Dept. for Speech, Music and Hearing

Quarterly Progress andStatus Report

Analysis of different speechperformance in patients with

multi-channel cochlearimplants

Agelfors, E. and Beidestam, G. and Rollvén,B. and Vainio, M. and Lindström, B. and

Bredberg, G.

journal: TMH-QPSRvolume: 37number: 4year: 1996pages: 097-110

http://www.speech.kth.se/qpsr

http://www.speech.kth.sehttp://www.speech.kth.se/qpsr

TMH-QPSR 4/1996

97

Analysis of different speech performance inpatients with multi-channel cochlear implantsEva Agelfors2, Gunnar Beidestam1, Birgitta Rollvén1, Marja Vainio1, Bo Lindström1 and Göran Bredberg1

1. Department of Audiology, Söder Hospital, Stockholm, Sweden2. Department of Speech, Music and Hearing, KTH, Stockholm, Sweden

AbstractTotally 41 postlingually deafened adult patients who have received multi -channelcochlear implants at Stockholm Söder Hospital have participated in this study. Allpatients have followed the CI-programme at the clinic and have been tested 0, 6,12, and 24 months postoperatively. The test-battery consisted of vowels (cVc) andconsonants (vCv), Everyday Sentences(ES), and Connected Discourse Tracking(CDT) presented in two modaliti es: cochlear implant plus lip-reading and implantalone. Results for the 41 patients from the speech tests showed an improvement, onaverage, within the first 6 months of implant use. Some patients, however, get highscores from the beginning, while others gradually improved over a 2-year period.The patients have been divided into three groups POOR (11), AVERAGE (13) andGOOD (17) based on their performance auditory alone on Connected DiscourseTracking test. Group mean results on consonants, showed that for patients withAVERAGE and GOOD speech understanding, errors in recognition of voicingwere rare but nasals were poorly perceived. For the POOR-group, envelope cuesseem to dominate the consonant recognition. The fricative [

� ] was well i dentified

by all patients in the three groups. Results are discussed in relation to etiology ofdeafness, years of deafness and other factors.

IntroductionNowadays, cochlear implantation is a generallyaccepted alternative for patients with a severe toprofound hearing loss who derive li ttle or nobenefit from a conventional hearing aid.Published results, however, show a wide varia-tion in performance with the same type of coch-lear implant and the possibili ties to predict theseresults from preoperative data and measurementsare limited (Tyler, 1991; Dorman, 1993; Dowell ,1995). In Sweden, cochlear implants have beenused for 12 years and 114 patients have beenimplanted at Söder Hospital, Stockholm. From thebeginning, a single-channel extracochlear implantwas used, Vienna/3M, but today multi-channelcochlear implants are used.

Some researchers suggest that the per-formance with a cochlear implant stabili sesduring the first year, whereas others note thatperformance continues to improve after 1 year.Waltzman & Hochberg (1990) found that SpeechPattern Contrast (SPAC) scores of six subjectswearing the Nucleus were similar at 3, 12 and 24months after cochlear implant connection. Tye-Murray et al. (1992) reported about resultsobtained over time by 13 Nucleus and 14

Ineraid subjects. On average, the abilit y of thesubjects to recognise words and phonemes in anaudition-only condition improved during thefirst 9 months, but the phoneme scorescontinued to improve during the next 9 months.About 50% of the subjects, who demonstratedpoor word recognition at 1 month, showedsignificantly improved percent word correctscores by 18 months. They also reported that theNucleus and Ineraid subjects did not differ intheir pattern of change over time. Tyler &Summerfield (1996) reported that 80% of theadult patients implanted in Iowa showsignificant performance improvement over time.On average, performance reached asymptoteafter 30 to 40 months of implant use, althoughindividual differences in the rate and amount ofimprovement are large. Many patients had greatbenefit from their aids and have some open-setspeech understanding, others had li ttle benefitand performed less well on speech perceptiontests.

The primary aim of this study was to evaluatethe test results obtained by all postlinguallydeafened adults, who had received a multi-channel cochlear implant at Söder Hospital,Stockholm, and used their implant for at least 2

Agelfors et al.: Analysis of different speech performance .....

98

years. All patients followed the CI-programme atthe clinic and were tested at 0, 6, 12 and 24months postoperatively. The second aim was toevaluate what effects residual hearing, durationof deafness, age at onset of deafness and etiologycould have on the obtained results with theimplant.

MethodPatientsTotally 41 postlingually deafened adults, allSwedish, participated in this study and all havereceived multi-channel cochlear implants at SöderHospital, Stockholm, operated by the samesurgeon (G Bredberg). They had no benefit froma conventional hearing aid at the time ofimplantation. All have followed the CI-programme at the clinic and have been tested at0, 6, 12 and 24 months postoperatively. Somedata for the patients is shown in Table 1.

Table 1. Some patient data

Patients: 41 (19 females, 22 males)Age: mean 54.5 years (range 24-78)Years of deafness: mean 14 years (range 1-41)

Etiology N=Unknown, progressive loss 11Otosclerosis 7Meningitis 6Unknown, sudden deafness 5Ototoxicity 5Skull trauma 3Meniérè 1Rem. Artrit 1Sarkoidos 1Unknown 1

CI-systemsTwo different systems have been used: Nucleusand Ineraid. Thirty-nine patients have beenimplanted with the Nucleus 22-channel cochlearimplant (Clark et al., 1981) and two patientshave been implanted with the Ineraid (Eddington,1980). The most important difference betweenthe two processors is that the Nucleus design is a22-channel feature extraction device, whereas theIneraid processor is a simultaneous, analog four-channel implant. The Ineraid processor does notperform feature extraction, but presents thebroadband speech waveform in four fil terchannels (Dorman et al., 1990).

Two different processing strategies in theNucleus device have been used, F0/F1/F2 andMultipeak (MPEAK). Both of these strategiesextract frequency and ampli tude information

concerning the first and second formantfrequencies (F1 and F2). In addition to F1 andF2, the MPEAK strategy extracts and codesinformation in three high-frequency bandsbetween 2-6 kHz. Both strategies use non-simultaneous pulsatile stimulation and codefundamental frequency as the rate of stimulation.Table 2 shows the number of patients that usedthe different coding strategies. Stimuli mode BP= bipolar; the active and reference electrodes areadjacent to each other. The spread of current isconfirmed to a relatively small area. BP+1 = theactive and reference electrodes are separated byone nonactive electrode. The further the activeand reference electrodes are spaced, the greater isthe spread of current. CG = common ground; oneelectrode is active all the others are connectedtogether to constitute a single reference electrode.The spread of current is more diffuse.

Table 2. Characteristics of the CI-systems andnumber of patients.

1. Nucleus Mini system 22Strategy: F0/F1/F2 (N = 9)

MPEAK (N = 30)Stim.mode: BP+1 (N = 38)

CG (N = 1)Number of active electrodes: mean 19 (range14-20)

2. IneraidAnalog 4 band-pass filters (N = 2)

Test batteryThe test battery that was used consisted ofvowels (cVc) and consonants (vCv), EverydaySentences (ES), and Connected DiscourseTracking (CDT) presented in two modali ties:cochlear implant plus lipreading and cochlearimplant alone. All subjects have been tested at 0,6, 12 and 24 months postoperatively. The firsttest session “0“ means results obtained after oneweek of the speech processor´s connection to theimplant. The patients have also been testedauditory alone at the clinic in a soundfieldsituation on closed-set tests; 3-digits span testand 12-choice spondees, and on open-set speech;PB-words and spondees. The results from thesetests are not discussed here.

1) Connected Discourse Tracking (CDT)The tracking technique developed by De Fili ppoand Scott (1978) has been used by many groupsto train and to evaluate the reception ofconnected speech via lipreading alone and incombination with various types of sensory aids.The speaker reads from a book, phrase by

TMH-QPSR 4/1996

99

phrase, and the speechreader (the subject) isrequired to repeat the text verbatim. If the phraseis not correctly repeated, the speaker employs ahierarchy of strategies to assist the subject inrepeating every word correctly. The score iscalculated according to the number of words inthe text conveyed in a given time.

The speech material chosen had a relativelyconsistent level of reading diff iculty from sessionto session. During each test session tracking wasperformed for ten minutes under each condition:lipreading plus aid and auditory alone with thedevice. All patients were tested in conditionauditory alone but testing was interrupted assoon as it was evident that this situation was toodiff icult and the patient was given a score ofzero. The CDT-test was presented live by a maleor a female habili tationist. The tracking rateachieved by a normal-hearing subject, who usesthe same method with the same speaker and thesame text material, is 85-95 words per minute.

2) Everyday Sentences (ES)Two different test lists with 20 sentences each atvarying length (2 - 14 words) were used. Afemale speaker was video-taped and thesentences were presented on a TV-screen in theaudiovisual test situation. A tape-recorder wasused in the test condition auditory alone. Thecorrect sentences were not given to the patients.The results are scored based on the number ofwords correctly identified.

3) Vowels (cVc)Nine Swedish vowels /i:, y:, � :, e:, � :, � :, u:, o:,� :/ were presented in a /bVb/-context. The voweltest was video taped, and each vowel waspresented 3 times in a random order by a femalespeaker on a TV-screen in the audiovisualsituation. In the test situation auditory alone,only one of each vowel was presented. Theconfusion matrices were analysed.

4) Consonants (vCv)Sixteen Swedish consonants were presented in an/aCa/-context. The following consonantphonemes were used /p, b, m, t, d, n, k, g, � , f, v,s, � , r, l, j/. The consonants sampled a variety ofdistinctions in voicing, place of articulation, andmanner of articulation. The consonant test wasvideo taped and each consonant was presentedtwice in a random order by a female speaker on aTV-screen in the audiovisual test situation. In thetest situation auditory alone, only one of eachconsonant was presented. The confusion matriceswere analysed.

All speech tests were presented at thepatient´s most comfortable listening level

adjusted by themselves. All tests were conductedin a special test room at the clinic.

Results and discussionIndividual results obtained 24 months post-operatively from the CDT-test presented in thetwo situations, audiovisual and auditory alone,are shown in Figure 1. There was a largevariation in abili ty to perceive speech in the testsituation auditory alone among the cochlearimplantees. Some patients had no abili ty at all ,others achieved a score of more than 60 wordsper minute auditory alone. All patients obtainedsome improvement with their aids in theaudiovisual situation. The correlation betweenthe results from two test situations was r = 0.81.

Some patients derived li ttle benefit from theircochlear implants and if, in addition, they arepoor lipreaders, they have severe problems indaily communication situations. By experiencewe know that a tracking speed around 40 wordsper minute is a minimum speed required for anacceptable communication. Results over 60words per minute indicate a good communicationabili ty and these patients are able tocommunicate nearly as well as normal-hearingpersons in this particular test situation.

The analysis showed that the mainimprovement on the CDT-test in the two testsituations were obtained during the first sixmonths of CI-use (Fig. 2). There were no signi-ficant differences in the mean scores obtained bythe CI-users over time after the first six monthsof CI-use. The average results at 6, 12 and 24months varied from 22, 24 to 25 words perminute auditory alone. In the combined situation,the average results varied from 56, 57 to 56words per minute.To obtain more information about the differentlevels of performance among the patients theywere divided into three subgroups which arelabelled POOR (11 patients), AVERAGE (13patients) and GOOD (17 patients) based on theirmean score with the implant on CDT-testauditory alone at 6, 12 and 24 monthspostoperatively. POOR = 30 words/minute. Individual results (mean scoreat 6, 12 and 24 months postoperatively),obtained by patients on the CDT- test auditoryalone, are shown in Figure 3. There was a largevariation in the result across the patients onCDT-test, from 0 to 64 words per minute.

Agelfors et al.: Analysis of different speech performance .....

100

Fig. 1. Individual results obtained by the CI-users on Connected Discourse Tracking in test situationsaudiovisual and auditory alone with the implant at 24 months postoperatively.

CDT

0

20

40

60

80

100

0 6 12 24

Months post implant

Auditory

Audiovisual

Words/min.

Fig. 2. Mean performance on CDT over time after implantation presented in the conditions auditoryalone (CI) and audiovisual.

CDT / Auditory only (CI) (mean: 6,12, 24 months)

020406080

100

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41Patients

Words/min.

"AVERAGE" "GOOD"(N = 11) (N = 13) (N = 17 )

"POOR"

Fig. 3. Individual mean score obtained on CDT auditory alone with the implant at 6, 12 and 24 mopostoperatively. The CI-users were divided into three subgroups POOR, AVERAGE and GOOD basedon their performance on the CDT-test.

Connected Discourse Tracking (CDT)(Postop. 24 mo)

0

20

40

60

80

100

0 20 40 60 80 100

Aud

iovi

sual

Words/min.

Auditory only (CI)

Goodcommunication

Acceptable communication

Words/min.

r = 0.81

TMH-QPSR 4/1996

101

Fig. 4 a-b. Group-mean performance on number of correct words perceived on CDT-test as afunction of time obtained by the three groups of CI-users in the two test conditions: auditory alonewith the implant and audiovisual.

Connected Discourse Tracking (CDT)The group-mean results obtained on CDT areshown in Figures 4a and b. The performance onCDT presented audiovisually showed a statisticalimprovement within the groups (p

Agelfors et al.: Analysis of different speech performance .....

102

Fig. 6. Relationship between individual mean results at 6, 12 and 24 months on Connected DiscourseTracking and Everyday Sentences tested audiovisually.

Everyday Sentences (ES)The results obtained on ES test showed similartendencies as on the CDT, which can be seen inFigures 5a and 5b. The correlation between theresults from two speech tests is shown in Figure6. The correlation coeff icient is r = 0.83.Individual results on ES showed a ceili ng-effectfor some of the better CI-users already after 6 to12 months of CI-use.

The average performance on the ES for thesubgroups over time in two test conditions areshown in Figures 5a-b. In the combined situationall three groups of CI-users showed a significantimprovement, on average, between the first andsecond test sessions, but after six months ofimplant use no significant improvement hadoccurred over time. As shown in Figures 4a-band 5a-b, performance stabili sed after about 6months for many of the patients, but for thebetter performers it continued to increasegradually over several years of implant use.Some patients showed fluctuations inperformance, but some of the poorer showedreductions in their performance over time. Themean score obtained by the POOR groupdeteriorated between the first and second yearwith the implant and the reason for this isunclear. There were significant differences (p<0.01) on average between the groups at the firstand second test sessions, but after 6 months of

CI-use the differences were negligible betweenthe two groups of better performers. Theperformance in the combined condition showedafter 6 months of CI-use an improvement around40% for all three groups of CI-users compared tocochlear implant alone.

In the test situation auditory alone, significantdifferences were obtained between the groupsafter 6 months of implant use, the GOODperformers scored around 60% correct, theAVERAGE performers around 30% and thePOOR performers perceived only a few wordscorrectly. The improvement within the first 6months of CI-use was for the GOOD and theAVERAGE groups around 20% compared to noimprovement at all for the POOR group.

Analysis of vowel and consonant testFigure 7 shows the mean results obtained by thethree subgroups on different speech tests at 6, 12and 24 months postoperatively. No significantimprovements were noted over time from 6months until 24 months for any of the subgroupson any of the tests. However, there are significantdifferences in performance between thesubgroups on the different speech tests. Toobtain a larger number of data for the confusionmatrix analysis the vowel and consonant testresults at 6, 12 and 24 months were collapsed.

CDT /ES, Audiovisual (mean 6,12, 24)

0

20

40

60

80

100

0 20 40 60 80 100

CDT, words/min.

Eve

ryda

y S

ente

nces

, %

cor

rect

r = 0.83

TMH-QPSR 4/1996

103

Fig. 7. Group-mean data on different speech tests obtained by the three groups of CI-users at 6, 12and 24 months postoperatively.

Figure 8 shows the confusion matrices for vowelspresented auditory alone for the three groups ofCI-performers. The front vowels /i:, y:, � :, e:, � :, :,/ and the back vowels / u:, o:, :/ are related totongue position from front to back of the oralcavity and associated with distance between thefrequencies of the second formant (F2) and thefirst formant (F1) of the vowels. There weresignificant differences (p

Agelfors et al.: Analysis of different speech performance .....

104

(GOOD) (AVERAGE)R e s p o n s e R e s p o n s e

i y � � e � � u o � � i y � � e � � � � u o � �S i 43 26 21 S i 39 18 13 11t y 32 57 t y 55 34i � � 70 i � � 18 66

m e 66 19 m e 16 39 18u � � 13 57 28 u � � 21 32 37l � � 11 38 47 l � � 24 47u u 83 17 u u 61 32s o 26 72 s o 16 18 63� �

11 85� �

13 71

All correct: 64.0% All correct: 50.2%Front vowels: 98.6 % Front vowels: 93.9%Back vowels: 99.3 % Back vowels: 87.7%

(POOR)R e s p o n s e

i y � � e � � � � u o � �S i 50 21t y 46 13 13 Figure 8. Confusion matrices from i � � 38 21 vowel recognition test /bV:b/ in condition

m e 17 54 auditory alone obtained by three groupsu � � 13 25 25 25 of patients: GOOD, AVERAGE and POOR.l � � 17 29 13 29 Confusions less than 10 % are not shownu u 17 42 13 in the matrices. Shading indicates front s o 13 13 29 29 and back vowels.� �

17 17 50

All correct: 34.8%Front vowels: 79.9 %Back vowels: 68.1%

Table 3. Mean percentage scores (standard deviations are given in parenthesis) obtained by the threegroups of CI-users on the consonantal features in two conditions: auditory alone (A) and audiovisual(AV). Signifi cant difference between the test conditions within groups on consonantal feature* (p

TMH-QPSR 4/1996

105

The consonantal features fricative and stopwere relatively well perceived by the GOODperformers. The feature nasali ty was poorlyperceived on average by all three groups ofimplantees. The AVERAGE performers madealso more errors for stops and nasals, while thePOOR performers were substantially poorer toidentify nasals and fricatives with high frequencycomponents and often heard as /t/ and /k/. In thecombined condition, all groups obtained animprovement on average around 30% forconsonants overall . The group-mean scores ofvoicing and nasali ty were not improvedstatistically as could be expected in the combinedcondition compared to auditory alone for any of

the groups, and the audiovisual characteristics ofthe stops were not distinguishable for POORgroup of CI-users.

The confusion matrices for the consonants,Figures 9 and 10, were analysed to obtaindetailed information about the most frequenterror responses from the different groups of per-formers in the two test situations. Figure 9 showsthe confusions of the consonants presentedauditory alone. The GOOD performers weremuch better to perceive voice/voicelessdistinction and manner information andespecially the stops /b, g/ and /p, k/ were betterrecognised by the better performers compared tothe poorer performers.

AuditoryaCa

(GOOD) R e s p o n s e (AVERAGE) R e s p o n s eb d g m n � � l r j v f � � s p t k b d g m n � � l r j v f � � s p t k

b 66 14 b 27 11 19 8d 38 26 20 8 d 14 8 30 11 11 11g 8 20 42 10 8 g 16 11 8 14 24m 20 28 10 22 m 11 22 24 11 14

S n 12 50 18 8 S n 24 14 24 8 16t � � 32 36 14 t � � 22 30 16 8i l 14 16 8 54 i l 21 34 13 13

m r 80 m r 8 79 8u j 10 10 10 44 10 u j 8 13 13 37 8l v 32 12 8 30 l v 8 11 8 14 11 16 8 8u f 44 14 14 12 u f 8 11 30 16 14s � � 86 8 s � � 74 16

s 30 46 20 s 16 51 19p 8 8 14 52 12 p 11 8 11 8 11 21 21t 12 28 32 14 t 11 21 35k 8 10 10 62 k 13 11 47

(POOR) R e s p o n s eb d g m n � � l r j v f � � s p t k

b 20 12 8 12 8 12d 20 12 8 8 8 12 8g 12 12 12 8 20 8m 8 28 8 20 8 Fig. 9. Consonant /vCv/ confusion matrices

S n 24 12 8 12 8 8 12 for the three groups of CI-users in auditory testt � � 8 8 12 24 8 12 8 situation. Confusions less than 8% are not showni l 8 8 8 14 36 8 in the matrices. Shading indicates manner and

m r 8 44 12 voice/voiceless categories.u j 8 8 12 12 12 44l v 8 8 20 8 8 8 8 0 8u f 20 14 24 12 20s � � 12 68

s 8 16 28 28p 8 8 8 16 12 28 12t 8 16 8 12 28 20k 8 8 24 32

Agelfors et al.: Analysis of different speech performance .....

106

AudiovisualaCa

(GOOD) R e s p o n s e (AVERAGE) R e s p o n s em p b f v n t d r l s � � j k g m p b f v n t d r l s ! ! j k g

m 78 20 m 80 20p 84 16 p 68 30b 8 91 b 91f 99 f 96

S v 92 S v 17 74 8t n 62 15 9 t n 43 9 18 16i t 60 9 21 i t 24 8 24 9 25

m d 57 18 19 m d 22 36 12 14u r 65 29 u r 51 11 24l l 93 l l 12 86u s 65 31 u s 46 46s " " 98 s " " 8 91

j 8 86 j 88k 86 11 k 87 11g 94 g 9 24 58# #

19 16 56# #

11 11 28 11 38

(POOR) R e s p o n s em p b f v n t d r l s " " j k g # #

m 59 34p 72 24b 16 78 Fig. 10. Consonant /vCv/ confusion matrices f 79 12 for the three groups of CI-users in audiovisual

S v 36 50 9 test situation. Confusions less than 8% are not t n 26 10 31 12 shown in the matrices. Shading indicatesi t 28 24 33 visemes with the same place of articulation.

m d 24 9 22 18u r 36 19 19 10l l 95u s 43 45s " " 93

j 17 60k 81 12g 14 31 34# #

34 12 16 9 16

Nasals were heard predominantly as the /l/-soundor as voiced consonants sharing the samecharacteristics. The /r/-sound was poor recog-nised by the poorer performers and often heardas other voiced consonants as /l and j/. Thefricatives /f, $ , s/ were often confused with eachother and the intense fricative /s/ waspredominantly heard as / % /. The explanation forthis seems be that they perceived all high-energyfricatives as / % /-sound. The voiceless fricative /

% / is mostly generally characterised by aconcentration of energy above 1.5 kHz andextending to above 4 kHz and /s/ is characterisedby energy above 4.0 kHz. In the combinedsituation (Fig. 10), the place feature was wellperceived by all three groups of CI-users and theconfusions were often made within the

groups according to whether they are produced inthe front, middle or back of the vocal tract.

Relationship between speech performanceand some back ground variables of theCI-usersThe factors determining good performance with acochlear implant have not yet been clearlyidentified. To get some knowledge about thedifferences among the CI-users, the resultsobtained by the three groups of performers werecompared with respect to different backgrounddata; age at implantation, years of deafness,number of active electrodes, etc. Some basic datais presented in the following.

TMH-QPSR 4/1996

107

1. Duration of deafness and age atimplantationYears of deafness and maybe age at implantationare possible predictors for the outcome. In Figure11, mean-group data and range min/max for ageat implantation and duration of deafness arepresented from the different groups of CI-users.The highest mean age was represented in thePOOR group with a range of 54-78 years and amean age of 65 years. Younger persons wererepresented in the GOOD and AVERAGEgroups and the mean ages were around 50 years.The widest range of age, however, was found inthe group of GOOD performers with a rangefrom 25 years of age up to 78 years at the time ofimplantation.

Shorter durations of deafness was found inthe GOOD-group and CI-users with longerdurations of deafness was represented in thePOOR-group. The widest range of durations wasfound in the POOR-group from about 10 yearsup to 40 years of deafness.

In Figures 12 a-b and 13 a-b are shown therelationship between CDT presented in two testconditions and duration of deafness and age atimplantation. The correlation coeff icient betweenage at implantation and CDT auditory alone isweak (r = -0.24). There is a tendency that manyof the older CI-users performed somewhat poorlywith their implants which could be explained bya longer deprivation of auditory input beforeimplantation. In the audiovisual situation, wheretwo different modali ties are combined, a higher

correlation coeff icient (r = -0.49) compared toauditory alone was obtained.

Speech perception performance with theimplant and duration of deafness are strongerassociated to each other compared to age atimplantation, r = 0.60. Blamey (1995) hasgathered data for a total group of 425 post-linguistic adult implant users and reported aboutage at implantation and duration of deafness.When grouped by both age at implantation andduration of deafness, the data showed aconsistent effect of age at implantation only atshorter durations of deafness. For durationslonger than 10 years, the effect of age atimplantation was relatively small and did notshow an orderly decrease from younger to olderages. Tyler & Summerfield (1996) reported thatpartial correlations demonstrate that bothmeasures have independent effects on outcomes.Many others have reported significantassociations between performance with animplant and both age at implantation andduration of deafness. However, some otherstudies have not found this pattern (Summerfield& Marshall , 1995). For example, among 119patients in United Kingdom who were testedpreoperatively and 9 months postoperatively,there was no significant association between ageand performance (r = 0.11), whereas a significantcorrelation did exist between performance andduration of deafness (r = -0.47, p

Agelfors et al.: Analysis of different speech performance .....

108

Fig. 12 a-b. Relationship between age at implantation and CDT (mean value at 6, 12 and 24 months) intwo test situations: auditory alone with the implant and audiovisual.

Fig. 13 a-b. Relationship between duration of deafness and CDT (mean value at 6, 12 and 24 months)in two test situations: auditory alone with the implant and audiovisual.

2. EtiologyThe cause of deafness varies among the im-plantees (Table 4), and there would appear to beno particular etiology bias in the three groups,except for otosclerosis which seems to be a goodpredictor for good performance.

Table 4. Etiology represented in the differentgroups of CI-users

GOOD AVERAGE POOREtiology: (N=17) (N=13) (N=11)

Unknown,progressive loss

3 3 5

Otosclerosis 5 2 -Meningitis 3 2 1Unknown,sudden deafness

2 1 2

Ototoxicity 3 1 1Skull trauma - 1 2Meniérè 1 - -Rem. artrit - 1 -Sarkoidos - 1 -Unknown - 1 -

3. Relationship between speech performanceand residual hearingTable 5 shows group-mean data of audiometricthresholds and range (min-max) at three differentfrequencies 250, 500 and 1000 Hz measuredpreoperatively on the implanted ear for the threegroups. There are great variations within eachgroup, but a majority of the GOOD performershad some residual hearing, especially at the lowerfrequencies, while most of the POOR performershad on average poorer thresholds. Residualhearing means that there is still some nerveactivity in the cochlea which could have a positiveeffect on the results with the implants.

4. Processor strategyRelationship between speech performance and CI-systems is presented in Table 6. Only twosubjects used Ineraid. They were represented inthe GOOD and POOR groups. The twoprocessing strategies available for the Nucleus

CDT/ Auditory (mean 6,12, 24)

0

20

40

60

80

100

20 30 40 50 60 70 80

Age at implantation (years)

r = - 0.24

Words/min. CDT/ Audiovisual (mean 6,12, 24 )

0

20

40

60

80

100

20 30 40 50 60 70 80

Age at implantation (years)

r = - 0.49

Words/min.

CDT/Auditory (mean 6,12,24)

0

20

40

60

80

100

0 10 20 30 40 50

Duration of deafness (years)

r = - 0.60

Words/min.CDT/ Audiovisual (mean 6,12,24)

0

20

40

60

80

100

0 10 20 30 40 50Duration of deafness (years)

r = - 0.64

Words/min.

TMH-QPSR 4/1996

109

Table 5. Group-mean data and range of audio-metric thresholds at frequencies 250, 500 and1000 Hz at the implanted ear preoperatively.

GOOD AVERAGE POOR

250 Hz, dBHLrange/ min-max

500 Hz, dBHLrange/ min-max

1000Hz, dBHLrange/ min-max

100(60-120)

102(75-120)

109(90-130)

103(75-120)

111(95-130)

121(105-130)

113(80-120)

118(95-30)

122(105-130)

users were MPEAK and F0/F1/F2. All subjectswho from the beginning were equipped with theolder speech coding strategy have had thepossibili ty to upgrade to MPEAK strategy.However, some of the Nucleus users havepreferred their practised F0/F1/F2 strategy.MPEAK strategy was represented in all groupsbut F0/F1/F2 was represented only in theAVERAGE and POOR groups. There are nodifferences between the groups regarding thenumber of active electrodes.

Table 6. CI-systems, programming data andnumber af active electrodes for the differentgroups of CI-users.

GOOD AVERAGE POOR(N =17) (N =13) (N =11)

CI-system Ineraid ( 1) - Ineraid ( 1)CI-system Nucleus (16) Nucleus (13) Nucleus (10)Strategy MPEAK (16) MPEAK ( 9) MPEAK ( 5)

F0/F1/F2 ( 4) F0/F1/F2 ( 5)

Stimuli mode BP+1 (15) BP+1 (13) BP+1 (10)CG ( 1)

No. of active electrodesmean 19 18 19range 15-20 15-20 14-20

Notes. Number in parentheses indicate number ofCI-users.

SummaryIn summary, 41 postlingually deafened adults whoreceived a multichannel cochlear implant at SöderHospital, Stockholm, equipped with second-generation devices have been evaluated in thisstudy, all with daily experience with CI up to twoyears. The abili ty to improve speech recognitionin different test situations was confirmed bydifferent speech tests. There was, however, alarge variation across the CI-users to understand

some open-set speech, without contextualinformation. Eleven patients (27%) had no abili tyat all to understand connected speech withoutsupport of lipreading on CDT-test. Oncomparison, the group of GOOD performers(41%) perceived more than 30 words/minute, andsome of them are very successful incommunication auditory alone in a quietenvironment with a famili ar speaker. Theseresults show that progress in speech perception isto be found within the first six months, whichindicates that this period is of the greatestimportance in terms of training and mapping,especially for the POOR performers. Further-more, 50% of the POOR performers with theNucleus device preferred the FO/F1/F2 to theMPEAK strategy which is probably due to thefact that they were fitted with the former strategyfirst and that they cannot utili se the moresophisticated MPEAK strategy. The obtainedresults of consonants between the better and thepoorer subjects could possibly be explained bydifferent coding strategies Multipeak versusF0/F1/F2. On the other hand, Skinner et al.(1991) report no difference in percent correct foreither vowels or consonants for patients testedwith both the F0/F1/F2 and Multipeak processor.

Many of the GOOD performers obtained openspeech from the very first day of CI-use andcontinued to improve over these two years oftesting. However, statistics showed that thehighest progression was during the first sixmonths of implant use which was the mostimportant period to predict the outcome for allgroups of implantees. Many of the GOODperformers became profoundly deaf a shorter timeprior to implantation compared to the POORperformers and some of them possess someresidual hearing in the low frequency range.Outcomes from CI vary widely among adultpatients even in a homogeneous group of CI-userwho use similar devices.

Final remarksThe results in this study show that progress is tobe found within the first six months, whichindicates that this period is of the greatestimportance in terms of training and mapping,especially for the POOR performers. The GOODperformers continued to improve their abili ty toperceive open speech auditory alone with thecochlear implants over time. The future researchshould be focused on identifying the componentsof individual variability among the CI-users.

Agelfors et al.: Analysis of different speech performance .....

110

AcknowledgementThis study is a result of co-operation between thedepartment of Audiology at Söder Hospital inStockholm and Department of Speech, Music andHearing, KTH. This report is an expanded versionof the paper presented at III rd InternationalCongress on Cochlear Implant, Paris, April 27-29, 1995.

ReferencesBlamey P (1995). Factors affecting auditory

performance of postlinguisticall y deaf adults usingcochlear implants: Etiology, age & duration ofdeafness. Presented at US National Institutes ofHealth Consensus Development Conference onCochlear Implants in Adults and Children, May 15-17, 1995.

Clark G, Tong Y, Dowell R, Martin L, Seligman P,Busby P & Patrick J (1981). A multiple channelcochlear implant: An evaluation using nonsensesyllables. Annals of Otology, Rhinology andLaryngology 90: 227-230.

Dorman MF, Soli S, Dankowski K, McCandless G &Parkin J (1990). Acoustic cues for consonantidentification by patients who use the Ineraidcochlear implant. J Acoust Soc Am 88/5: 2074-2079.

Dorman MF (1993). Speech perception by adults. In:Tyler RS, ed. Cochlear Implants, 145-189.

Dowell RC (1995). Speech perception for adultsusing cochlear implants. In: Plant G & Spens K-E,

eds. Profound Deafness and SpeechCommunication, 262-284.

Eddington D (1980). Speech discrimination in deafsubjects with cochlear implants. J Acoust Soc Am68: 885-891.

De Fili ppo CL & Scott BL (1978). A method fortraining and evaluating the reception of ongoingspeech. J Acoust Soc Am 63: 1186-1192.

Skinner M, Holden L, Holden T, Dowell R, SeligmanP, Brimacombe J & Beiter A (1991). Performanceof postlinguisticall y deaf adults with the wearablespeech processor (WSP III) and mini speechprocessor (MSP) of the Nucleus multi -channelcochlear implant. Ear and Hearing 12: 3-22.

Summerfield AQ & Marshall DH (1995). Cochlearimplantation in the UK 1990-1994. Report by theMCR Institute of Hearing Research on theEvaluation of the National Cochlear ImplantProgramme.

Tye-Murray N, Tyler RS, Woodworth GG & GantzBJ (1992). Performance over time with Nucleus orIneraid cochlear implant. Ear and Hearing 13/3:200-209.

Tyler RS (1991). What can we learn about hearingaids from cochlear implants? Ear and Hearing12/6, Suppl: 177-186.

Tyler RS & Summerfield AQ (1996). Cochlearimplantation: Relationships with research onauditory deprivation and acclimatization. Ear andHearing 17/3: 38-50.

Waltzman S & Hochberg I (1990). Perception ofspeech pattern contrasts using a multi channelcochlear implant. Ear and Hearing 11: 50-55.