25
Application of electrophysiological Application of electrophysiological measurements in processor fitting; measurements in processor fitting; parametric fitting parametric fitting Guido F. Smoorenburg Guido F. Smoorenburg Christina Christina Willeboer Willeboer Johannes E. van Johannes E. van Dijk Dijk

Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

  • Upload
    others

  • View
    6

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Application of electrophysiological Application of electrophysiological measurements in processor fitting;measurements in processor fitting;

parametric fittingparametric fitting

•• Guido F. SmoorenburgGuido F. Smoorenburg

•• Christina Christina WilleboerWilleboer

•• Johannes E. van Johannes E. van DijkDijk

Page 2: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

QuestionQuestion•• Starting rehabilitation of a cochlear implantStarting rehabilitation of a cochlear implant

patient by assessing the Tpatient by assessing the T-- and Cand C--levels for 22levels for 22electrodes is quite cumbersome, particularlyelectrodes is quite cumbersome, particularlyin children or those individuals who have littlein children or those individuals who have littlehearing experience at all.hearing experience at all.

•• Can we simplify the procedure,Can we simplify the procedure,make it more patientmake it more patient--friendly,friendly,e.g.e.g. by using electrophysiological data,by using electrophysiological data,and / or by using speech stimuli ?and / or by using speech stimuli ?

Page 3: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

ApproachApproach•• The literature (The literature (AbbasAbbas, Brown) shows that , Brown) shows that

the correlation between Tthe correlation between T--levels and NRTlevels and NRT--thresholds is weak.thresholds is weak.

•• However, the profile of the NRTHowever, the profile of the NRT--thresholds thresholds across the electrode array may be indicative across the electrode array may be indicative of the profile of the Tof the profile of the T-- and Cand C--levels.levels.

•• Take the NRT profile and adjust the overall Take the NRT profile and adjust the overall level of the profile to threshold and comfort level of the profile to threshold and comfort level using life speech.level using life speech.

Page 4: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

MethodsMethods•• Adult subjectsAdult subjects•• SPEAK strategySPEAK strategy•• PostPost--operative NRT measurementsoperative NRT measurements•• NRT measurements at electrodes 3NRT measurements at electrodes 3--2222•• OverOver--all adjustment of NRT profile to Tall adjustment of NRT profile to T--

and Cand C--levels using life voice > NRTlevels using life voice > NRT--MAPMAP•• Speech reception tested with CVC wordsSpeech reception tested with CVC words•• Speech test immediately after assessing Speech test immediately after assessing

NRTNRT--MAP and after 2 weeks of optional MAP and after 2 weeks of optional use of the NRTuse of the NRT--MAP MAP

Page 5: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Results 1Results 1•• Postoperatively, we found NRT thresholds at Postoperatively, we found NRT thresholds at

all 20 electrodes in 13 < 27 subjects (48%). all 20 electrodes in 13 < 27 subjects (48%). •• The result of the Cochlear field study is The result of the Cochlear field study is

similar: 65 < 147 (44%).similar: 65 < 147 (44%).•• TT--NRT, the TNRT, the T--levels assessed by adjusting the levels assessed by adjusting the

overall level of the NRToverall level of the NRT--profile using life profile using life voice, appeared to be much lower than the voice, appeared to be much lower than the conventionally determined Tconventionally determined T--levels.levels.

Page 6: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Comparison of NRTComparison of NRT--thresholds thresholds to Tto T-- and Cand C-- levelslevels

profiles

100

110

120

130

140

150

160

170

180

190

200

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

# electrode

cure

nt u

nits

T-levelC-levelNRT-levelT-NRTC-NRTT-NRT-T

Page 7: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

NRT profile touching NRTNRT profile touching NRT--levelslevels

profiles

100

110

120

130

140

150

160

170

180

190

200

210

220

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

# electrode

cure

nt u

nits

T-levelC-levelNRT-levelT-NRT

Page 8: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Comparison of NRTComparison of NRT--thresholds to Tthresholds to T--and Cand C-- levelslevels

profiles

100

110

120

130

140

150

160

170

180

190

200

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

# electrode

cure

nt u

nits

T-levelC-levelNRT-levelT-NRTC-NRTT-NRT-T

Page 9: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Results 2Results 2•• The TThe T--NRT levels are much lower than the conventional NRT levels are much lower than the conventional

TT--levels.levels.•• The CThe C--NRT levels are somewhat higher than the NRT levels are somewhat higher than the

conventional Cconventional C--levels.levels.•• Thus, the dynamic range of the NRT MAP is about twice Thus, the dynamic range of the NRT MAP is about twice

the range of the conventional MAP.the range of the conventional MAP.•• For 1/3 the lower TFor 1/3 the lower T--NRT levels may be explained by the NRT levels may be explained by the

assumption of response of the 1assumption of response of the 1stst electrode.electrode.•• For 2/3 the lower TFor 2/3 the lower T--NRT levels may be explained by NRT levels may be explained by

spatial/frequencyspatial/frequency--band integration of impulse rate.band integration of impulse rate.

Page 10: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

CVC scores with CVC scores with behaviouralbehavioural and and NRTNRT--based MAPS, 1based MAPS, 1stst sessionsession

0

10

20

30

40

50

60

70

80

90

100

PG JH MH NH PAK PK MO AR NS MS FT HT JW mean

phon

eme

scor

e (%

)

behavioural

ECAP-based

Page 11: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Hours of use of NRTHours of use of NRT--based MAPbased MAP

0

20

40

60

80

100

120

140

160

180

200

PG JH MH NH PAK PK MO AR NS MS FT HT JW mean

hours

Page 12: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

CVC scores with CVC scores with behaviouralbehavioural and and ECAP based MAPS, 2ECAP based MAPS, 2ndnd sessionsession

0

10

20

30

40

50

60

70

80

90

100

PG JH MH NH PAK PK MO AR NS MS FT HT JW mean

phon

eme

scor

e (%

)

behavioural

ECAP-based

*

* *

* ** * *

*

*

*

* few hours of use * worse speech

Page 13: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Results 3Results 3•• On average, phoneme perception is little (7%) On average, phoneme perception is little (7%)

affected by a large decrease in overall Taffected by a large decrease in overall T--levels levels and some upward shift of the overall Cand some upward shift of the overall C--levels. levels.

•• The MAP may change considerably without The MAP may change considerably without affecting speech reception.affecting speech reception.

•• The results show marked The results show marked intersubjectintersubject variability; variability; some subjects perform just as well with the NRTsome subjects perform just as well with the NRT--based map, some perform up to 25% worse.based map, some perform up to 25% worse.

•• WHY ?WHY ?

Page 14: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Examples of individual profiles Examples of individual profiles displayed in terms of the two factorsdisplayed in terms of the two factors

-2

-1

0

1

2

-2 -1 0 1 2

factor 1

fact

or 2

100

120

140

160

180

200

220

240

20 18 16 14 12 10 8 6 4 2

# electrode

curre

nt u

nits

Page 15: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Profile factor 1: overall levelProfile factor 1: overall level

-4

-3

-2

-1

0

1

2

100 120 140 160 180 200 220

avg. level

fact

or 1

T-level

C-level

NRT-level

T-NRT

T-NRT-T

C-NRT

Page 16: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Profile factor 2: slopeProfile factor 2: slope

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

-1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5

slope in current units / electrode

fact

or 2 T-level

C-level

NRT-level

Page 17: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Correlation between overCorrelation between over--all Tall T--, C, C-- and NRTand NRT--levelslevels

profiles

100

110

120

130

140

150

160

170

180

190

200

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

# electrode

cure

nt u

nits

T-levelC-levelNRT-levelT-NRTC-NRTT-NRT-T

-3

-2

-1

0

1

2

3

-3 -2 -1 0 1 2 3C-level

NR

T-le

vel

-3

-2

-1

0

1

2

3

-3 -2 -1 0 1 2 3T-level

NR

T-le

vel

scale unit equals ~ 20 current units

Corr. coeff = 0.64 Corr. coeff = 0.39

Page 18: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Speech perception Speech perception reredifference in overall leveldifference in overall level

profiles

100

110

120

130

140

150

160

170

180

190

200

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

# electrode

cure

nt u

nits

T-levelC-levelNRT-levelT-NRTC-NRTT-NRT-T

-4

-3

-2

-1

0

-3 -2 -1 0 1T-level

T-NR

T le

vel

worse speech similar speech-2

-1

0

1

2

-2 -1 0 1 2C- level

C-N

RT

leve

l

worse speech similar speech

scale unit equals ~ 20 current units

Page 19: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Correlation between tilts Correlation between tilts in Tin T--, C, C-- and NRTand NRT--levelslevels

-3

-2

-1

0

1

2

3

-3 -2 -1 0 1 2 3C-level

NR

T-le

vel

-3

-2

-1

0

1

2

3

-3 -2 -1 0 1 2 3T-level

NR

T-le

vel

Corr. coeff = 0.36

profiles

100

110

120

130

140

150

160

170

180

190

200

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

# electrode

cure

nt u

nits

T-levelC-levelNRT-levelT-NRTC-NRTT-NRT-T

Corr. coeff = 0.80

scale unit equals ~ 2/3 cu/electrode

Page 20: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

profiles

100

110

120

130

140

150

160

170

180

190

200

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

# electrode

cure

nt u

nits

T-levelC-levelNRT-levelT-NRTC-NRTT-NRT-T

Speech perception Speech perception reredifference in tiltdifference in tilt

-2

-1

0

1

2

-2 -1 0 1 2 3T-level

T-N

RT

leve

l

worse speech similar speech-2

-1

0

1

2

-2 -1 0 1 2 3

C-level

C-N

RT

leve

l

worse speech similar speech

scale unit equals ~ 2/3 cu/electrode

Page 21: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Individual example of poor Individual example of poor speech perceptionspeech perception

80

120

160

200

240

22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3

# electrode

curr

ent u

nits

T-level

C-level

NRT-level

T-NRT

C-NRT

Page 22: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Results 4Results 4•• The TThe T--, C, C-- and NRTand NRT--profiles can de described by profiles can de described by

only two factors: only two factors: overall leveloverall level and and tilttilt..•• For For overall leveloverall level the correlation between Tthe correlation between T--level level

and NRT threshold is weak (0.64), and NRT threshold is weak (0.64), for Cfor C--level and NRT threshold it is poor (0.39).level and NRT threshold it is poor (0.39).

•• For For tilttilt the correlation between Tthe correlation between T--level and NRT level and NRT threshold is satisfactory (0.80), threshold is satisfactory (0.80), for Cfor C--level and NRT threshold it is poor (0.36).level and NRT threshold it is poor (0.36).

Page 23: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Results 5Results 5•• NRT predicts the TNRT predicts the T--profile accurately but this profile accurately but this

may be unimportant to speech perception.may be unimportant to speech perception.•• NRT does not accurately predict the CNRT does not accurately predict the C--profile profile

whereas the Cwhereas the C--profile may be important to profile may be important to speech perception. Phoneme perception speech perception. Phoneme perception decreases with a marked increase in the tilt of decreases with a marked increase in the tilt of the Cthe C--level profiles.level profiles.

•• However, these results may have been affected However, these results may have been affected considerably by habituation to the initial MAP.considerably by habituation to the initial MAP.

•• Acceptance of stronger highAcceptance of stronger high--frequency frequency stimulation might stimulation might favourfavour speech perception.speech perception.

Page 24: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

Parametric fittingParametric fitting

120

140

160

180

200

22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3

# electrode

curr

ent l

evel

-2

-1

0

1

2

-2 -1 0 1 2

overall level

tilt

Page 25: Guido F. Smoorenburg • Christina Willeboer • Johannes E ...guido.smoorenburg.pagesperso-orange.fr/NRTfitting.pdf · • Guido F. Smoorenburg • Christina Willeboer • Johannes

An implementation of An implementation of parametric fittingparametric fitting