2.5 Data analysis
\(MAE=\frac{\sum_{i=1}^{n}\left|\alpha\par
\begin{matrix}i\\
\text{RESP}\\
\end{matrix}-\alpha\par
\begin{matrix}i\\
\text{TARG}\\
\end{matrix}\right|}{n}\) (1)
\(\alpha_{\text{RESP}}=g\bullet\alpha_{\text{TARG}}+b\) (2)
The mean absolute error (MAE) was calculated using Equation 1 to assess
the sound localisation accuracy under different conditions.
Theα RESP and α TARG referred
to the response azimuth (in degrees) and target azimuth (in degrees),
respectively. Additionally, the best linear fit of the target–response
relationship for each participant was also computed using Equation 2,
where g is the response gain (slope, dimensionless), and bis the response bias (offset in degrees). In this study, the right side
was defined as the impaired side; therefore, azimuth coordinates for
patients with left ear impairment and controls with NH with left ears
plugged were inverted.
Paired and independent t- tests were conducted to evaluate
differences under different test conditions. The p-values of
<0.05 and <0.001 were considered statistically
significant. SPSS version 26.0 and GraphPad Prism version 8.0 were used
to analyse the data and draw diagrams.
RESULTS
3.1
Hearing thresholds
For
patients with UMA, the mean ± SD of the unaided hearing threshold was
51.36
± 5.02 dB HL, which significantly improved to
27.64±2.38
dB HL with a mean FHG of
23.73
± 3.47 dB HL
(p
< 0.001). For the NH comparison group, the mean hearing
threshold was
15.36
± 3.88 dB HL. The aided mean hearing threshold of the UMA group was
still higher (worse) than the mean hearing threshold of the NH group
(p
< 0.001). Detailed outcomes across 0.5–4 kHz are depicted in
Fig. 1a.
3.2
Speech perception abilities
In Fig.1b-c, the average unaided WRS and SNR for patients with UMA were
18.27
± 14.63 % and -5 ± 1.18 dB SPL, respectively, while the average aided
WRS and SNR conspicuously changed to
85.45±7.38
% and -7.73 ± 1.42 dB SPL, respectively (WRS:
p
< 0.001; SNR:
p
< 0.05).
The
mean WRS and SNR of the comparison group were
99.27
± 1.35 % and -10.55 ± 2.77 dB SPL, respectively. Fig. 1b-c also
indicates significant differences in the speech levels between the
patients aided with BCDs and their peers with NH (WRS: p <
0.001; SNR: p < 0.05).
3.3
Sound localisation in patients with UMA
andstimulated
UHL
Fig.
2 shows the individual sound localisation target-response plots for two
children with UMA and one control (P6, P7, and N2) under monaural
(unaided and P, left column) and binaural (aided and UP, right column)
listening conditions. Under the unaided condition, P6 showed a poor
localisation ability and perceived most stimuli from the healthy ear
side. However, P7 exhibited relatively better sound localisation
accuracy than P6. When aided with the BCD, the sound localisation
accuracy improved in P6 (delta gain = 0.422, delta MAE = -14.28°) and
worsened in P7 (delta gain = -0.167, delta MAE = 18.58°). Under the P
condition, all data points of N2 fell along the diagonal dotted line,
indicating a sharply deteriorated localisation performance (gain = 0.12,
MAE = 65°).
Individual
data on gender, age, MAE, response gain, bias, and r2for all participants are presented in Table 2. The MAE and response gain
under monaural listening conditions (unaided and P) are plotted against
those under binaural listening conditions (aided and UP) in Fig. 3a–b.
Varying sound localisation performance was observed in the 11 children
with UMA under the unaided condition. When the mean gain and MAE of all
children with UMA were compared between the unaided and aided
conditions, no significant differences were identified (gain: p = 0.104,
MAE: p = 0.436).
Control
listeners showed good sound localisation performance in the UP
condition. All of them exhibited considerable deterioration after being
plugged (gain: p < 0.001; MAE: p < 0.001), with most
of them unable to localize the stimuli presented from the plugged side.
Although no significant difference was observed between the unaided and
P conditions (gain: p = 0.073; MAE: p = 0.073), the results indicated
that children with UMA showed better sound localisation performance
(smaller MAE and gain) than the stimulated UHL listeners. This
phenomenon might be related to the adaptation to congenital unilateral
asymmetric hearing loss; however, the benefit of this adaptation was
insufficient for children with UMA to localize sound as accurately as
the normal controls did.
3.4
Influence of a BCD on sound localisation accuracy
Fig. 3c–d
shows
the localisation accuracy of the patients with UMA and the NH controls
on the impaired (the
atretic
and plugged) side and the
contralateral
(the healthy and unplugged) side,
respectively.
A better sound localisation accuracy was observed in children with UMA
(43.18
± 30.58°
vs.
83.18 ± 37.82°,
p
< 0.05) on the impaired side than in controls in the P
condition (Fig. 3c).
When
aided with a BCD, there was no difference in the MAE between the unaided
and aided conditions on the impaired side
(43.18
±
30.58°
vs.
34.14
± 17.9°, p = 0.303) or the contralateral side
(26.97
± 24.68° vs.
27.42
± 14.52°, p = 0.79), indicating that the BCD use was not detrimental to
the original sound localisation ability of the patients with UMA.
According
to the criterion in a previous study, 11 children with UMA were divided
into two subgroups: good performers (n = 5; gain > 0.75)
and poor performers (n = 6; gain ≤ 0.75) 14. When the
MAE outcomes were separately compared bilaterally, a significantly
better sound localisation accuracy on the atretic side was observed in
good performers
(15.67°
± 10.71° vs. 66.11° ± 19.77°, p < 0.05, Fig. 4a). Linear
regression was further conducted to explore the predictive effect of
gain on the benefits of sound localisation accuracy by fitting BCDs
(delta MAE, MAEaided − MAEunaided). The
results revealed an evident relationship between response gain and delta
MAE (r2 = 0.553, p < 0.05), indicating that
children with UMA who have poor sound localisation performance (lower
gain) show more improvement after being fitted with BCDs (Fig. 4b).