109 / 162
107
5
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.
Another interesting observation relates to figure 2 , which shows that most T-levels are 20–35% of the
van der Beek/Briaire/Frijns
Audiol Neurotol 2015;20:1–16
DOI: 10.1159/000362779
12
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):
(4)
7
1 8 cos 15
cos 2
1
32
32
0 118
7
and
lectrode
M-level lectrode .
lectrode
.
lectrode
M-level
¬
¬
¬
®
®
®
e
e
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
¬
¬
¬
¬
®
®
®
®
¬
®
236
7
.
lectrode
¬
¬¬
®
®
®
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
M-level profile (with emphasis on higher frequencies)
was set during fitting in our clinic (see Subjects andMeth-
ods).
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
T-level – predicted T-level (dB)
–6
–3
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
3
6
T-level – predicted T-level (CU)
–60
–40
–20
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
16
20
40
60
a
c
b
M-level – predicted M-level (dB)
–6
–3
0
3
6
M-level – predicted M-level (CU)
–60
–40
–20
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
20
40
60
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
).
Downloadedby:
LeidenUniversity
145.88.209.33 - 11/23/20142:34:38PM
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).