Proefschrift_vd_Beek

6

60

40

3

20

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

–20

–3

T-level – predicted T-level (CU) –60 –40 b

T-level – predicted T-level (dB) –6 a

6

60

40

3

20

0

0

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

16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

–20

–3

M-level – predicted M-level (CU) –60 –40 d

c M-level – predicted M-level (dB) –6

useful way to fit children and other recipients, which are not able to give sufficient feed-back during fitting procedures. The use of these predictive equations matches well with the trend of using simplified fitting methods, which are mainly based on interpolation between electrodes and the fact that profiles are relatively flat for monopolar strategies [Plant et al., 2005; Boyd, 2010]. The main im-provement presented in this paper is the closed-set for-mulation in relation to the observation that the fitting lev-els (when expressed in decibels) show a level-indepen-dent increase towards the basal end of the electrode array (fig.5 a). Although equations 2 and 3 were derived on the basis of measured T-levels of individual electrodes, it is not clear to what extent these equations can be generalized, since all patients were fitted in the same clinic, which might have enhanced intersubject similarities. Further-more, it must be taken into account that, unlike the T-levels, the M-levels in our population were fitted with a preset profile with emphasis on the basal electrodes, which is definitely reflected in equations 4 and 5. Up to now, eCAP-based profiles still cannot provide proper settings and are only used as a starting point and continue to be used in combination with subjective pa-tient-derived information. Although eSRT or eABR pro-vides complementary information to eCAP measures, the latter are nowadays used more often for predicting levels, mainly for practical reasons. van der Beek/Briaire/Frijns Audiol Neurotol 2015;20:1–16 DOI: 10.1159/000362779 12 (4) 7 32 32 0 118 7 and lectrode . lectrode M-level ž Ÿ ® Ÿ ® Ÿ ® e e ( in dB ) (5) 1 18 cos 15 cos 2 1 0118 7 20 32 32 7 36 cos 15 cos 2 1 32 32 7 10 10 . ectrode . l ctr de lectrode lectrode lectrode M-level lectrode M-level M-level  ¬  ¬  ¬  ¬ ž ž ž ž ž ž ž ž ž ž ž ž ž ž Ÿ ® Ÿ ® Ÿ ® Ÿ ®  ¬  ž ž ž ž ž ž Ÿ ® Ÿ 236 7 . lectrode  ¬  ¬¬ ž ž ž ž ž ž ž ž ® Ÿ ® Ÿ ® e × × e e e e e e (in CU). Contrary to the situation for the T-levels, which were measured individually, this high predictability for the M- level profile was mainly due to the fact that a predefined The bars in figure 6a, b show the mean differences be- tween the predicted and m asur d T-levels, wh le the dashed lines indicate the lower and upper borders of the 95% prediction interval for the individual electrode con- tacts, expressed in decibels (fig. 6a) and clinical units (fig. 6b). Figure 6c, d shows the same data for the M-lev- els. It is clear that the size of the 95% prediction interval increases with the distance from electrode contact 7, at which T- and M-levels are measured, in spite of the fact that the model predicts the mean levels accurately along the whole array. Discussion The present paper demonstrates how the group profile of T- and M-levels in a relatively large population can be described in closed-set formulas and how this can serve as a starting point for fitting individual cochlear implant A fit comparable to figure 5c was made for the M-level profile (not shown), and, again, a high predictability could be obtained with a measurement on only one elec- trode contact (table 3b). On the basis of a similar mixed linear model, the M-l vels along the array could be pre- dicted with equations 4 and 5 (in decibels and clinical units, respectively): 1 8 cos 15 ž ž cos 2 ž ž 1 M-level lectrode . e lectrode  ¬  ž ž ¬  ž ¬ ž e M-level profile (with emphasis on higher frequencies) was set during fitting in our clinic (see Subjects andMeth- ods). Fig. 6. Prediction error means (bars) and 95% prediction intervals (dashed lines) of T-levels in decibels ( a ) and clinical units ( b ) as well as of M-levels in decibels ( c ) and clinical units ( d ). Fig. 6. Prediction error means (bars) and 95% prediction intervals (dashed lines) of T-levels in decibels (a) and clinical units (b) as well as of M-levels in decibels (c) and clinical units (d).

5

Downloadedby: LeidenUniversity 145.88.209.33 - 11/23/20142:34:38PM

Another interesting observation relates to figure 2 , which shows that most T-levels are 20–35% of the

107