Proefschrift_vd_Beek
–40
T-level –6 a
T-level
b
f the measured T-levels vs. predicted T-levels for the remaining 30% of the population for
–60
6
60
while the mean r was reduced at both ends of the array (to r = 0.86). R-values for electrodes 2, 5, 9 and 14 are not shown in figure 5b, since these electrodes were active in only less than 33% of the subjects (table 3). 40 The goodness of fit of equation 2 for the individual T-levels was tested in the remaining 30% of the measured data (fig. 5c; table 3a). Figure 5c shows scatter plots of the predicted T-levels versus the measured T-levels for all 16 electrode contacts, while table 3a provides the associated numerical data. Again, it is clear that the predictions are slightly better for the center region of the array. This procedure was repeated with a number of other random selections of 30% of the population, with essentially the same result. predicted T-levels versus the measured T-levels for all 16 electrode contacts, while table 3a provides the associated numerical data. Again, it is clear that the predictions are slightly be ter for the c nte region of the array. This pro- cedure was repeated with a number of other random se- lections of 30% of the population, with essentially the same result. To obtain a T-level profile expressed in clinical units, equation 2 can be reformulated as follows: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 c M-level – predicted M-level (dB) –6 –3 0 –20 0 d To obtain a T-level profile expressed in clinical units, equation 2 can be reformulated as follows: 3 20
de 7, yielding the
+
1 2 3 4 5 6 7
B ). (2) electrodes in the e results, while the array (to r = 0.86). are not shown in active in only less the individual T- of the measured scatter plots of the
M-level – predicted M-level (CU) –60 –40
Fig. 6. Prediction error means (bars) and 95% prediction intervals (dashed lin clinical units ( b ) as well as of M-levels in decibels ( c ) and clinical units ( d ).
7 T level lectrode T-level T-level × × e e - lectrode
1 0 01
2 2
e
.
lectrode
7 0 025 .
lectrode
7
e
10 10
20
2 2
0 2 .
lectrode
7 0 5 .
lectrode
7
e
e
(3)
(in CU).
5 tween the predicted and dashed lines indicate the lo 95% prediction interval for tacts, expressed in decibel (fig. 6b). Figure 6c, d shows els. It is clear that the size o increases with the distance which T- and M-levels are that the model predicts the the whole array. Downloaded by: Leiden University 145.88.209.33 - 11/23/2014 2:34:38 PM M-level profile (with emp was set during fitting in our ods). The bars in figure 6a, b s
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 electrode contact (table 3b). On the basis of a similar mixed linear model, the M-levels along the array could be predicted with equations 4 and 5 (in decibels and clinical units, respectively): van der Beek/Briaire/Frijns 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-levels along the array could be pre- dicted with equations 4 and 5 (in decibels and clinical units, respectively): 7 lectrode e
:1–16
¬ ®
¬ ®
¬
cos 2
e
M-level lectrode . e
lectrode
1 8 cos 15
1
®
32
32
(4)
0 118 .
lectrode
M-level
7
and
e
( in dB )
lectrode
7
e
(5)
M-level lectrode e
1 1 8 cos 15 20 .
¬ ®
¬ ®
¬ ®
¬ ®
lectrode
0 118 .
lectrode
cos 2
1
7
e
e
32
32
M-level
10
×
e
lectrode
7
Discussion
¬ ®
¬¬
¬ ®
36 cos 15
lectrode
2 36 .
lectrode
cos 2
1
7
e
e
®
®
32
32
M-level
×
10
(in CU).
e
lectrode
7
The present paper demo of T- and M-levels in a relat described in closed-set for as a starting point for fittin
M-level profile (with emphasis on higher frequencies) was set during fitting in our clinic (see Subjects and Methods). 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 between the predicted and measured T-levels, while 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 Tand M-levels are measured, in spite of the fact that the model predicts the mean levels accurately alongthe whole array. Audiol Neur tol 2015;20:1–16 DOI: 10.1159/000362779 12
van der B
103
Made with FlippingBook