Nobbe, A., 2004b. Pulse rate discrimination with deeply inserted electrodes. Hear. Res. 196, 49–57. Be´ke´sy, G. von, 1960. Experiments in Hearing. McGraw-Hill, New York. Boe¨x, C., Baud, L., Cosendai, G., Sigrist, A., Ko´s, M.-I., Pelizzone, M., 2006. Acoustic to electric pitch comparisons in cochlear implant subjects with residual hearing. J. Assoc. Res. Otol., accepted for publication. Cohen, L.T., Xu, J., Xu, S.A., Clark, G.M., 1996. Improved and simplified methods for specifying positions of the electrode bands of a cochlear implant array. Am. J. Otol. 17, 859–865. Dorman, M.F., Smith, M., Smith, L., Parkin, J.L., 1994. The pitch of electrically presented sinusoids. J. Acoust. Soc. Am. 95, 1677– 1679. Greenwood, D.D., 1961. Critical Bandwidth and the frequency coordinates of the basilar membrane. J. Acoust. Soc. Am. 33, 1344– 1356. Greenwood, D.D., 1990. A cochlear frequency-position function for several species—29 years later. J. Acoust. Soc. Am. 87, 2592–2605. Hochmair, I., Arnold, W., Nopp, P., Jolly, C., Muller, J., Roland, P., 2003. Deep electrode insertion in cochlear implants: apical morphology, electrodes and speech perception results. Acta Otolaryngol 123, 612–617. Otte, J., Schuhknecht, H.F., Kerr, A.G., 1978. Ganglion cell populations in normal and pathological human cochleae. Implications for cochlear implantation. Laryngoscope 88, 1231–1246. Moore, B.C., Huss, M., Vickers, D.A., Glasberg, B.R., Alcantara, J.I., 2000. A test for the diagnosis of dead regions in the cochlea. Br. J. Audiol. 34, 205–224. Moore, B.C., Alcantara, J.I., 2001. The use of psychophysical tuning curves to explore dead regions in the cochlea. Ear Hear. 22, 268–278.
42
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
Rattay, F., Lutter, P., Felix, H., 2001a. A model of the electrically excited human cochlear neuron. I. Contribution of neural substructures to the generation and propagation of spikes. Hear. Res. 153, 43–63. Rattay, F., Leao, R.N., Felix, H., 2001b. A model of the electrically excited human cochlear neuron. II. Influence of the three-dimensional cochlear structure on neural excitability. Hear. Res. 153, 64–79. Spoendlin, H., Schrott, A., 1989. Analysis of the human auditory nerve. Hear. Res. 43, 25–38. Turner, C., Burns, E.M., Nelson, D.A., 1983. Pure tone pitch perception and low frequency hearing loss. J. Acoust. Soc. Am. (73), 966–975.
Yukawa, K., Cohen, L., Blamey, P., Pyman, B., Tungvachirakul, V., O’Leary, S., 2004. Effects of insertion depth of cochlear implant electrodes upon speech perception. Audiol. Neurootol. 9, 163–172. Zeng, F.G., 2002. Temporal pitch in electric hearing. Hear Res. 174, 101– 106. Zierhofer, C.M., Hochmair-Desoyer, I., Hochmair, E.S., 1995. Electronic design of a cochlear implant for multichannel high-rate pulsatile stimulation strategies. IEEE Transact. Rehab. Engin. 3, 112–116. Zwicker, E., Fastl, H., 1999. Psychoacoustics—Facts and Models, Second ed. Springer, Berlin, Germany.
Research paper
The cochlear implant electrode–pitch function Uwe Baumann a
a,*
, Andrea Nobbe
b
Department of Otorhinolaryngology, University of Munich, Marchioninistr. 15, 81377 Mu¨nchen, Germany b Med-El GmbH, Fu¨rstenweg 77A, 6020 Innsbruck, Austria Received 27 May 2005; received in revised form 12 December 2005; accepted 12 December 2005 Available online 25 January 2006
Abstract The cochlear frequency-place function in normal hearing ears has been found to be an exponential relationship in a wide variety of species [D.D. Greenwood, J. Acoust. Soc. Am. 87 (1990) 2592–2605]. Although it seems reasonable to assume a similar function for electrical stimulation by means of an intra-cochlear electrode array, the exact frequency-place function for this special type of stimulation needs to be investigated. Six users of the MED-EL COMBI 40+ cochlear implant device with moderate to profound hearing loss between 125 and 1000 Hz in the non-implanted ear took part in a binaural pitch adjustment experiment. The COMBI 40+ electrode array provides a deep insertion into the scala tympani and a wide spatial separation between the stimulating electrodes. Insertion depth was controlled by Stenver’s view plain radiographs and the insertion angle was estimated. The task of the subjects was to adjust the frequency of a sinusoid presented in the non-implanted ear by means of an adjusting knob until they perceived the same pitch as was elicited by a reference stimulus in the implanted ear. The results show adjustments corresponding to electrode positions along the cochlea, with the exception of the two most apical electrodes for most of the subjects. Pitch increased in an orderly fashion with an average of 98 Hz per electrode separation (40 Hz/mm). In contrast to the exponential predictions according to [D.D. Greenwood, J. Acoust. Soc. Am. 87 (1990) 2592–2605] for normal hearing, the average electrode–pitch function shows a linear relationship. Ó 2006 Elsevier B.V. All rights reserved. Keywords: Cochlear implant; Electric hearing; Frequency-place function; Pitch sensation
1. Introduction The internal parts of current multi-channel cochlear implant systems consist of a stimulator housing and an electrode array which is at one end connected to the stimulator and at the other end inserted into the scala tympani of the cochlea. Depending on the design of the implant a number of electrodes are distributed on a silicone electrode array. Damaged non-functional hair cells are bypassed by an electrical field emitted from these electrodes in order to stimulate neural structures of the distal part of the auditory nerve. The success of multi-channel cochlear implants is based on the fact that the stimulation of electrodes on
different locations inside the cochlea evokes different pitch percepts. Electrodes located near the round window elicit the perception of high pitch, whereas electrodes at the apical end of the cochlea transmit low pitch sensations. In normal hearing, the oscillation pattern of the traveling wave defines the mapping of an acoustic pure tone stimulus to a certain place of strongest deflection of the basilar membrane and the subsequent movement of inner and outer hair cells. The cochlear frequency-place function in normal hearing ears has been investigated for human cadavers and a wide variety of other species (Be´ke´sy, 1960; Otte et al., 1978), and an exponential relationship F ¼ Að10ax kÞ
*
Abbreviations: pps – Pulses per second Corresponding author. Tel.: +49 89 7095 3878; fax: +49 89 7095 6869. E-mail address: [email protected] (U. Baumann).
0378-5955/$ - see front matter Ó 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.heares.2005.12.010
ð1Þ
has been proposed (Greenwood, 1961, 1990). According to Greenwood, suitable constants (for human ears) are
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
A = 165.4 (to yield frequency in Hz), a = 0.06 (x expressed in millimeters distance from apex) and k = 0.88 to yield the conventional lower frequency limit. Although it seems reasonable to assume a similar function for electrical stimulation by means of an intra-cochlear electrode array, the exact frequency-place function for this special type of stimulation needs to be investigated. An electric frequency-place function—or better termed electrode–pitch function—is of high importance for the design of the signal processing strategies implemented in cochlear implant speech processor devices. In current signal processing strategies the signal is filtered by a band pass filter bank and the information of each band pass filter is then transmitted to an electrode location inside the cochlea. Usually, the arrangement of band pass filter center frequencies is based on a logarithmic function in order to simulate the frequency-place function for normal hearing. Due to the facts that the individual insertion depth of the electrode array shows a large variation and that the total length of the electrode array depends on the type of the implant, large between-subject variations of the electrode–pitch functions are expected. An exact allocation of spectral information in the signal to an electrode with corresponding pitch perception might contribute to a better acceptance of the sound of a cochlear implant and might enhance the representation of spectral information. One way to examine the exact electrode–pitch function is to gain data from implanted subjects with considerable residual acoustic hearing in the non-implanted ear. Most patients of this group show residual hearing in the lower frequency range up to 1 kHz. Regarding Greenwood’s formula (1), the apical electrodes of cochlear implants with long electrode arrays (c.f. MED-EL COMBI 40+) might evoke pitch percepts which correspond to the range of residual acoustic hearing at the non-implanted ear. The most apical electrode of the MED-EL COMBI 40+ array is located at a 30.3-mm distance from the round window. If the correspondence of frequency to place derived from normal hearing is also valid for electrical stimulation this electrode might evoke a pitch percept in electric hearing which corresponds to about a 170 Hz pure tone in normal hearing (according to formula (1) and assuming a total average length of the human cochlea of 35 mm). In the same way, the next more basally located electrode (27.9 mm) might evoke a pitch percept corresponding to a 296 Hz pure tone and the sixth electrode (16.7 mm) a pitch percept corresponding to a 1.54 kHz pure tone. The most basal electrode (3.9 mm) of this implant type might evoke a pitch percept corresponding to a 12.6 kHz pure tone. If Greenwood’s formula is applicable to electrical stimulation, the same place of excitation inside the cochlea will elicit the same pitch perception regardless whether acoustical or electrical stimulation takes place. In order to test this hypothesis, an experiment with selected cochlear implant subjects with considerable residual hearing at the non-implanted ear was conducted. Subjects had to compare the pitch height of acoustical and electrical stimula-
35
tion by matching the frequency of a pure tone delivered to the non-implanted ear with the pitch percept elicited by electrical stimulation of a certain apical electrode with a fixed stimulation rate. The main problem resulting from the hearing loss in the acoustic hearing ear is that the validity of the acoustic pitch as a reference in these subjects is questionable. In particular, it is not possible to gauge whether the acoustic stimuli are truly appropriate as standards for pitch. The fact that subjects can discriminate nearby acoustic frequencies with reasonable accuracy does not necessarily mean that they perceive pitch in the same way as a normal hearing listener would: for instance, their discrimination may have been based on other sensations (timbre, loudness, or other quality differences). These restrictions have to be considered in the interpretation of the data. 2. Materials and methods 2.1. Subjects The six participating subjects (S1, S2, S3, S7, S13 and S15) had residual hearing at the non-implanted ear with hearing loss between 45 and 100 dB HL at 125 Hz and between 70 and 105 dB HL at 1000 Hz. Four subjects had a sloping hearing loss characteristic with residual hearing mainly in the lower frequency region. One subject (S4) of this group had only limited residual hearing up to only 400 Hz. S1 and S15 had a flat profound hearing loss with very limited residual hearing. The individual pure tone audiograms are shown in Fig. 1. Three subjects were regularly using a hearing aid at the contralateral ear (S2, S13, S15). Table 1 shows the demographic data of the subjects, including age, aetiology, duration of implant use, duration of deafness of the implanted ear, score for a sentence test in noise (SNR 10 dB) and the insertion angle of the most apical electrode estimated with a procedure according to Cohen et al. (1996). The subjects were all postlingually deafened with varying aetiologies. Subject S13 lost his hearing on the left side at the age of 3 and developed a sensorineural hearing loss on the right ear with the age of 30. The subjects were selected on the basis of their availability for psychophysical testing and had all participated in previous psychophysical and speech perception studies. Their age ranged from 31 to 77 years with an average of 52 years. The subjects had been using their cochlear implants between 3 and 50 months. All subjects were implanted at the ENT clinic of the University of Munich (surgeon Prof. G. Rasp) with the MED-EL COMBI 40+ implant (Zierhofer et al., 1995). This device is equipped with an electrode array comprising 12 channels spaced 2.4 mm apart which is inserted into the scala tympani of the cochlea, and a separate lead with a reference electrode at the end which is placed under the temporalis muscle. Electrical stimulation is carried out in monopolar mode. The electrodes are numbered E1– E12 in apical to basal direction (c.f. figure of array in Baumann and Nobbe, 2004a). The subjects received an allowance for participation in the study and gave informed consent. The design of the study was approved by the local ethical committee. The electrode array was successfully placed into the cochlea for all subjects, which was confirmed by a modified Stenver’s view plain radiograph of the implanted ear (‘cochlear view’, Cohen et al., 1996). Fig. 2 displays a summary of the individual radiographs with electrode positions highlighted in order to enhance visibility. Although precise measurements of insertion depth are difficult with plain radiographs, insertion angle markers are given in Fig. 2 to roughly divide the subjects into groups with deeper or shallower electrode insertion. Four subjects show an insertion angle of more than 630° and even the shallower inserted electrodes of S2 and S15 show insertion angles of more than 540°. Compared to insertion angle measurements of Nucleus CI22M or CI24 implants given by Yukawa
36
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
250
500
1000
2000
4000
-1 0 0 10 20 30 40 50 60 70 80 90 100 110 120 130
S7
125
250
500
1000
2000
4000
500
1000
2000
4000
125
250
500
1000
2000
Frequency / [Hz]
8000
4000
0 10
20
20
20
30
30
30
40
40
40
[dB HL]
0 10
[dB HL]
0
70
125
250
500
1000
2000
4000
8000
500
1000
2000
4000
8000
-10
10
60
250
Frequency / [Hz]
8000
-10
50
125
-10 0 10 20 30 40 50 60 70 80 90 100 110 120 130
S15
Frequency / [Hz]
8000
-10
[dB HL]
250
-10 0 10 20 30 40 50 60 70 80 90 100 110 120 130
S13
Frequency / [Hz] 125
S4
Frequency / [Hz]
8000
[dB HL]
[dB HL]
S2
Frequency / [Hz] 125
[dB HL]
S1
50 60 70
50 60 70
80
80
90
90
90
100
100
100
80
110
110
110
120
120
120
130
130
130
Fig. 1. Pure tone audiograms showing the individual amount of residual hearing at the acoustically stimulated ear.
Table 1 Subject data Index
Sex
Age at testing [years]
Cause of deafness
Duration of implant use [months]
Duration of deafness on implanted ear [months]
Sentence test with SNR 10 dB [%]
Insertion angle of most apical electrode [°]
S1 S2 S4 S7 S13 S15
M M F F M M
62 31 77 37 54 50
Progr. degen. Trauma, progr. degen. Progr. degen. Progr. degen. Progr. degen. Progr. degen.
50 39 41 24 33 3
3 60 8 33 552 36
89 98 100 83 99 99
650 576 680 663 643 600
et al. (2004), the radiographs confirm that the electrode tip of the MEDEL COMBI 40+ device is on average inserted more than a half turn deeper into the cochlea.
2.2. Procedure The pitch-matching experiment was controlled by a IBM PC compatible computer via ‘Matlab’Ò software and customized routines were applied to deliver instructions to an interface for provision of the electrical stimulation by the implant (Research Interface Box, RIB, Univ. of Innsbruck). The task was performed at different apical electrodes and with varying starting frequencies of the acoustic stimulus. The stimulated electrode was chosen randomly. The stimuli for the implanted and the residual hearing ear were presented alternately between both sides. The subject had to turn an adjusting knob to change the frequency of the acoustic stimulus. The matching task was terminated when the subject pressed a key to indicate that the acoustic stimulus was perceived with the same pitch as the electric reference stimulus in the implanted ear. The subject was allowed to listen to the alternating stimuli without a time limit. For five of the subjects (S1, S2, S7, S13, S15), ten fixed starting frequencies were chosen ran-
domly between 125 and 1000 Hz. For each electrode and each starting frequency two adjustments were collected. The average adjustment for each reference electrode was calculated as the median of 20 adjustments. Due to the limited range of residual hearing of subject S4, a reduced set of seven starting frequencies was chosen randomly between 75 and 300 Hz resulting in 14 estimations for each electrode. The electric stimuli consisted of biphasic current pulses with pulse duration of 26.7 ls per phase. Zeng (2002) showed that electrical stimulation with rates above 300–500 pps evoked mostly place pitch perceptions. A previous study with some of the subjects had also shown that with pulse rates above roughly 300 pps no further increase in pitch height was perceived (Baumann and Nobbe, 2004b). Therefore, the stimulation rate was fixed to 800 pps. Depending on the residual hearing in the non implanted ear of the subject, three (subject S4: upper limit of hearing 400 Hz) to six apical electrodes (E1 to E3–E6) were stimulated. Due to an unwanted co-stimulation of the facial nerve by electrodes E5 and E6, Subject S13 performed the test only for the four most apical electrodes. The current amplitude was adjusted to the perception of comfortable loudness for each stimulated electrode with an ascending-descending technique described elsewhere (Baumann and Nobbe, 2004a).
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
37
Fig. 2. Modified Stenver’s view radiographic images with highlighted electrodes. Estimation of insertion angle indicated. Subjects S2, S4, S7 with deep electrode insertion.
The acoustic stimuli consisted of pure tones with a 25-ms rise/fall time and were digitally generated. The pure tones were delivered via D/A converter (24-bit, 44.1 kHz sample rate, ADI 8 DS, RME, Germany) and amplifier and presented over audiometric headphones (HDA 200, Sennheiser). The frequency of the sinusoids was adjustable between 125 and 1000 Hz (75–400 Hz for S4) in 1-Hz steps. The level of the acoustic stimuli was determined via ‘Matlab’Ò software as follows: prior to testing the level to achieve comfortable loudness was determined at 125, 250, 500, 750 and 1000 Hz (additionally 75 Hz for S4). Depending on the adjusted frequency, the actual acoustical presentation level during the experimental run was then calculated based on a linear interpolation between the predetermined comfortable loudness levels. To achieve comfortable loudness level, the sound pressure level was in the range of 80–110 dB SPL (B&K coupler 4152), depending on frequency and individual hearing loss. All stimuli had a duration of 500 ms. The inter-stimulus gap between electric and acoustic stimulus was 500 ms.
3. Results The individual results of the pitch matching task are shown in Fig. 3. The median and the twenty adjustments are plotted for each electrode and subject. Although the adjustments for each electrode scatter within a wide range, the average adjusted frequency of the acoustic stimulus is increasing with increasing electrode number in each subject. Most of the subjects matched the frequency of the pure tone for two most apical electrodes E1 and E2 as equal in pitch. Two subjects do not show differences in the adjustments between more basal electrodes: S1 between E4 and E5, S15 between E3 and E4. The average adjustment corresponding to E1 was between 150 Hz (S4) and 380 Hz (S13). The adjusted frequency corresponding to E6 varies between 520 Hz (S1) and 780 Hz (S15). The results of the subjects S1, S2, S7 and S15 (S13 too, but only three significantly different test electrodes) show a similar
slope for those electrodes which elicit a significantly different pitch perception. Fig. 4a and Table 2 gives a comparison of the individual average data (medians). For the purpose of comparison, the Greenwood formula (1), with parameters set for human ears,1 is also included in Fig. 4a (dashed line). It is obvious that the individual data is poorly characterized by the Greenwood function. Every individual average electrode frequency adjustment for the most apical test electrode is higher, and the adjustments derived from the two most basal test electrodes are lower than the Greenwood function. In addition, the slope of the individual electrode pitch function seems to appear shallower than the Greenwood function. The individual slope of the electrode pitch function for each subject is reflected by parameter a in Table 2 which is the result of a linear regression model calculated with SigmaPlotÒ Version 8 for a set of test electrodes indicated in Table 2. The data used as input into the regression analysis were all electrodes in basal direction after the first apical electrode pair with significantly different pitch perception including this pair. Despite the relatively large between subject variation in terms of the individual frequency adjustments for each electrode the slope parameter a is similar for all subjects with available data (average a = 37.0 ± 1.9). The coefficient of determination R2, the most common measure of how well a regression model describes the data, is high for subjects S1, S2, S7 and S15. Due to the extremely sloping characteristic of
1 Since the data is relative to electrode number and not to an absolute place, for the application of the Greenwood formula an assumption of 2.5 mm distance of E1 to the helicotrema was assumed to gain the absolute locations in mm.
38
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
* t -test, p < 0.05
Adjusted frequency in Hz
1000 700 500
*
*
300
*
*
*
*
200
*
100
S1 E1
E2
E3
E4
E5
S2
E6
E1
E2
E3
E4
E5
S4 E1
E6
E2
E3
E4
E5
Adjusted frequency in Hz
1000 700
*
*
500
* *
*
300
*
*
200
*
S7
100 E1
E2
*
*
*
E6
E3
E4
E5
S13
E6
E1
Electrode number
E2
E3
E4
E5
S15
E6
E1
Electrode number
E2
E3
E4
E5
E6
Electrode number
Fig. 3. Individual adjusted frequencies of pure tones matching in pitch height perception with electrical stimulation plotted as a function of electrode number. Average depicted with continuous line. Significant differences between two neighbored electrodes are indicated with a star. Intracochlear electrode position schematically depicted for each subject.
500
700 500 300 S2 S7 S13 S15 S3 S1 Greenwood
200
100 E1 a
E2
E3 E4 E5 Electrode Number
Frequency difference (Ref. E1)
Adjusted frequency in Hz
1000
400 300 200 100 0
E6
E1 b
E2 E3 E4 E5 Electrode number
E6
Fig. 4. (a) Average of the individual pitch adjustments. Dashed line: Greenwood function formula (1), parameters set to human ears, E1 distance set to 2.5 mm from helicotrema. (b) Average adjustments of subjects with complete data sets (S1, S2, S7, S15, error bars standard deviation) in relation to the average adjustment of E1. Dashed line: linear regression between E2 and E6 (R2 = 0.98).
Table 2 Average adjusted frequencies and results of linear regression Subject
fE1 [Hz]
fE2 [Hz]
fE3 [Hz]
fE4 [Hz]
fE5 [Hz]
fE6 [Hz]
E in linear regression
R2
S1 S2 S4 S7 S13 S15
202 331 150 223 380 331
190 325 134 220 408 496
260 392 254 241 454 608
372 568
423 565
520 686
320 863 587
534
629
738
773
E2–E6 E3–E6 E2–E3 E2–E6 E1–E4 E1–E6
0.99 105.2 34.28 0.90 221.3 38.94 Only 2 data points available 0.84 97.8 37.59 0.54 342.9 38.32 0.93 367.9 36.00
Y0
Slope a
SE 0.04 0.10 0.19 0.25 0.09
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
the hearing loss of subject S13 only 4 data points were available for the linear regression model which maybe explains the low R2 coefficient for this subject. Subject S4 was excluded from the regression model due to the fact that only two data points were available. It is reasonable to assume, that the large between subject variations in terms of the individual adjusted frequency for each test electrode are related to the individual insertion depth of the electrode array. Fig. 4b shows the results of an attempt to compensate individual differences, whereby the averaged adjustments (median) are depicted in relation to the adjustment of E1 for subjects with complete data sets (S1, S2, S7, S15). The average data shows no statistical significant difference (p < 0.05) in terms of frequency adjustment between the two most apical electrodes. The average adjustment starts to increase for electrodes more basal than E2. A linear regression analysis of the adjustments between E2 and E6 shows a high correlation (R2 = 0.98). The slope of the regression is 98 Hz per electrode, which corresponds to an increase of 40 Hz per mm. 4. Discussion One of the most important questions regarding the liability of the results presented here is whether the acoustic frequencies presented to the profoundly hearing impaired ear can serve as a standard in the comparison procedure between electrical and acoustical stimulation. Turner et al. (1983) investigated binaural pitch matches in a unilateral hearing impaired subject and reported that the pitch adjustments were within the limits of matching variability and normal diplacusis effects. The matching reliability in the order of one semitone (6%) was significantly higher than in normal hearing subjects. The large variance of the individual estimations (standard deviation in average 20%) in the present study indicates that the subjects’ task to adjust the acoustic frequency was difficult. During the experimental runs subjects often reported that the perception of the acoustic and electric stimuli differed to such a large extent that the comparison of pitch height was distracted. Whereas the electrical stimulation elicited a clear and pleasant pitch sensation, the acoustic stimulus was perceived by subjects with profound hearing loss as buzzy and was sometimes accompanied by a feeling of uncomfortable non-auditory vibrotactile or pressure sensations. Especially in subjects with profound sloping hearing loss (S4, S7, S13), regions of nonfunctional inner hair cells in the basal cochlea (dead regions) might be present (Moore et al., 2000; Moore and Alcantara, 2001). To address the question whether the residual acoustic hearing is accurate enough to serve as a reference, subjects S1, S7, and S15 repeated the experiment with a modified signal presentation. In order to estimate pitch discrimination accuracy, two consecutive acoustic signals were presented on the ear with residual hearing with the first interval serving as reference stimulus. The task of the sub-
39
ject was to adjust the frequency of the second interval stimulus according to the pitch perception elicited by the reference stimulus. The frequency of the acoustic reference stimulus was individually set to the previously determined average adjustment derived with electrical reference stimulus, the starting frequency of the second stimulus was set randomly out of ten fixed frequencies in the range of 125 Hz to 1000 Hz as described earlier for the electrical/ acoustical test procedure. Fig. 7 allows a comparison of the median and interquartile ranges between adjustments with acoustical or electrical reference. In regard of the adjustments obtained with acoustical reference (indicated by numbers displaying the frequency of the reference stimulus) it is obvious that the interquartile ranges are narrower than for the electric reference. Therefore it can be concluded that the pitch perception elicited by acoustical stimulation in the profoundly hearing impaired ear of the subjects is reasonably accurate, although the profound hearing loss clearly deteriorates the individual frequency resolution of the subject. This conclusion is further supported by the JND Df data from S2 obtained from an additional 2 interval AFC frequency discrimination test in his acoustic hearing ear. For reference frequencies set to 331 Hz, 392 Hz and 564 Hz the JND Df was 22.5 Hz, 16.1 Hz and 34.2 Hz respectively (stimulus presentation level 95–100 dB SPL). This means that the acoustic frequency discrimination accuracy of S2 is in the order of one semitone. This result is in accordance with the results for acoustic reference for S1 and S15 displayed in Fig. 7. Our conclusion is, that although the frequency discrimination of the hearing impaired ear is limited to some extent, the accuracy of the acoustic reference is reasonably high to support the general observations. Four out of six subjects (S1, S2, S4, S7) estimated the pitch of the two most apical electrodes E1 and E2 as equal. Subject S13 showed a small, but significant pitch estimation difference between E1 and E2 (paired t-test T = 2.16, p = 0.044, df = 18). Interestingly, with the exception of S2, these subjects show a comparably deep insertion of the electrode. It is likely that this observation is related to the neurophysiology of the most apical part of the cochlea where no spiral ganglion cells are present (Hochmair et al., 2003; Spoendlin and Schrott, 1989). The most apical half turn of the cochlea is innervated through Rosental’s canal with dendrites, which are connected to narrowly neighbored spiral ganglion cells located more centrally in the modiolus. Due to the spread of the electrical field emitted by the most apical electrodes2 it can be assumed that the same set of dendrites is stimulated and the same spiral ganglion cells will be activated. Consequently, no difference in terms of pitch height will be perceivable. Some important observations will be discussed in more detail in the following: 2
Rattay et al. (2001a,b) have presented a model for the spread of the electrical field with calculations for an electrode located in the first turn of the cochlea.
40
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
best frequency in normal hearing at this place of the cochlea. All other more basal located electrodes are matched to a lower pitch than predicted by formula (1). This mismatch was confirmed also recently by Boe¨x et al. (2006) in a study with six clarion implant users with residual hearing in the non-implanted ear. (2) The frequency adjustments show a large within as well as between subject variability. The most apical electrode was estimated between 150 and 380 Hz. It seems reasonable to assume, that the individual differences can be attributed to differences in insertion depth of the electrode array. Subjects with shallower insertion (S2, S13 and presumably S15) show higher adjustments than subjects with deeper insertion (S1, S4, S7). An attempt for the compensation of major
(1) A dependency of pitch height on electrode location was observed. Between E2 and E6 the pitch is linearly increasing with a slope of about 40 Hz per mm from apical to basal. Although there is a lack of sufficient data to confirm the observation completely, the average frequency adjustments seem not to follow the exponential frequency-place function of normal hearing proposed by Greenwood (1961, 1990). The pitch of the most apical electrode E1 (average 277 Hz) is matched with a higher acoustic frequency than was expected (30.4 mm distance to the round window corresponds to a best frequency of about 170 Hz in normal hearing). Only the adjusted acoustic frequency of reference electrode E2 (average 272 Hz) as well as E3 (average 326 Hz) corresponds roughly to the assumed
1000
1000
Adjusted frequency in Hz
700
800 500 S2
300
600
S7 S13
200
400
S15 S3
200
S1
100
0 E1
a
E2
E3
E4
E5
E6 E7* E8*
E1
b
E2
E3
E4
E5
E6 E7* E8*
Fig. 5. (a) Same as Fig. 4, but pitch adjustment data from subjects S2, S13, S15 shifted 2 more basal electrode places to consider the shallower insertion depth. The asterisk at E7, E8 indicates, that these electrode were not tested. (b) Median and interquartile range of the collapsed data (all subjects included). Linear regression for pitch adjustments of electrodes E2–E8* plotted with dashed line.
2000
Electric vs. acoustic reference: Median and interquartile ranges 900
1000
E3
700
757 E6
500
Adjusted frequency [Hz]
Frequency [Hz]
E6
E6
700
300 200 100 50
Greenwood Cochlear Implant (lin. regression)
30
500
5
10
15
20
E1
530 368
300
611
757
E4
400
370 E1
330
E1
200
213 200
20 0
E4
S1
25
Distance from apex Distance from E1 [mm] Fig. 6. Comparison of the cochlear frequency-place function for human ears according to Greenwood (solid) and cochlear implant electrode–pitch function (dashed, replotted from Fig. 5b. Assumption of 2.5 mm minimal distance of the tip of the electrode array from the apex.
100
Gray: electrode number
S7
S15
Black: frequency of acoustic reference [Hz]
Fig. 7. Comparison of reproducibility between bimodal (electrical– acoustical alternating between the ears) and pure acoustical stimulation (presentation of two consecutive stimuli on the ear with residual hearing). Median and interquartile ranges. Test electrodes E1, E4 (E3 S15) and E6, acoustic reference frequency given as label.
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
insertion depth differences is to apply an X-axis shift to the data derived in the pitch matching task. Fig. 5a displays data from Fig. 4a with the exception that data derived from subjects with an insertion angle below 650° (S2, S13, S15) was shifted 2 electrodes more basal. Applying the shift, the results of subjects with shallower insertion seem to fit better into the whole data set. The collapsed data for all subjects derived with the described basal electrode shift is displayed in Fig. 5b. The extension to ‘‘virtual’’ electrodes E7* and E8* (electrodes E5 and E6 shifted) is still in agreement with the assumption of a linear place–pitch relationship as already shown by linear regression for E2–E6 in Fig. 4b. The slope of the linear regression for E2–E8* is with 95.4 Hz/electrode nearly equal to that of the slope of the linear regression for data from E2 to E6 (98 Hz/electrode) and the coefficient of determination is R2 = 99.3. (3) In normal hearing, the best frequency increases by 70 Hz per mm in the apical region (Zwicker and Fastl, 1999). The slope of the electrode–pitch function differs from the cochlear frequency-place function in normal hearing. This finding is demonstrated inFig. 6, where formula (1), with parameters set for human ears, is compared with the results of the linear regression shown in Fig. 5a. The observation of a linear electrode–pitch function was also made in a previous study where a comparable pitch matching task with one Inneraid cochlear implant subject was performed (Dorman et al., 1994). (4) Exclusion of effects contributed to octave confusions: the difference between electric/acoustic frequency matches of the present study and the frequency-place function of normal hearing might be caused by octave confusions in their adjustments of the acoustic frequencies. The range of remaining hearing and therefore the range of adjustable frequencies for the acoustic stimulus were limited. As a consequence the subjects had to find matching frequencies within the limited range of their residual hearing. This effect is therefore more likely to occur at basal electrodes because the acoustic range of the apical electrodes corresponds better to the assumed frequency-place allocation. Octave confusion would be characterized by bimodal frequency adjustment distributions. An evaluation of histograms derived for each electrode and subject gave no evidence for the occurrence of octave confusions (data not shown).
5. Conclusion (1) Electrical stimulation on different places in the apical last turn of the cochlea does not elicit the perception of pitch differences.
41
(2) Successively stimulated electrodes located in more basal cochlear regions convey pitch increments in an orderly linear fashion. (3) The lowest pitch perception is determined by the insertion depth of the most apical electrode. (4) The application of the exponential frequency-position function derived by Greenwood (1961), Greenwood (1990) for normal hearing seems to be questionable for intracochlear electrical stimulation.
Acknowledgements We express our thanks and appreciation to the subjects for their time and effort. Prof. G. Rasp provided us with the Stenver’s radiographs. We are indebted to the Department of Applied Physics at the University of Innsbruck (Prof. E. Hochmair) for provision of the research interface and programming support. We gratefully acknowledge comments from Ernst Terhardt, Bob Carlyon, Fan-Gang Zeng and another anonymous reviewer on earlier versions of this paper. This research was conducted while author Andrea Nobbe was working at the ENT department of the Ludwig-Maximilians-University in Munich. It was supported by DFG Grant No. FO 306/2-3 (Research group ‘‘Auditory objects’’). References Baumann, U., Nobbe, A., 2004a. Pitch ranking with deeply inserted electrode arrays. Ear Hear. 25, 275–283. Baumann, U., Nobbe, A., 2004b. Pulse rate discrimination with deeply inserted electrodes. Hear. Res. 196, 49–57. Be´ke´sy, G. von, 1960. Experiments in Hearing. McGraw-Hill, New York. Boe¨x, C., Baud, L., Cosendai, G., Sigrist, A., Ko´s, M.-I., Pelizzone, M., 2006. Acoustic to electric pitch comparisons in cochlear implant subjects with residual hearing. J. Assoc. Res. Otol., accepted for publication. Cohen, L.T., Xu, J., Xu, S.A., Clark, G.M., 1996. Improved and simplified methods for specifying positions of the electrode bands of a cochlear implant array. Am. J. Otol. 17, 859–865. Dorman, M.F., Smith, M., Smith, L., Parkin, J.L., 1994. The pitch of electrically presented sinusoids. J. Acoust. Soc. Am. 95, 1677– 1679. Greenwood, D.D., 1961. Critical Bandwidth and the frequency coordinates of the basilar membrane. J. Acoust. Soc. Am. 33, 1344– 1356. Greenwood, D.D., 1990. A cochlear frequency-position function for several species—29 years later. J. Acoust. Soc. Am. 87, 2592–2605. Hochmair, I., Arnold, W., Nopp, P., Jolly, C., Muller, J., Roland, P., 2003. Deep electrode insertion in cochlear implants: apical morphology, electrodes and speech perception results. Acta Otolaryngol 123, 612–617. Otte, J., Schuhknecht, H.F., Kerr, A.G., 1978. Ganglion cell populations in normal and pathological human cochleae. Implications for cochlear implantation. Laryngoscope 88, 1231–1246. Moore, B.C., Huss, M., Vickers, D.A., Glasberg, B.R., Alcantara, J.I., 2000. A test for the diagnosis of dead regions in the cochlea. Br. J. Audiol. 34, 205–224. Moore, B.C., Alcantara, J.I., 2001. The use of psychophysical tuning curves to explore dead regions in the cochlea. Ear Hear. 22, 268–278.
42
U. Baumann, A. Nobbe / Hearing Research 213 (2006) 34–42
Rattay, F., Lutter, P., Felix, H., 2001a. A model of the electrically excited human cochlear neuron. I. Contribution of neural substructures to the generation and propagation of spikes. Hear. Res. 153, 43–63. Rattay, F., Leao, R.N., Felix, H., 2001b. A model of the electrically excited human cochlear neuron. II. Influence of the three-dimensional cochlear structure on neural excitability. Hear. Res. 153, 64–79. Spoendlin, H., Schrott, A., 1989. Analysis of the human auditory nerve. Hear. Res. 43, 25–38. Turner, C., Burns, E.M., Nelson, D.A., 1983. Pure tone pitch perception and low frequency hearing loss. J. Acoust. Soc. Am. (73), 966–975.
Yukawa, K., Cohen, L., Blamey, P., Pyman, B., Tungvachirakul, V., O’Leary, S., 2004. Effects of insertion depth of cochlear implant electrodes upon speech perception. Audiol. Neurootol. 9, 163–172. Zeng, F.G., 2002. Temporal pitch in electric hearing. Hear Res. 174, 101– 106. Zierhofer, C.M., Hochmair-Desoyer, I., Hochmair, E.S., 1995. Electronic design of a cochlear implant for multichannel high-rate pulsatile stimulation strategies. IEEE Transact. Rehab. Engin. 3, 112–116. Zwicker, E., Fastl, H., 1999. Psychoacoustics—Facts and Models, Second ed. Springer, Berlin, Germany.