- Email: [email protected]
Mismatch negativity: a tool for the assessment of stimuli discrimination in cochlear implant subjects
Clinical Neurophysiology 111 (2000) 743±751
Technical report
www.elsevier.com/locate/clinph
Mismatch negativity: a tool for the assessment of stimuli discrimination in cochlear implant subjects Jocelyne Wable a,*, Thierry van den Abbeele b, SteÂphane GalleÂgo c, Bruno Frachet a a
Centre de Recherche et d'IngeÂnierie des Sciences et Techniques de l'Audition et du Langage (CRISTAL), Universite Paris 13, HoÃpital Avicenne, Service ORL, 125 rue de Stalingrad, 93009 Bobigny, France b HoÃpital Robert DebreÂ, Service ORL, 48 Boulevard Serurier, 75019 Paris, France c MXM 2720 Chemin St. Bernard, F-06224 Vallauris Cedex, France Accepted 9 November 1999
Abstract Objectives: The performance of cochlear implants varies among users. This variability may be due to the ability to process auditory information. The mismatch negativity should provide an index of discrimination in cochlear implantees (Kraus N, McGee T, Carrell T, Sharma A. Neurophysiologic bases of speech discrimination. Ear Hear. 1995;16:19±37). Our aim was to analyze MMN in cochlear implant (Digisonic) subjects to assess electrode discrimination and to study the relationship between MMN and speech performance. Methods: The mismatch was determined by stimulating three pairs of different electrodes. Two sessions were performed with both standard and deviant stimuli reversed. Speech recognition abilities were evaluated using 4 speech tests. The statistics included the results of 6 subjects. They indicated that MMN may be obtained when stimulating two different electrodes. A difference occurred between standard and deviant stimuli within a session but also when the response to the deviant stimulus was compared to the response of the same stimulus in a standard condition, validating the discrimination process. MMN latency was about 140 ms, and amplitude about 22.8 mV. No differences were shown with respect to electrode spacing. No relationship between MMN and speech performance was found. A clinical application of this method might be to assess the auditory processing of electrical stimuli in congenitally deaf subjects at the pre-implantation stage. q 2000 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Event-related electrically evoked auditory potentials; Mismatch negativity; Cochlear implant
1. Introduction A cochlear implant consists of an electrode array placed in the scala tympani. It delivers electrical stimulations to the remaining auditory nerve ®bers of deaf patients. Non-invasive and objective tools have been developed to assess auditory nerve and central auditory system function before and after implantation. Auditory evoked potentials have been widely used for evaluating the excitability of the auditory nerve before implantation and the integrity of the implant, as well as for setting the device parameters (Pelizzone et al., 1989; Oviatt and Kileny, 1991). The success of cochlear implantation depends on the speech intelligibility and speech production abilities of the patient. Implant users vary greatly in their performance. The poorer performers could bene®t from optimization of parameters settings at an individual level. Differences between * Corresponding author. Tel.: 133-1-4348-8664; fax: 133-1-48955202. E-mail address: [email protected] (J. Wable)
subjects may not be entirely explained in terms of device ear status or number of surviving neurons. Abilities to adapt to electrical stimulation, discrimination of stimuli, as well as cognitive capacities may be involved. Recently, relationships were found between electrically evoked brain-stem responses (EABR) and speech performances (Herman and Thornton, 1992; Groenen et al., 1996; GalleÂgo et al., 1998). The ability in discriminating small acoustic differences is very important for the perception and processing of speech signals. The mismatch negativity (MMN) method is an objective tool that provides a measure of automatic stimuli discrimination (NaÈaÈtaÈnen et al., 1978; NaÈaÈtaÈnen and Michie, 1979; Sams et al., 1985; NaÈaÈtaÈnen and Picton, 1987). This method seems likely to be useful in the study of neurophysiologic processes of stimulus change occurring during normal perception or in pathological situations (Kraus et al., 1993a,1995a). MMN should provide an index of discrimination abilities. Kraus et al. (1993b) demonstrated that MMN may be obtained in patients with cochlear implants. With 100 ms
1388-2457/00/$ - see front matter q 2000 Elsevier Science Ireland Ltd. All rights reserved. PII: S 1388-245 7(99)00298-9
CLINPH 99530
744
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
synthesized speech stimuli /da/ and /ta/, they observed some response patterns similar to those obtained in normal-hearing subjects. Ponton and Don (1995) demonstrated that the activation of two different electrode pairs along the implant array elicited an MMN response. They found, with a stimulation consisting of one pulse, a larger and earlier MMN response for the apical electrodes than for the basal ones. Kileny et al. (1997) investigated late and cognitive evoked potentials in children with cochlear implants. They found that for the frequency contrast, shorter and stronger P3 and MMN were associated with high sentence recognition scores. Cochlear implantees may bene®t from this technique in terms of electrode functioning, auditory system integrity, discrimination abilities (Kraus et al., 1993b; Ponton and Don, 1995). Moreover. it may assess evolution of capacities along training (Kraus et al., 1995b,1998; Tremblay et al., 1997). Our aim was to analyze MMN in cochlear implant subjects to assess electrode discrimination along the implant array. The main objective was to compare event-related auditory evoked potentials (ERPs) waves and difference waves as a function of the following experimental parameters: occurrence of the stimulus and stimulated electrodes. A second objective was to study the relationship between MMN and speech perception performance. Our experiments were performed with users of Digisonic DX10 multi-electrode cochlear implants.
2. Material and methods 2.1. Subjects Eight subjects were included in the experiment, ranging in age from 40 to 71 years, having been deaf from 1 to 13 years, and having been implanted from 4 to 36 months. They were all post-lingnally deafened. Details concerning etiologies, length of deafness, implant experience and stimulus parameters used in this study are displayed in Table 1. All subjects wore a Digisonic DX 10 cochlear implant (MXM).
2.2. Digisonic DX 10 system The DX 10 Digisonic processor signal processing consists (Beliaeff et al., 1994) in assigning to the 15 implanted electrodes 15 frequency bands ranging from 100 to 7800 Hz. The signal spectral analysis relies on a Fast Fourier Transform. Each sweep, six electrodes are stimulated, corresponding to the louder bands of the spectral analysis. The amplitude of the spectral peaks determines the amount of current delivered to the corresponding electrode, i.e. the pulse duration. The frequency bands assigned to these electrodes are spaced with a linear or a logarithmic scale. Overlapping or reassigning the frequency bands is possible. The highest pitch corresponds to the ®rst electrode and the lowest one to the 15th. 2.3. Event-related evoked potentials During measurements the subjects were seated in a comfortable chair and were asked to relax with their eyes closed, but not to go to sleep. They were instructed not to pay attention to the stimuli. 2.4. Stimulation The stimuli did not go through the implant processor but were directly delivered through a Digisonic interface, the Digistim (Fig. 1). Stimuli were presented using a passive oddball paradigm. Attention is directed away from the acoustic stimuli with an explicit instruction to ignore all auditory stimuli. Standard (n 850) and deviant (n 150) stimuli were presented in a pseudo-random sequence with at least three standard stimuli between two deviant ones. The mismatch was elicited by stimulating two different electrodes. Three pairs of electrodes were tested: (13, 12), (13, 10), (13, 8). In order to avoid overlapping of the MMN with the waveform components related to the physical difference between stimuli, two sessions were performed for each pair: the ®rst session with stimulus 1 the repetitive standard and stimulus 2 the randomly interspersed deviant one, and the contra protocol with both stimuli reversed (Ponton and Don, 1995). The order of the sessions was
Table 1 Age, etiology, deprivation duration, implantation duration and electrodes tested for the subjects evaluated. Pulse durations were set to deliver stimuli at about 70% of the dynamic range Subject
Age
Etiology
Total deaf duration (years)
Implantation duration (months)
Electrode pairs tested [electrode:pulse duration (ms)]
JL AN HB SO MI AZ SP AC
59 40 69 71 55 53 65 59
Otosclerosis Meningitis Chronic otitis Otosclerosis Sudden deafness Sudden meningitis Sudden deafness Progressive deafness
4 6 2 1 1 7 13 3
8 4 21 8 6 12 28 14
13:33, 12:35, 10:31, 8:26 13:64, 12:66, 10:68, 8:68 13:22, 12:26, 10:29, 8:36 13:31, 10:44, 8:38 13:16, 12:18, 10:18,8:16 13::14, 12:14, 10:16, 8:18 13:24, 12:28 13:23, 12:25, 10:24, 8:27
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
Fig. 1. Morphology of the Digisonic cochlear implant stimulation. In our experiment the inter-pulse interval is set at 3.3 ms, and the inter-stimulus interval at 1 s. Pulse amplitude is set at its ®tting value and pulse duration corresponds to 70% of the dynamic range.
randomized between subjects. The order of pairs and electrodes was randomized as well. Both stimuli were 100 ms long, with 3.3 ms inter-pulse duration (300 Hz). The inter-stimuli interval was 1 s (Fig 1). Stimulus level was set at about 70% of the dynamic range. The subjects discriminated these stimuli behaviorally. An equal level of loudness of the electrodes 13±8 was researched before testing at this adjusted level. At the beginning of the electrophysiological measurements it was veri®ed that the subjects detected all stimuli and that the loudness was comfortable. 2.5. Recording The potentials were recorded from forehead/mastoid contralateral to the implant with the ground electrode on the chin. In this work, we used only this recording position, which is known to be a good location to measure MMN. The recordings were controlled using a home-made software. The recording window included 500 ms of post-stimulus time with 250 sampling points per sweep. The responses were ampli®ed and analog bandpass ®ltered on-line from 0.1 Hz to 30 Hz with a CyberAmp 320 (Axon Instruments, Foster City). Using an automatic artifact rejection algorithm, the sweeps containing activity that exceeded ^75 mV were excluded from subsequent averaging. ERPs from deviant and standard stimuli were averaged separately for on-line visualization of the responses. However, each ERP was also stored separately to allow off-line analysis. 2.6. Derivation of the MMN Averaged ERPs from standard and from deviant stimuli were obtained from the recording. ERPs from standard stimuli following deviant ones were not included in the averaging. The MMN. which overlaps the N1 and P2 event-related potentials components, was measured from a difference wave obtained by subtracting the response evoked by the standard stimulus from the response evoked by the deviant stimulus. Two different waveforms were obtained: the intra-session in which stimuli were physically different (two electrodes stimulated) and the inter-session
745
involving physically identical stimuli presented in different contexts (randomized and rare in one session vs. repetitive and frequent in the other one). The resulting waveforms should re¯ect stimulus-related activities and contextual differences for the ®rst wave, and only contextual differences for the second one. However, the contextual difference means not only deviant vs. standard stimulation but also repetition rates and number of waveforms involved (Picton et al., 1974). A lower N1 amplitude is to be expected in the case of a standard stimulation compared with a deviant one. 2.7. Speech recognition Four tests were administered to evaluate speech recognition abilities (Table 2). For each test, the stimuli were presented in random order, at a comfortable listening level with the processor set at a normal-use setting. No lip-reading help and no feedback were provided. The VCV test assesses the ability to identify consonants. Sixteen consonants were presented through a computer with a 16-bit sound card, three times each, using the vowel /a/. The subject had to choose between 16 responses presented on the computer screen. The Lafon test assesses recognition of monosyllabic words. A list contains 17 words of three phonemes. A list of 75 familiar monosyllabic words was also presented. Thirty-®ve short sentences with 119 familiar key words were used to assess phonetic and cognitive skills. The stimuli were presented once, in an open-set condition. The percentage of consonants, phonemes, words and key words recognition was evaluated, respectively. 2.8. Statistics Statistics only concerned the 6 subjects for whom the three pair discriminations were evaluated. The P1-N1-P2 wave was identi®ed for each evoked potential. The N1 latencies were measured at the trough. The amplitudes were evaluated based upon the difference between P1 peak and N1 trough. The MMN waves were identi®ed in both difference waves. The MMN latencies were measured at the trough. If there were two troughs in the difference waveform, one at the N1 latency and one later, the second one was taken as the MMN (Scherg et al., 1989). The amplitudes were evaluated based upon the difference Table 2 Speech test performance for the 6 subjects included in the statistical analysis Subject
VCV
Words
Sentences
Lafon IC
AC AN AZ LIB JL MI
42 27 48 65 46 42
57 20 45 41 49 11
80 21 58 81 84 21
82 50 63 75 45 59
746
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
between MMN onset and MMN trough. Frequency of occurrence, session and electrode effects on latency or amplitude values were tested using ANOVAs. However, missing MMNs (6 over 72 recordings) do not allow to perform two way ANOVA tests on MMN latency or amplitude values. Moreover, the morphologies of ERP evoked by the standard and deviant stimuli were examined. Ten periods of 50 ms long were de®ned. In each period, the responses to standard and deviant stimuli were compared as a function of time. For each electrode, for each temporal domain, the ERPs evoked by the standard stimuli or by the deviant ones were compared using an ANOVA analysis with subjects, condition, and occurrence as factors. The factor condition represented a value of 6 dimensions (3 electrodes £ 2 sessions). Correlations between MMN latency or amplitude and speech recognition were tested using Pearson's coef®cient.
3. Results All subjects perceived the differences between the stimuli. The data set consisted of responses from 100 to 150 deviant stimuli and from 484 to 729 standard stimuli. 3.1. Event-related evoked potential peaks analysis Fig. 2a shows the averaged response to the standard stimulus for a representative subject. The classic P1-N1P2 wave is obvious, with N1 latency at about 100 ms. Fig. 2b displays the grand mean averaged responses from the stimulation of electrode 13, 12, 10 and 8 as standard stimuli. Statistical analysis (occurrence £ electrode ANOVA on repeated measures, N 6) did not show any effect of occurrence or electrode on the N1 or P2 latencies. N1 occurs
Fig. 2. (a) Averaged response of the standard stimulus for a representative subject. The classic P1-N1-P2 wave is obvious, with N1 at about 100 ms. (b) Grand mean averaged responses from stimulation of electrode 13, 12, 10 and 8 as standard stimuli.
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
around 117 ms (SD 12) and P2 around 217 ms (SD 21). Fig. 3 shows the mean P1N1 amplitude (a) and the mean N1P2 amplitude (b) for the standard and deviant stimuli. As no effect of the session was found, data relative to the same occurrence were gathered between sessions. The occurrence £ electrode pair ANOVA on repeated measures on the averaged values demonstrates an effect of occurrence (F1 19:4, P1 0:007; F2 10:7, P2 0:022) and an effect of electrode pair (F1 4:4, P1 0:042) with interaction (F1 5:5, P1 0:024; F2 5:4, P2 0:026) on the P1N1 and N1P2 amplitudes, respectively. The posthoc analysis, using the Student-Newman-Keuls test shows a difference between amplitude of deviants of pairs 2 and 3 vs. pair 1. For standard stimuli no differences were found. mean P1N1 amplitude equals 24.3 mV (SD 1.3), and mean N1P2 amplitude equals 4.7 mV (SD 1.7 mV). 3.2. Morphology analysis The analysis of variance with condition and occurrence as factors indicates a condition effect, above all from 100 to 450 ms, an occurrence effect from 100 to 300 ms with a
747
lower value for deviant stimuli in the 100±200 ms interval and a higher value in the 200±300 ms interval, and no interaction between the condition and the occurrence. Fig. 4 displays the mean ERP gathering all condition values for the standard and deviant stimuli by 50 ms periods. 3.3. Difference waves analysis 3.3.1. Intra-session difference Fig. 5a shows the averaged responses to the standard stimulus (stimulation of electrode 13), the averaged responses to the deviant stimulus (stimulation of electrode 12), and the difference wave for a representative subject. The MMN de¯ection is obvious in the response to the deviant stimulus, overlapping the N1-P2 wave. Fig. 5b displays the mean MMN intra-session difference wave (N 6) for the 3 pairs of electrodes. The paired t tests did not show any differences between sessions on MMN latency (N 16, mean 145.4 ms, SD 20:4 for the ®rst session; mean 134.4 ms, SD 26:1 for the second session) or amplitude (N 16, mean 23.1, SD 2:4 for the ®rst session; mean 22.8, SD 1:0 for the second session). However, the test power was
Fig. 3. Mean P1N1 amplitude (a) and mean N1P2 amplitude (b) for standard and deviant stimuli in respect to electrode pairs.
748
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
Fig. 4. Mean ERP values for standard and deviant stimuli by 50 ms periods. Asterisks indicate the signi®cant differences for each interval. *P , 0:05, **P , 0:01, ***P , 0:001.
very low. One way ANOVAs on repeated measures did not show any differences between electrodes on the latency or amplitude values. The mean MMN latency is 139 ms (SD 23). Its amplitude equals 22.9 mV (SD 1:8). Fig 4b displays the mean MMN intra-session difference wave (N 6) for the 3 pairs of electrodes.
4. Discussion
3.3.2. Inter-session difference Fig. 6a shows the averaged responses to the standard stimulus (stimulation of electrode 12), the averaged responses to the deviant stimulus (stimulation of electrode 12), and the difference wave for a representative subject. A negativity is obvious in this difference wave. One way ANOVAs on repeated measures did not show any differences between electrodes on the latency or amplitude values. The mean MMN latency of 142 ms (SD 22) and amplitude of 22.8 mV (SD 1.4) are the same as the MMN in the intra-session difference. Fig. 6b shows the mean MMN inter-session difference wave (N 6) for the 3 pairs of electrodes.
The aim of the study was to characterize the MMN recorded from electrical stimulation of Digisonic cochlear implant subjects. A multi-electrode implant consists of several electrodes spaced along the cochlea. By stimulating neural ganglia at a speci®c location through these implanted electrodes, some sensation of pitch may be restored. Activation of adjacent electrode can be discriminated behaviorally. The recording of MMN by stimulating two different electrodes may be related to the measure of frequency discrimination in normal-hearing subjects. Ponton and Don (1995), comparing responses to electrode pair stimulation in cochlear implant subjects and responses to tone burst stimulation in normal-hearing subjects, found earlier MMN waves with electrical stimulation (169 and 155 ms for 2 and 1 kHz stimuli, respectively, vs. 114 and 89 ms for the basal and apical pairs, respectively) which may be explained by a more synchronized activation of neurons in the case of electrical stimulation. The stimulation of the apical located electrode pairs lead to larger responses than
3.4. Speech performance
Table 3 Characteristics and evaluation criterion of speech tests.
Table 3 reports the speech tests results. The correlation analysis did not emphasize a relationship between MMN amplitude or latency and speech performance, regardless what the speech test was. However, the low power of the correlation test did not allow to conclude to an absence of relationship.
Test
Characteristics
Level of evaluation
VCV Lafon
16 consonants, 3 times each 17 monosyllabic words of three phonemes 75 familiar monosyllabic words 35 sentences, 119 key words
Consonant/48 Phonemes/51
Words Sentences
Words/75 Key words/119
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
749
Fig. 5. (a) Averaged responses of the stimulation of electrode 12 as standard stimulus, as deviant stimulus, and difference wave for a representative subject. The MMN de¯ection is obvious in the response of the deviant stimulus, overlapping the N1P2 wave. (b) Grand mean MMN intra-session difference wave (N 6) for the 3 pairs of electrodes.
stimulation of the basely located electrode pairs. We found a mean MMN latency of about 140 ms for apical electrode stimulation. The difference in latency with the Ponton and Don (1995) study can be due to the difference in the stimulation characteristics between the two cochlear implants: adjacent electrode pairs in the Ponton and Don study, one array electrode vs. adjacent electrodes and common electrode in the Digisonic implant in the present study. The latter mode of stimulation may lead to a larger spread of current and a less synchronized activation. We found a shorter MMN latency than Kraus et al. (1993b). With 100 ms synthesized stimuli /da/ (standard) and /ta/ (deviant), they found a MMN latency of 220 ms, a duration of 121 ms, and an amplitude from onset to trough of 21.7 mV. A similar MMN occurs when the /ta/-alone waveform was taken as standard reference. The difference between the
two studies ± earlier and greater MMN in the present ± may be explained by the differences in stimuli used: phonemes in free ®eld presentation vs. direct and localized electrode stimulation which may provide a more synchronized and replicable activation of the neurons. Indeed, stimulating bypassing the speech processor allows precise and replicable activation of speci®c electrodes along the implant array. Moreover, the delay observed in case of speech discrimination may be the consequence of a longer speech processing compared to the non speech processing. The difference in MMN amplitude may also be due to the difference in the recording electrode position: Fz/earlobe in the Kraus et al. (1993b) study, forehead/mastoid in this one. Several studies showed that MMN peak latency increases as the standard deviant discrimination becomes more dif®cult (Sams et al., 1985; Kaukoranta et al., 1989; Scherg et
750
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
Fig. 6. (a) Averaged responses of the standard stimulus (stimulation of electrode 12), of the deviant stimulus (stimulation of electrode 12), and difference wave for a representative subject. A negativity is obvious in this difference wave. (b) Mean MMN intra-session difference wave (N 6) for the 3 pairs of electrodes.
al., 1989). The absence of electrode effect on MMN latency or amplitude suggests that the three tasks were as dif®cult, i.e. electrode 8 is not easier to discriminate from electrode 13 than electrode 12. The evoked potentials recordings aims are also to try to predict speech recognition performance. Using the Digisonic cochlear implant, GalleÂgo et al. (1998) found a correlation between brain-stem evoked potential wave III to wave V delay and speech recognition score (r 0:69; P , 0:005) with a shorter interval linked with a higher recognition score. Groenen et al. (1997) found a relationship between phoneme recognition and middle latency responses. Makh-
doum et al. (1997) demonstrated a positive correlation between cortical response (N1P2 amplitude) and speech performance in cochlear implant users, and a negative correlation with P2 latency. Kileny et al. (1997) investigated late and cognitive evoked potentials in children with cochlear implants. They studied the relationship between amplitude and latency of the cognitive response and speech recognition abilities. The latency tended to be shorter for the frequency contrast than for the loudness contrast which in turn tended to be shorter than for the speech contrast. In frequency, the discrimination showed shorter and stronger P3 and MMN that were associated with high sentence recognition scores.
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
However, our study failed to show a relationship between latency or amplitude of MMN and speech recognition abilities. This might be due to the low power of the statistical test, linked with the low number of subjects, or to the absence of such a relationship. A more extensive study could be planned to examine this using the mean and variance values of this one as reference. A way to study this issue more deeply could be in applying this method to assess the auditory processing of electrical stimuli in congenitally deaf subjects at a preimplantation stage, and even to try to predict performance. Acknowledgements The authors wish to thank the Assistance Publique±HoÃpitaux de Paris for their grant (PHRC no. 094034) and the MXM Laboratory for their ®nancial support. The authors also wish to thank D. Moyse for helping with the statistics, S. Labassi for editing the manuscript, and the patients who volunteered for the experiments. References Beliaeff M, Dubus P, Levean J-M, Repetto J-C, Vincent P. Sound signal processing and stimulation coding of the Digisonic DX 10 15-channel cochlear implant. In: Hochmair-Desoyer IJ, Hochmair ES, editors. Advances in cochlear Implants, Vienna: Mainz, 1994. pp. 198±203. GalleÂgo S, Frachet B, Micheyl C, Tray E, Collet L. Cochlear implant performance and electrically evoked auditory brain-stem response characteristics. Electroenceph clin Neurophysiol 1998;108:521±525. Groenen PAP, Makhdoum M, Van Den Brink JL, Stollman MHP, Snik AFM, van den Broek P. The relation between electric auditory brain stem and cognitive responses and speech perception in cochlear implant users. Acta Otolaryngol (Stockh) 1996;116:785±790. Groenen PAP, Snik AFM, van den Broek P. Electrically evoked auditory middle latency responses versus perception abilities in cochlear implant users. Audiology 1997;36:83±97. Herman B, Thornton A. Electrically-evoked auditory brain-stem responses in cochlear implanted subjects (abstr) Iowa City, IA. Second International Cochlear Implant Symposium 1992;:57. Kaukoranta E, Sams M, Hari R, HaÈmaÈlaÈinen M, NaÈaÈtaÈnen R. Reactions of human auditory cortex to a change intone duration. Hear Res 1989;41:15±22.
751
Kileny PR, Boerst A, Zwolan T. Cognitive evoked potentials to speech and tonal stimuli in children with implants. Otolaryngol Head Neck Surg 1997;117:161±169. Kraus N, Micco AG, Koch DB, McGee T, Carrell T, Sharma A, Wiet RJ, Weingarten CZ. The mismatch negativity cortical evoked potential elicited by speech in cochlear-implant users. Hear Res 1993a;65:118± 124. Kraus N, McGee T, Micco A, Sharma A, Carrel T, Nicol T. Mismatch negativity in school-age children to speech stimuli that are just perceptibly different. Electroenceph clin Neurophysiol 1993b;88:123± 130. Kraus N, McGee T, Carrell T, Sharma A. Neurophysiologic bases of speech discrimination. Ear Hear 1995a;16:19±37. Kraus N, McGee T, Carrel T, King C, Tremblay K, Nicol T. Central auditory system plasticity associated with speech discrimination training. J Cogn Neurosci 1995b;7:25±32. Kraus N, McGee TJ, Koch DB. Speech sound representation, perception, and plasticity: a neurophysiologic perspective. Audiol Neuro-Otol 1998;3:168±182. Makhdoum MJ, Groenen PAP, Snik AEM, van den Broek P. Intra- and interindividual correlation between auditory evoked potentials and speech perception in cochlear implant users. Scand Audiol 1997;27:13±20. NaÈaÈtaÈnen R, Michie PT. Early selective attention effects on the evoked potential. A critical review and reinterpretation. Biol Psychol 1979;8:81±136. NaÈaÈtaÈnen R, Picton T. The N1 wave of the human electric and magnetic response to sound: a review and an analysis of the component structure. Psychophysiology 1987;24:375±425. NaÈaÈtaÈnen R, Gaillard AWK, MaÈntysalo S. Early selective attention effect on evoked potential reinterpreted. Acta Psychol 1978;42:313±329. Oviatt D, Kileny P. Auditory event-related potentials elicited from cochlear implant recipients and hearing subjects. Am J Audiol 1991;1:48±55. Pelizzone M, Kasper A, Montandon P. Electrically evoked responses in cochlear implant patients. Audiology 1989;28:230±238. Picton TW, Hillyard SA, Krausz HT, Galambos R. Human auditory evoked potentials I: evaluation of components. Electroenceph clin Neurophysiol 1974;36:179±190. Ponton CW, Don M. The mismatch negativity in cochlear implant users. Ear Hear 1995;16:131±146. Sams M, HaÈmaÈlaÈinen M, Antervo A, Kaukoranta E, Reinikainen K, Hari R. Cerebral neuromagnetic responses evoked by short auditory stimuli. Electroenceph clin Neurophysiol 1985;61:254±266. Scherg M, Vajsar J, Picton TW. A source analysis of the late human auditory evoked potentials. Cogn Neurosci 1989;1:336±355. Tremblay K, Kraus N, Carrell TD, McGee T. Central auditory system plasticity: generalization to novel stimuli following training. J Acoust Soc Am 1997;102:3762±3773.
Technical report
www.elsevier.com/locate/clinph
Mismatch negativity: a tool for the assessment of stimuli discrimination in cochlear implant subjects Jocelyne Wable a,*, Thierry van den Abbeele b, SteÂphane GalleÂgo c, Bruno Frachet a a
Centre de Recherche et d'IngeÂnierie des Sciences et Techniques de l'Audition et du Langage (CRISTAL), Universite Paris 13, HoÃpital Avicenne, Service ORL, 125 rue de Stalingrad, 93009 Bobigny, France b HoÃpital Robert DebreÂ, Service ORL, 48 Boulevard Serurier, 75019 Paris, France c MXM 2720 Chemin St. Bernard, F-06224 Vallauris Cedex, France Accepted 9 November 1999
Abstract Objectives: The performance of cochlear implants varies among users. This variability may be due to the ability to process auditory information. The mismatch negativity should provide an index of discrimination in cochlear implantees (Kraus N, McGee T, Carrell T, Sharma A. Neurophysiologic bases of speech discrimination. Ear Hear. 1995;16:19±37). Our aim was to analyze MMN in cochlear implant (Digisonic) subjects to assess electrode discrimination and to study the relationship between MMN and speech performance. Methods: The mismatch was determined by stimulating three pairs of different electrodes. Two sessions were performed with both standard and deviant stimuli reversed. Speech recognition abilities were evaluated using 4 speech tests. The statistics included the results of 6 subjects. They indicated that MMN may be obtained when stimulating two different electrodes. A difference occurred between standard and deviant stimuli within a session but also when the response to the deviant stimulus was compared to the response of the same stimulus in a standard condition, validating the discrimination process. MMN latency was about 140 ms, and amplitude about 22.8 mV. No differences were shown with respect to electrode spacing. No relationship between MMN and speech performance was found. A clinical application of this method might be to assess the auditory processing of electrical stimuli in congenitally deaf subjects at the pre-implantation stage. q 2000 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Event-related electrically evoked auditory potentials; Mismatch negativity; Cochlear implant
1. Introduction A cochlear implant consists of an electrode array placed in the scala tympani. It delivers electrical stimulations to the remaining auditory nerve ®bers of deaf patients. Non-invasive and objective tools have been developed to assess auditory nerve and central auditory system function before and after implantation. Auditory evoked potentials have been widely used for evaluating the excitability of the auditory nerve before implantation and the integrity of the implant, as well as for setting the device parameters (Pelizzone et al., 1989; Oviatt and Kileny, 1991). The success of cochlear implantation depends on the speech intelligibility and speech production abilities of the patient. Implant users vary greatly in their performance. The poorer performers could bene®t from optimization of parameters settings at an individual level. Differences between * Corresponding author. Tel.: 133-1-4348-8664; fax: 133-1-48955202. E-mail address: [email protected] (J. Wable)
subjects may not be entirely explained in terms of device ear status or number of surviving neurons. Abilities to adapt to electrical stimulation, discrimination of stimuli, as well as cognitive capacities may be involved. Recently, relationships were found between electrically evoked brain-stem responses (EABR) and speech performances (Herman and Thornton, 1992; Groenen et al., 1996; GalleÂgo et al., 1998). The ability in discriminating small acoustic differences is very important for the perception and processing of speech signals. The mismatch negativity (MMN) method is an objective tool that provides a measure of automatic stimuli discrimination (NaÈaÈtaÈnen et al., 1978; NaÈaÈtaÈnen and Michie, 1979; Sams et al., 1985; NaÈaÈtaÈnen and Picton, 1987). This method seems likely to be useful in the study of neurophysiologic processes of stimulus change occurring during normal perception or in pathological situations (Kraus et al., 1993a,1995a). MMN should provide an index of discrimination abilities. Kraus et al. (1993b) demonstrated that MMN may be obtained in patients with cochlear implants. With 100 ms
1388-2457/00/$ - see front matter q 2000 Elsevier Science Ireland Ltd. All rights reserved. PII: S 1388-245 7(99)00298-9
CLINPH 99530
744
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
synthesized speech stimuli /da/ and /ta/, they observed some response patterns similar to those obtained in normal-hearing subjects. Ponton and Don (1995) demonstrated that the activation of two different electrode pairs along the implant array elicited an MMN response. They found, with a stimulation consisting of one pulse, a larger and earlier MMN response for the apical electrodes than for the basal ones. Kileny et al. (1997) investigated late and cognitive evoked potentials in children with cochlear implants. They found that for the frequency contrast, shorter and stronger P3 and MMN were associated with high sentence recognition scores. Cochlear implantees may bene®t from this technique in terms of electrode functioning, auditory system integrity, discrimination abilities (Kraus et al., 1993b; Ponton and Don, 1995). Moreover. it may assess evolution of capacities along training (Kraus et al., 1995b,1998; Tremblay et al., 1997). Our aim was to analyze MMN in cochlear implant subjects to assess electrode discrimination along the implant array. The main objective was to compare event-related auditory evoked potentials (ERPs) waves and difference waves as a function of the following experimental parameters: occurrence of the stimulus and stimulated electrodes. A second objective was to study the relationship between MMN and speech perception performance. Our experiments were performed with users of Digisonic DX10 multi-electrode cochlear implants.
2. Material and methods 2.1. Subjects Eight subjects were included in the experiment, ranging in age from 40 to 71 years, having been deaf from 1 to 13 years, and having been implanted from 4 to 36 months. They were all post-lingnally deafened. Details concerning etiologies, length of deafness, implant experience and stimulus parameters used in this study are displayed in Table 1. All subjects wore a Digisonic DX 10 cochlear implant (MXM).
2.2. Digisonic DX 10 system The DX 10 Digisonic processor signal processing consists (Beliaeff et al., 1994) in assigning to the 15 implanted electrodes 15 frequency bands ranging from 100 to 7800 Hz. The signal spectral analysis relies on a Fast Fourier Transform. Each sweep, six electrodes are stimulated, corresponding to the louder bands of the spectral analysis. The amplitude of the spectral peaks determines the amount of current delivered to the corresponding electrode, i.e. the pulse duration. The frequency bands assigned to these electrodes are spaced with a linear or a logarithmic scale. Overlapping or reassigning the frequency bands is possible. The highest pitch corresponds to the ®rst electrode and the lowest one to the 15th. 2.3. Event-related evoked potentials During measurements the subjects were seated in a comfortable chair and were asked to relax with their eyes closed, but not to go to sleep. They were instructed not to pay attention to the stimuli. 2.4. Stimulation The stimuli did not go through the implant processor but were directly delivered through a Digisonic interface, the Digistim (Fig. 1). Stimuli were presented using a passive oddball paradigm. Attention is directed away from the acoustic stimuli with an explicit instruction to ignore all auditory stimuli. Standard (n 850) and deviant (n 150) stimuli were presented in a pseudo-random sequence with at least three standard stimuli between two deviant ones. The mismatch was elicited by stimulating two different electrodes. Three pairs of electrodes were tested: (13, 12), (13, 10), (13, 8). In order to avoid overlapping of the MMN with the waveform components related to the physical difference between stimuli, two sessions were performed for each pair: the ®rst session with stimulus 1 the repetitive standard and stimulus 2 the randomly interspersed deviant one, and the contra protocol with both stimuli reversed (Ponton and Don, 1995). The order of the sessions was
Table 1 Age, etiology, deprivation duration, implantation duration and electrodes tested for the subjects evaluated. Pulse durations were set to deliver stimuli at about 70% of the dynamic range Subject
Age
Etiology
Total deaf duration (years)
Implantation duration (months)
Electrode pairs tested [electrode:pulse duration (ms)]
JL AN HB SO MI AZ SP AC
59 40 69 71 55 53 65 59
Otosclerosis Meningitis Chronic otitis Otosclerosis Sudden deafness Sudden meningitis Sudden deafness Progressive deafness
4 6 2 1 1 7 13 3
8 4 21 8 6 12 28 14
13:33, 12:35, 10:31, 8:26 13:64, 12:66, 10:68, 8:68 13:22, 12:26, 10:29, 8:36 13:31, 10:44, 8:38 13:16, 12:18, 10:18,8:16 13::14, 12:14, 10:16, 8:18 13:24, 12:28 13:23, 12:25, 10:24, 8:27
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
Fig. 1. Morphology of the Digisonic cochlear implant stimulation. In our experiment the inter-pulse interval is set at 3.3 ms, and the inter-stimulus interval at 1 s. Pulse amplitude is set at its ®tting value and pulse duration corresponds to 70% of the dynamic range.
randomized between subjects. The order of pairs and electrodes was randomized as well. Both stimuli were 100 ms long, with 3.3 ms inter-pulse duration (300 Hz). The inter-stimuli interval was 1 s (Fig 1). Stimulus level was set at about 70% of the dynamic range. The subjects discriminated these stimuli behaviorally. An equal level of loudness of the electrodes 13±8 was researched before testing at this adjusted level. At the beginning of the electrophysiological measurements it was veri®ed that the subjects detected all stimuli and that the loudness was comfortable. 2.5. Recording The potentials were recorded from forehead/mastoid contralateral to the implant with the ground electrode on the chin. In this work, we used only this recording position, which is known to be a good location to measure MMN. The recordings were controlled using a home-made software. The recording window included 500 ms of post-stimulus time with 250 sampling points per sweep. The responses were ampli®ed and analog bandpass ®ltered on-line from 0.1 Hz to 30 Hz with a CyberAmp 320 (Axon Instruments, Foster City). Using an automatic artifact rejection algorithm, the sweeps containing activity that exceeded ^75 mV were excluded from subsequent averaging. ERPs from deviant and standard stimuli were averaged separately for on-line visualization of the responses. However, each ERP was also stored separately to allow off-line analysis. 2.6. Derivation of the MMN Averaged ERPs from standard and from deviant stimuli were obtained from the recording. ERPs from standard stimuli following deviant ones were not included in the averaging. The MMN. which overlaps the N1 and P2 event-related potentials components, was measured from a difference wave obtained by subtracting the response evoked by the standard stimulus from the response evoked by the deviant stimulus. Two different waveforms were obtained: the intra-session in which stimuli were physically different (two electrodes stimulated) and the inter-session
745
involving physically identical stimuli presented in different contexts (randomized and rare in one session vs. repetitive and frequent in the other one). The resulting waveforms should re¯ect stimulus-related activities and contextual differences for the ®rst wave, and only contextual differences for the second one. However, the contextual difference means not only deviant vs. standard stimulation but also repetition rates and number of waveforms involved (Picton et al., 1974). A lower N1 amplitude is to be expected in the case of a standard stimulation compared with a deviant one. 2.7. Speech recognition Four tests were administered to evaluate speech recognition abilities (Table 2). For each test, the stimuli were presented in random order, at a comfortable listening level with the processor set at a normal-use setting. No lip-reading help and no feedback were provided. The VCV test assesses the ability to identify consonants. Sixteen consonants were presented through a computer with a 16-bit sound card, three times each, using the vowel /a/. The subject had to choose between 16 responses presented on the computer screen. The Lafon test assesses recognition of monosyllabic words. A list contains 17 words of three phonemes. A list of 75 familiar monosyllabic words was also presented. Thirty-®ve short sentences with 119 familiar key words were used to assess phonetic and cognitive skills. The stimuli were presented once, in an open-set condition. The percentage of consonants, phonemes, words and key words recognition was evaluated, respectively. 2.8. Statistics Statistics only concerned the 6 subjects for whom the three pair discriminations were evaluated. The P1-N1-P2 wave was identi®ed for each evoked potential. The N1 latencies were measured at the trough. The amplitudes were evaluated based upon the difference between P1 peak and N1 trough. The MMN waves were identi®ed in both difference waves. The MMN latencies were measured at the trough. If there were two troughs in the difference waveform, one at the N1 latency and one later, the second one was taken as the MMN (Scherg et al., 1989). The amplitudes were evaluated based upon the difference Table 2 Speech test performance for the 6 subjects included in the statistical analysis Subject
VCV
Words
Sentences
Lafon IC
AC AN AZ LIB JL MI
42 27 48 65 46 42
57 20 45 41 49 11
80 21 58 81 84 21
82 50 63 75 45 59
746
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
between MMN onset and MMN trough. Frequency of occurrence, session and electrode effects on latency or amplitude values were tested using ANOVAs. However, missing MMNs (6 over 72 recordings) do not allow to perform two way ANOVA tests on MMN latency or amplitude values. Moreover, the morphologies of ERP evoked by the standard and deviant stimuli were examined. Ten periods of 50 ms long were de®ned. In each period, the responses to standard and deviant stimuli were compared as a function of time. For each electrode, for each temporal domain, the ERPs evoked by the standard stimuli or by the deviant ones were compared using an ANOVA analysis with subjects, condition, and occurrence as factors. The factor condition represented a value of 6 dimensions (3 electrodes £ 2 sessions). Correlations between MMN latency or amplitude and speech recognition were tested using Pearson's coef®cient.
3. Results All subjects perceived the differences between the stimuli. The data set consisted of responses from 100 to 150 deviant stimuli and from 484 to 729 standard stimuli. 3.1. Event-related evoked potential peaks analysis Fig. 2a shows the averaged response to the standard stimulus for a representative subject. The classic P1-N1P2 wave is obvious, with N1 latency at about 100 ms. Fig. 2b displays the grand mean averaged responses from the stimulation of electrode 13, 12, 10 and 8 as standard stimuli. Statistical analysis (occurrence £ electrode ANOVA on repeated measures, N 6) did not show any effect of occurrence or electrode on the N1 or P2 latencies. N1 occurs
Fig. 2. (a) Averaged response of the standard stimulus for a representative subject. The classic P1-N1-P2 wave is obvious, with N1 at about 100 ms. (b) Grand mean averaged responses from stimulation of electrode 13, 12, 10 and 8 as standard stimuli.
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
around 117 ms (SD 12) and P2 around 217 ms (SD 21). Fig. 3 shows the mean P1N1 amplitude (a) and the mean N1P2 amplitude (b) for the standard and deviant stimuli. As no effect of the session was found, data relative to the same occurrence were gathered between sessions. The occurrence £ electrode pair ANOVA on repeated measures on the averaged values demonstrates an effect of occurrence (F1 19:4, P1 0:007; F2 10:7, P2 0:022) and an effect of electrode pair (F1 4:4, P1 0:042) with interaction (F1 5:5, P1 0:024; F2 5:4, P2 0:026) on the P1N1 and N1P2 amplitudes, respectively. The posthoc analysis, using the Student-Newman-Keuls test shows a difference between amplitude of deviants of pairs 2 and 3 vs. pair 1. For standard stimuli no differences were found. mean P1N1 amplitude equals 24.3 mV (SD 1.3), and mean N1P2 amplitude equals 4.7 mV (SD 1.7 mV). 3.2. Morphology analysis The analysis of variance with condition and occurrence as factors indicates a condition effect, above all from 100 to 450 ms, an occurrence effect from 100 to 300 ms with a
747
lower value for deviant stimuli in the 100±200 ms interval and a higher value in the 200±300 ms interval, and no interaction between the condition and the occurrence. Fig. 4 displays the mean ERP gathering all condition values for the standard and deviant stimuli by 50 ms periods. 3.3. Difference waves analysis 3.3.1. Intra-session difference Fig. 5a shows the averaged responses to the standard stimulus (stimulation of electrode 13), the averaged responses to the deviant stimulus (stimulation of electrode 12), and the difference wave for a representative subject. The MMN de¯ection is obvious in the response to the deviant stimulus, overlapping the N1-P2 wave. Fig. 5b displays the mean MMN intra-session difference wave (N 6) for the 3 pairs of electrodes. The paired t tests did not show any differences between sessions on MMN latency (N 16, mean 145.4 ms, SD 20:4 for the ®rst session; mean 134.4 ms, SD 26:1 for the second session) or amplitude (N 16, mean 23.1, SD 2:4 for the ®rst session; mean 22.8, SD 1:0 for the second session). However, the test power was
Fig. 3. Mean P1N1 amplitude (a) and mean N1P2 amplitude (b) for standard and deviant stimuli in respect to electrode pairs.
748
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
Fig. 4. Mean ERP values for standard and deviant stimuli by 50 ms periods. Asterisks indicate the signi®cant differences for each interval. *P , 0:05, **P , 0:01, ***P , 0:001.
very low. One way ANOVAs on repeated measures did not show any differences between electrodes on the latency or amplitude values. The mean MMN latency is 139 ms (SD 23). Its amplitude equals 22.9 mV (SD 1:8). Fig 4b displays the mean MMN intra-session difference wave (N 6) for the 3 pairs of electrodes.
4. Discussion
3.3.2. Inter-session difference Fig. 6a shows the averaged responses to the standard stimulus (stimulation of electrode 12), the averaged responses to the deviant stimulus (stimulation of electrode 12), and the difference wave for a representative subject. A negativity is obvious in this difference wave. One way ANOVAs on repeated measures did not show any differences between electrodes on the latency or amplitude values. The mean MMN latency of 142 ms (SD 22) and amplitude of 22.8 mV (SD 1.4) are the same as the MMN in the intra-session difference. Fig. 6b shows the mean MMN inter-session difference wave (N 6) for the 3 pairs of electrodes.
The aim of the study was to characterize the MMN recorded from electrical stimulation of Digisonic cochlear implant subjects. A multi-electrode implant consists of several electrodes spaced along the cochlea. By stimulating neural ganglia at a speci®c location through these implanted electrodes, some sensation of pitch may be restored. Activation of adjacent electrode can be discriminated behaviorally. The recording of MMN by stimulating two different electrodes may be related to the measure of frequency discrimination in normal-hearing subjects. Ponton and Don (1995), comparing responses to electrode pair stimulation in cochlear implant subjects and responses to tone burst stimulation in normal-hearing subjects, found earlier MMN waves with electrical stimulation (169 and 155 ms for 2 and 1 kHz stimuli, respectively, vs. 114 and 89 ms for the basal and apical pairs, respectively) which may be explained by a more synchronized activation of neurons in the case of electrical stimulation. The stimulation of the apical located electrode pairs lead to larger responses than
3.4. Speech performance
Table 3 Characteristics and evaluation criterion of speech tests.
Table 3 reports the speech tests results. The correlation analysis did not emphasize a relationship between MMN amplitude or latency and speech performance, regardless what the speech test was. However, the low power of the correlation test did not allow to conclude to an absence of relationship.
Test
Characteristics
Level of evaluation
VCV Lafon
16 consonants, 3 times each 17 monosyllabic words of three phonemes 75 familiar monosyllabic words 35 sentences, 119 key words
Consonant/48 Phonemes/51
Words Sentences
Words/75 Key words/119
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
749
Fig. 5. (a) Averaged responses of the stimulation of electrode 12 as standard stimulus, as deviant stimulus, and difference wave for a representative subject. The MMN de¯ection is obvious in the response of the deviant stimulus, overlapping the N1P2 wave. (b) Grand mean MMN intra-session difference wave (N 6) for the 3 pairs of electrodes.
stimulation of the basely located electrode pairs. We found a mean MMN latency of about 140 ms for apical electrode stimulation. The difference in latency with the Ponton and Don (1995) study can be due to the difference in the stimulation characteristics between the two cochlear implants: adjacent electrode pairs in the Ponton and Don study, one array electrode vs. adjacent electrodes and common electrode in the Digisonic implant in the present study. The latter mode of stimulation may lead to a larger spread of current and a less synchronized activation. We found a shorter MMN latency than Kraus et al. (1993b). With 100 ms synthesized stimuli /da/ (standard) and /ta/ (deviant), they found a MMN latency of 220 ms, a duration of 121 ms, and an amplitude from onset to trough of 21.7 mV. A similar MMN occurs when the /ta/-alone waveform was taken as standard reference. The difference between the
two studies ± earlier and greater MMN in the present ± may be explained by the differences in stimuli used: phonemes in free ®eld presentation vs. direct and localized electrode stimulation which may provide a more synchronized and replicable activation of the neurons. Indeed, stimulating bypassing the speech processor allows precise and replicable activation of speci®c electrodes along the implant array. Moreover, the delay observed in case of speech discrimination may be the consequence of a longer speech processing compared to the non speech processing. The difference in MMN amplitude may also be due to the difference in the recording electrode position: Fz/earlobe in the Kraus et al. (1993b) study, forehead/mastoid in this one. Several studies showed that MMN peak latency increases as the standard deviant discrimination becomes more dif®cult (Sams et al., 1985; Kaukoranta et al., 1989; Scherg et
750
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
Fig. 6. (a) Averaged responses of the standard stimulus (stimulation of electrode 12), of the deviant stimulus (stimulation of electrode 12), and difference wave for a representative subject. A negativity is obvious in this difference wave. (b) Mean MMN intra-session difference wave (N 6) for the 3 pairs of electrodes.
al., 1989). The absence of electrode effect on MMN latency or amplitude suggests that the three tasks were as dif®cult, i.e. electrode 8 is not easier to discriminate from electrode 13 than electrode 12. The evoked potentials recordings aims are also to try to predict speech recognition performance. Using the Digisonic cochlear implant, GalleÂgo et al. (1998) found a correlation between brain-stem evoked potential wave III to wave V delay and speech recognition score (r 0:69; P , 0:005) with a shorter interval linked with a higher recognition score. Groenen et al. (1997) found a relationship between phoneme recognition and middle latency responses. Makh-
doum et al. (1997) demonstrated a positive correlation between cortical response (N1P2 amplitude) and speech performance in cochlear implant users, and a negative correlation with P2 latency. Kileny et al. (1997) investigated late and cognitive evoked potentials in children with cochlear implants. They studied the relationship between amplitude and latency of the cognitive response and speech recognition abilities. The latency tended to be shorter for the frequency contrast than for the loudness contrast which in turn tended to be shorter than for the speech contrast. In frequency, the discrimination showed shorter and stronger P3 and MMN that were associated with high sentence recognition scores.
J. Wable et al. / Clinical Neurophysiology 111 (2000) 743±751
However, our study failed to show a relationship between latency or amplitude of MMN and speech recognition abilities. This might be due to the low power of the statistical test, linked with the low number of subjects, or to the absence of such a relationship. A more extensive study could be planned to examine this using the mean and variance values of this one as reference. A way to study this issue more deeply could be in applying this method to assess the auditory processing of electrical stimuli in congenitally deaf subjects at a preimplantation stage, and even to try to predict performance. Acknowledgements The authors wish to thank the Assistance Publique±HoÃpitaux de Paris for their grant (PHRC no. 094034) and the MXM Laboratory for their ®nancial support. The authors also wish to thank D. Moyse for helping with the statistics, S. Labassi for editing the manuscript, and the patients who volunteered for the experiments. References Beliaeff M, Dubus P, Levean J-M, Repetto J-C, Vincent P. Sound signal processing and stimulation coding of the Digisonic DX 10 15-channel cochlear implant. In: Hochmair-Desoyer IJ, Hochmair ES, editors. Advances in cochlear Implants, Vienna: Mainz, 1994. pp. 198±203. GalleÂgo S, Frachet B, Micheyl C, Tray E, Collet L. Cochlear implant performance and electrically evoked auditory brain-stem response characteristics. Electroenceph clin Neurophysiol 1998;108:521±525. Groenen PAP, Makhdoum M, Van Den Brink JL, Stollman MHP, Snik AFM, van den Broek P. The relation between electric auditory brain stem and cognitive responses and speech perception in cochlear implant users. Acta Otolaryngol (Stockh) 1996;116:785±790. Groenen PAP, Snik AFM, van den Broek P. Electrically evoked auditory middle latency responses versus perception abilities in cochlear implant users. Audiology 1997;36:83±97. Herman B, Thornton A. Electrically-evoked auditory brain-stem responses in cochlear implanted subjects (abstr) Iowa City, IA. Second International Cochlear Implant Symposium 1992;:57. Kaukoranta E, Sams M, Hari R, HaÈmaÈlaÈinen M, NaÈaÈtaÈnen R. Reactions of human auditory cortex to a change intone duration. Hear Res 1989;41:15±22.
751
Kileny PR, Boerst A, Zwolan T. Cognitive evoked potentials to speech and tonal stimuli in children with implants. Otolaryngol Head Neck Surg 1997;117:161±169. Kraus N, Micco AG, Koch DB, McGee T, Carrell T, Sharma A, Wiet RJ, Weingarten CZ. The mismatch negativity cortical evoked potential elicited by speech in cochlear-implant users. Hear Res 1993a;65:118± 124. Kraus N, McGee T, Micco A, Sharma A, Carrel T, Nicol T. Mismatch negativity in school-age children to speech stimuli that are just perceptibly different. Electroenceph clin Neurophysiol 1993b;88:123± 130. Kraus N, McGee T, Carrell T, Sharma A. Neurophysiologic bases of speech discrimination. Ear Hear 1995a;16:19±37. Kraus N, McGee T, Carrel T, King C, Tremblay K, Nicol T. Central auditory system plasticity associated with speech discrimination training. J Cogn Neurosci 1995b;7:25±32. Kraus N, McGee TJ, Koch DB. Speech sound representation, perception, and plasticity: a neurophysiologic perspective. Audiol Neuro-Otol 1998;3:168±182. Makhdoum MJ, Groenen PAP, Snik AEM, van den Broek P. Intra- and interindividual correlation between auditory evoked potentials and speech perception in cochlear implant users. Scand Audiol 1997;27:13±20. NaÈaÈtaÈnen R, Michie PT. Early selective attention effects on the evoked potential. A critical review and reinterpretation. Biol Psychol 1979;8:81±136. NaÈaÈtaÈnen R, Picton T. The N1 wave of the human electric and magnetic response to sound: a review and an analysis of the component structure. Psychophysiology 1987;24:375±425. NaÈaÈtaÈnen R, Gaillard AWK, MaÈntysalo S. Early selective attention effect on evoked potential reinterpreted. Acta Psychol 1978;42:313±329. Oviatt D, Kileny P. Auditory event-related potentials elicited from cochlear implant recipients and hearing subjects. Am J Audiol 1991;1:48±55. Pelizzone M, Kasper A, Montandon P. Electrically evoked responses in cochlear implant patients. Audiology 1989;28:230±238. Picton TW, Hillyard SA, Krausz HT, Galambos R. Human auditory evoked potentials I: evaluation of components. Electroenceph clin Neurophysiol 1974;36:179±190. Ponton CW, Don M. The mismatch negativity in cochlear implant users. Ear Hear 1995;16:131±146. Sams M, HaÈmaÈlaÈinen M, Antervo A, Kaukoranta E, Reinikainen K, Hari R. Cerebral neuromagnetic responses evoked by short auditory stimuli. Electroenceph clin Neurophysiol 1985;61:254±266. Scherg M, Vajsar J, Picton TW. A source analysis of the late human auditory evoked potentials. Cogn Neurosci 1989;1:336±355. Tremblay K, Kraus N, Carrell TD, McGee T. Central auditory system plasticity: generalization to novel stimuli following training. J Acoust Soc Am 1997;102:3762±3773.