Availability of binaural cues for bilateral implant recipients and bimodal listeners with and without preserved hearing in the implanted ear

Abstract

The purpose of the current study was to examine the availability of binaural cues for adult, bilateral cochlear implant (CI) patients, bimodal patients and hearing preservation patients using a multiple baseline, observational study design. Speech recognition was assessed using the Bamford-Kowal-Bench Speech-In-Noise (BKB-SIN) test as well as the AzBio sentences [Spahr et al., 2012] presented in a multi-talker babble at +5 dB signal-to-noise ratio (SNR). Test conditions included speech at 0° with noise presented at 0° (S₀N₀), 90° (S₀N₉₀), and 270° (S₀N₂₇₀). Estimates of summation, head shadow (HS), squelch, and spatial release from masking (SRM) were calculated. Though none of the subject groups consistently showed access to binaural cues, the hearing preservation patients exhibited a significant correlation between summation and squelch whereas the bilateral and bimodal participants did not. That is, the two effects associated with binaural hearing—summation and squelch—were positively correlated for only the listeners with bilateral acoustic hearing. This finding provides evidence for the supposition that implant recipients with bilateral acoustic hearing have access to binaural cues which should, in theory, provide greater benefit in noisy listening environments. It is likely, however, that the chosen test environment negatively affected the outcomes. Specifically, the spatially separated noise conditions directed noise toward the mic port of the behind-the-ear (BTE) hearing aid and implant processor. Thus it is possible that in more realistic listening environments for which the diffuse noise is not directed toward the processor/hearing aid mic, hearing preservation patients have binaural hearing cues for improved speech understanding.

INTRODUCTION

There are a number of published studies documenting the respective benefits of bilateral cochlear implantation, bimodal hearing, and hearing preservation cochlear implantation. Several studies have documented summation, head shadow (HS), binaural squelch (also commonly termed binaural unmasking) and spatial release from masking (SRM) for bilateral CI recipients [Buss et al., 2008; Eapen et al., 2009; Litovsky et al., 2006; Schleich et al., 2004; Wackym et al., 2007; Zeitler et al., 2008]. Since binaural hearing is not required for head shadow nor SRM, only summation and squelch are those cues that are indicative of binaural hearing. In previous studies with bilateral implant recipients, squelch estimates were quite small—in most cases ranging from 0.9 to 1.9 dB for adaptive or pseudo-adaptive measures [Litovsky et al., 2006; Schleich et al., 2004] and 8- to 18-percentage points for fixed SNR measures [Buss et al., 2008; Eapen et al., 2009; Laszig et al., 2004; Verhaert et al., 2012].

In contrast to bilateral implant users, bimodal listeners, without preserved hearing in the implanted ear, have exhibited greater variability and generally more modest estimates of the use of binaural cues. In a meta-analysis of 13 studies examining adult bimodal listeners, Schafer and colleagues reported significant bimodal summation (bimodal vs. CI alone) of 14 percentage points across studies for fixed SNR measures [Schafer et al., 2007]. For adaptive speech reception thresholds (SRTs), both Gifford and Dorman [Gifford and Dorman, 2012] and Morera and colleagues [Morera et al., 2012] demonstrated 3-dB summation effects for 11 and 15 adult bimodal listeners, respectively.

Estimates of binaural squelch for bimodal listeners have been much less frequently reported than that typically reported in the literature for bilateral implant recipients. Schafer and colleagues [Schafer et al., 2007] evaluated 3 studies [Dunn et al., 2005; Morera et al., 2005; Tyler et al., 2002] in a meta-analysis reporting squelch as the performance difference for the CI alone with noise directed to the hearing aid (HA) S₀N_HA versus the bimodal condition in the same noise configuration (S₀N_HA). Using this calculation, Schafer and colleagues [Schafer et al., 2007] calculated an across-study mean squelch estimate of 10.1-percentage points across the three studies, which was not found to be statistically significant. Using the same squelch calculation for squelch in an adaptive SRT, Morera and colleagues reported significant estimates of squelch ranging from 2.6 to 3.6 dB, across two test sessions [Morera et al., 2012]. They reported, however, that these estimates were largely driven by the results for two bimodal participants who demonstrated much better speech recognition performance with the HA over the implanted ear.

Bimodal listeners have demonstrated similar magnitude of head shadow effect as bilateral implant recipients—though with different estimates across ears, as expected, given the asymmetry in performance across the HA and CI ears. In a meta-analysis across 6 studies referencing the CI ear and 3 studies referencing the HA ear, Schafer et al. (2007) showed significant mean estimates of head shadow of 17.4-percentage points for the implanted ear and 61.1-percentage points for the HA ear.

Unilateral implant recipients with preserved hearing in the implanted ear (hearing preservation patients) have been shown to exhibit significant benefit for speech recognition with various noise sources [Dorman et al., 2012; Dunn et al., 2010; Gifford et al., 2010; Gifford et al., 2013; Rader et al., 2013], horizontal-plane localization [Dunn et al., 2010] and preserved interaural time differences [Gifford et al., 2013]. Thus, there is some evidence that these patients have access to binaural cues. However, no published study has attempted to compare the availability of binaural cues for these patients as well as patients with bilateral CIs and patients with bimodal hearing.

The hypothesis underlying this project was that implant recipients with bilateral acoustic hearing would exhibit significantly greater summation and squelch than either bilateral CI patients or bimodal listeners (the latter not having preserved hearing in the implanted ear). Our hypothesis was driven by our finding of access to interaural time difference (ITD) cues [Gifford et al., 2013] in hearing preservation patients and the fact that interaural spectral “mismatch” caused by disparate electrode insertion depths and electrode-to-neural interfaces for bilateral recipients has been shown to negatively affect squelch and summation [Yoon et al., 2012]. Thus, the primary objective of the current study was to examine the availability of binaural cues for adult bilateral cochlear implant (CI) recipients and bimodal listeners both with, and without, preserved hearing in the implanted ear.

MATERIALS AND METHODS

Subjects

Demographic information for the 81 study participants is shown in Tables 1 – 3 for the 30 bilateral, 35 bimodal and 16 hearing preservation participants, respectively. Variables provided include age at testing, gender, implant type, experience with implants, aided speech intelligibility index (SII) at 60 dB SPL as provided by the Audioscan Verifit real-ear measures, and Consonant Nucleus Consonant (CNC, [Peterson and Lehiste, 1962]) monosyllabic word recognition performance at 60 dBA. Of note is that not all of the implanted electrodes for the hearing preservation were specifically designed for hearing preservation purposes. Nevertheless, with minimally traumatic surgical techniques and patients having more preoperative hearing to potentially preserve, we are going to be encountering more clinical patients with measurable, and aidable, hearing in the implanted ear postoperatively. For the hearing preservation patients, the inclusion criteria specified audiometric thresholds 80 dB HL or better at 250 Hz and below. This criterion was based on two factors. The first was the half-gain rule as the maximum low-frequency gain for most in-the-ear (ITE) hearing aids is approximately 40 dB. The second basis for this criterion was that previous studies have shown that the majority of the EAS- or bimodal-based benefit is derived from acoustic hearing at 250 Hz and below [Dawson et al., 2004; Henry and Ricketts, 2003; Zhang et al., 2010].

Table 1.

Bilateral participant demographic information including age at testing, years of electric experience for the 1^st implant, years experience with the 2^nd implant, implants, processors, and CNC monosyllabic word recognition (% correct) in the 1^st CI, 2^nd CI and in the bilateral condition. Participants who were simultaneously implanted are shaded in grey.

Subject	Age	Years exp 1st CI	Years exp 2nd CI	Implants	Processors	CNC 1st CI	CNC 2nd CI	CNC bilateral
1	50	3.3	2.6	CI24RE(CA) x2	Freedom x2	88	80	84
2	69	13.3	8.9	CI22M, CI24RCA	Freedom x2	76	72	72
3	56	4.9	2.9	CI24RE(CA) x2	Freedom x2	88	80	84
4	46	10.5	4.1	C1.2, HR90K	Platinum BTE, Harmony	66	50	78
5	59	9.9	4.8	CI24RE(CA) x2	Freedom x2	56	72	90
6	52	8.8	2.7	CII, HR90K	Harmony x2	80	44	92
7	59	19.7	7.6	CI22M, CI24RE(CA)	Freedom x2	60	22	70
8	53	4.6	2.7	CI24RE(CA) x2	Freedom x2	74	80	84
9	22	3.4	3.4	CI24RE(CA) x2	Freedom x2	98	72	94
10	19	1.7	1.7	CI24RE(CA) x2	Freedom x2	98	90	98
11	47	21.0	3.1	CI22M explanted-reimplanted CI24RE(CA), CI512	CP810 x2	60	50	80
12	19	9.2	4.0	CI24RCS, CI24RE(CA)	Freedom x2	74	66	92
13	75	5.6	1.4	CI24RE(CA), CI512	Freedom, CP810	76	76	86
14	63	19.6	1.2	CI22M, CI512	Freedom, CP810	66	88	94
15	51	0.6	0.5	HR90K x2	Harmony x2	80	78	92
16	48	2.8	1.5	CI24RE(CA) x2	Freedom x2	80	68	84
17	62	6.6	1.4	HR90K x2	Harmony x2	70	48	80
18	59	7.1	0.7	HR90K x2	Harmony x2	82	70	90
19	67	2.5	1.2	CI24RE(CA) x2	Freedom x2	92	84	92
20	65	3.7	3.0	HR90K x2	Harmony x2	96	82	96
21	62	8.0	2.9	CI24RE(CA) x2	Freedom x2	36	56	58
22	60	2.4	2.1	Sonata (H) x2	Opus2 x2	74	74	78
23	37	1.1	0.9	CI512 x2	CP810 x2	94	90	96
24	63	0.9	0.7	CI512 x2	CP810 x2	46	46	72
25	55	7.5	2.0	CI24RE(CA), CI512	CP810 x2	74	88	92
26	29	8.3	8.1	HR90K x2	Harmony x2	42	68	78
27	45	8.5	8.5	Sonata (H) x2	Opus2 x2	84	78	86
28	61	2.8	0.6	Sonata (H) x2	Opus2 x2	64	46	72
29	50	9.3	9.3	Combi40+ (H) x2	Tempo+ x2	84	66	90
30	50	2.1	1.0	Sonata (H) x2	Opus2 x2	48	40	40
MEAN	51.8	7.0	3.2	N/A	N/A	73.5	67.5	83.1
STDEV	14.3	5.5	2.7	N/A	N/A	16.6	17.3	12.4

Table 3.

Hearing preservation participant demographic information including age at testing, years of electric experience, implant, processor, aided speech intelligibility index (SII) at 60 dB SPL in both the implanted and non-implanted ears, and CNC monosyllabic word recognition (% correct) in the following conditions: HA in implanted ear (ipsi HA), HA in non-implanted ear (contra HA), CI only, CI + ipsi HA, CI + contra HA, and CI + bilateral HAs.

Subject	Age	Years exp CI	Implant	Processor	Aided SII non-CI ear	Aided SII CI ear	CNC ipsi HA	CNC contra HA	CNC CI only	CNC CI + ipsi HA	CNC CI + contra HA	CNC CI + bilateral HA
1	71	1.1	Hybrid L24	Hybrid Freedom	31	24	24	26	68	82	74	78
2	70	4.7	CI24RCA	Freedom	18	6	0	10	94	90	98	94
3	69	1.0	Hybrid S8	Hybrid Freedom	57	44	16	50	42	74	78	82
4	52	0.7	CI512	CP810	46	9	6	54	98	96	98	98
5	49	4.5	CI24RE(CA)	Freedom	30	10	4	4	90	90	96	98
6	79	1.7	CI24RE(CA)	Freedom	12	16	0	26	80	92	96	94
7	61	2.1	Hybrid L24	Hybrid Freedom	32	34	21	48	76	94	90	88
8	61	1.0	CI512	CP810	19	6	10	30	84	82	80	84
9	45	1.4	Sonata (H)	Opus2	30	6	2	44	82	84	82	84
10	53	4.0	Hybrid S8	Hybrid Freedom	38	34	28	36	44	60	62	80
11	58	6.4	CI24RCA	CP810	31	16	18	24	82	86	84	86
12	52	2.0	Hybrid S8	Hybrid Freedom	32	12	12	48	62	58	86	86
13	34	4.1	CI24RCA	Esprit 3G	35	9	2	28	80	84	88	90
14	52	1.4	Hybrid L24	Hybrid Freedom	44	36	26	52	76	90	88	88
15	77	0.6	Sonata (H)	Opus2	22	12	0	14	48	56	46	56
16	83	0.9	CI422	CP810	29	15	0	12	26	42	40	46
MEAN	58.9	2.4	N/A	N/A	31.6	18.1	10.6	31.6	70.8	78.8	80.4	83.3
STDEV	12.5	1.8	N/A	N/A	11.2	12.4	10.23	16.1	20.7	16.1.3	17.4	14.0

Participants ranged in age from 19 to 90 years old with a mean of 60.5 years. Twenty-six of thirty bilateral participants received their implants in sequential surgeries. Participants had an average of 4.6 years (range of 0.5 to 21.0 years) experience with the first implant (across all 81 participants). For the sequential bilateral participants, average experience with the second implant was 3.2 years (range of 0.5 to 9.3 years) with a mean difference between the two implants of 4.4 years (range of 0.1 to 18.4 years). Though the CNC monosyllabic word recognition scores are not demographic in nature, they were included here to characterize outcomes for what is generally considered the most commonly reported metric for English-speaking recipients. More detail will be provided for CNC monosyllabic word recognition in the Results section.

Individual and mean audiograms for the non-implanted ears of the unilateral implant recipients, with and without preserved acoustic hearing in the implanted ear, are shown in Figure 1A. Figure 1B displays individual and mean postoperative audiograms for the implanted ear for the participants with preserved acoustic hearing in the implanted ear. Audiograms were obtained for all participants immediately prior to experimentation.

FIGURE 1.

Individual and mean audiometric thresholds (in dB HL) obtained on the day of testing for the non-implanted ears of the bimodal and hearing preservation patients as well as the implanted ear of the hearing preservation patients. Error bars represent ± 1 standard deviation.

Methods

Speech perception was assessed during one or two visits, depending upon listener preference, using recorded stimuli presented in a sound-treated booth at a calibrated level of 60 dBA. For the 3 participants preferring to participate over two test sessions, the duration between sessions ranged between 1 day and 1 week. Participants used their everyday CI programs and were not permitted to manipulate settings during testing. For Nucleus implant recipients, all participants used the ‘EVERYDAY’ setting which makes use of the default directional mic setting with the addition of Autosensitivity (ASC) and Adaptive Dynamic Range Optimazation (ADRO, [Dawson et al., 2004]). Test conditions for the speech in noise testing included speech at 0° with noise presented at 0° (S₀N₀), 90° (S₀N₉₀), and 270° (S₀N₂₇₀). All three listening conditions were completed for each individual ear alone as well as the bilateral, best-aided condition (bilateral, bimodal, or CI + bilateral hearing aids).

For unilateral implant recipients wearing hearing aids in one or both ears, hearing aid audibility was verified for 60-dB-SPL speech immediately prior to experimentation using real-ear measurements with NAL-NL1 prescriptive targets [Dillon et al., 1998]. Participants for whom hearing aids were undershooting target audibility by more than 5 dB at one or more frequencies had their hearing aids reprogrammed to match NAL-NL1 targets. For participants whose own hearing aids could not be adjusted due to lack of reserve gain or incompatibility with NOAH programming software, laboratory stock hearing aids (Phonak Naida III UP with deactivated Sound Recover, i.e. nonlinear frequency compression) were programmed and used for testing purposes. This occurred for six of the bimodal subjects (18, 23, 24, 25, 32 and 33). For these six subjects, testing was only completed with the stock hearing aids programmed to NAL-NL1 target audibility.

For BKB-SIN, the SNR-50—or the SNR at which approximately 50% correct performance would be expected—was recorded for each condition. Two paired lists (e.g., 1A/B and 2A/B) were presented for each listening condition with the average SNR-50 across the paired lists reported as mean performance per participant. For the AzBio sentences at +5 dB SNR, performance was recorded in percent words correct. Two, 20-sentence lists were run for each listening condition with the average performance across the two lists recorded for each participant. Listening conditions were counterbalanced across participants in each listener group with the list numbers chosen in a quasi-random manner.

In addition to the speech in noise testing, CNC word recognition was also assessed for all 81 subjects for each ear alone as well as in the bilateral, best-aided condition. For the hearing preservation subjects, additional listening conditions were tested including hearing aid (HA) in the implanted ear (ipsilateral HA), CI plus ipsilateral HA, CI plus contralateral HA (commonly termed the “bimodal” condition), as well as CI plus bilateral HAs (i.e., bilateral best aided). All speech stimuli were presented at a calibrated level of 60 dBA.

RESULTS

CNC word recognition

As shown in Tables 1–3, CNC word recognition performance was obtained for all 81 participants in each individual ear condition as well as the bilateral, best-aided condition. Mean implant only performance across all 81 participants was 72.3% with a range of 26 to 98% correct¹. Considering the poorer hearing ear for all participants, mean CNC word recognition was 35.3% with a range of 0 to 90% with the lowest scores belonging to the unilateral recipients’ non-implanted ears. Considering the bilateral, best-aided condition (i.e. all devices both ears) for all 81 participants, mean CNC word recognition was 82.0% with a range of 40 to 98% correct.

For CNC word recognition, a two-way, repeated-measures analysis of variance (ANOVA) was completed with the subject group (bimodal, bilateral, hearing preservation) and listening condition (better ear, poorer ear and bilateral best aided) as variables. For this analysis, the better hearing ear for the hearing preservation patients was defined as the implant plus the ipsilateral HA (see Table 3) and the bilateral best-aided condition included both ears with all available devices. In other words, the bilateral best-aided condition included the implant plus bilateral HAs for the hearing preservation subjects. The repeated-measures ANOVA revealed a significant effect of subject group [F_{(2, 78)} = 21.1, p < 0.001], listening condition [F_{(2, 78)} = 437.7, p < 0.001], and a significant interaction [F₍₄₎ = 65.0, p < 0.001]. Post hoc multiple comparisons using the Holm-Sidak test revealed no significant difference across the three subject groups for the better hearing ear (p > 0.26 for all comparisons). Considering the bilateral best-aided condition, there was also no significant difference across the three groups of subjects (p > 0.24 for all comparisons). Not unexpectedly, however, there was a significant difference for the poorer hearing ear, with the bilateral subjects achieving significantly higher levels of word recognition than both the bimodal (t = 14.7, p < 0.001) and the hearing preservation subjects (t = 7.0, p < 0.001). The hearing preservation subjects scored significantly higher than the bimodal subjects (t = 4.7, p < 0.001) for the poorer hearing ear which undoubtedly reflects lower (i.e. better) audiometric thresholds in the non-implanted ears for the hearing preservation participants (Figure 1) and associated higher (i.e. better) aided SII values (see Tables 2 and 3).

Table 2.

Bimodal participant demographic information including age at testing, years of electric experience, implant, processor, aided speech intelligibility index (SII) at 60 dB SPL, and CNC monosyllabic word recognition (% correct) in the HA, CI and in the bimodal conditions.

Subject	Age	Years exp CI	Implant	Processor	Aided SII	CNC HA	CNC CI	CNC bimodal
1	76	2.1	CI24RE(CA)	Freedom	26	22	54	80
2	82	1.0	CI24RE(CA)	Freedom	51	28	64	48
3	39	4.8	CI24RCA	Esprit 3G	21	38	84	96
4	74	6.2	CI24RE(CA)	Freedom	4	0	84	84
5	64	1.9	HR90K	Harmony	22	0	80	90
6	66	1.5	CI24RE(CA)	Freedom	10	0	90	90
7	58	5.5	HR90K	Harmony	19	0	90	78
8	70	4.8	CI24RCA	Freedom	28	0	92	86
9	82	1.8	CI24RE(CA)	Freedom	22	2	68	60
10	65	1.3	CI24RE(CA)	Freedom	12	0	90	80
11	77	0.6	CI24RE(CA)	Freedom	29	24	84	84
12	76	1.0	CI24RE(CA)	Freedom	17	14	50	60
13	34	0.8	CI24RE(CA)	Freedom	36	10	80	84
14	90	1.0	CI24RE(CA)	Freedom	36	20	74	84
15	50	0.5	HR90K	Harmony	40	30	88	88
16	76	13.2	CI24RE(CA)	Freedom	18	6	74	74
17	47	10.7	CI24M	Freedom	15	6	72	82
18	41	3.2	HR90K	Harmony	7	0	88	82
19	52	0.6	CI512	CP810	43	54	98	98
20	64	0.8	CI512	CP810	20	8	84	86
21	78	5.6	CI24RE(CA)	Freedom	20	2	52	68
22	69	2.1	CI24RE(CA)	Freedom	35	26	92	90
23	74	5.9	CI24RE(CA)	Freedom	15	0	80	76
24	79	0.9	CI24RE(CA)	Freedom	16	10	50	72
25	84	7.3	Combi40+	Tempo+	6	0	78	68
26	62	0.6	CI24RE(CA)	Freedom	15	0	46	72
27	68	6.4	CI24RCA	Freedom	21	18	94	90
28	72	5.3	CI24RE(CA)	Freedom	28	6	92	88
29	86	0.8	CI512	CP810	33	22	38	64
30	68	1.9	CI512	CP810	23	0	68	84
31	68	3.5	HR90K	Harmony	37	18	74	80
32	72	3.4	CI24RE(CA)	Freedom	15	0	76	84
33	73	2.2	CI512	CP810	28	30	91	94
34	79	10.7	CI24RCA	CP810	38	34	60	78
35	64	5.3	HR90K	Harmony	40	6	90	96
MEAN	68.0	3.6	N/A	N/A	24.2	12.4	76.3	80.5
STDEV	13.5	3.2	N/A	N/A	11.4	13.9	15.8	11.2

Summation for CNC word recognition can be estimated by subtracting the score for the better hearing ear from the bilateral, best-aided score. Summation estimates were 6.1, 4.3, and 4.4-percentage points for the bilateral, bimodal and hearing preservation subjects, respectively. Statistical analysis was completed for summation estimates across subject groups. A Kruskal-Wallis ANOVA on ranks was completed as the assumption of normality was not met. There was no significant difference [H₍₂₎ = 1.6, p = 0.45] across the three subject groups for summation observed with CNC word recognition in quiet.

Speech in noise: S₀N₀

Figures 2A and 2B display box and whisker plots for BKB-SIN (SNR-50) and AzBio at +5 dB (% correct) for the S₀N₀ listening condition. The box extends from the 25th to 75th percentiles with the horizontal line in the middle representing the median. The whiskers extend from the minimum to the maximum value for all individual data thus displaying the range of scores for any given condition. In this and all subsequent figures, the shaded bars represent bilateral subjects, the unfilled bars represent bimodal subjects and the patterned bars represent hearing preservation subjects.

FIGURE 2.

Box and whisker plots for BKB-SIN (SNR-50) and AzBio at +5 dB (% correct) for the S₀N₀ listening condition (i.e. speech and noise at 0° azimuth). The box extends from the 25th to 75th percentiles with the horizontal line in the middle representing the median. The whiskers extend from the minimum to the maximum value for all individual data thus displaying the range of scores for any given condition. The shaded bars represent bilateral subjects, the unfilled bars represent bimodal subjects and the patterned bars represent hearing preservation subjects.

With reference to the BKB-SIN scores shown in Figure 2A, mean SNR-50 for the better-hearing ear was 6.8 dB for bilateral, 7.2 dB for bimodal, and 7.4 dB for hearing preservation subjects. Mean SNR-50 for the poorer hearing ear was 10.0 dB for bilateral, 19.3 dB for bimodal and 15.0 dB for hearing preservation subjects. For the best-aided condition, mean SNR-50 was 5.8 dB for bilateral, 6.3 dB for bimodal, and 5.3 dB for hearing preservation subjects. A two-way ANOVA was completed with listening condition and subject group as the variables. The analysis revealed a significant main effect of listening condition [F₍₂₎ = 129.8, p < 0.001], subject group [F₍₂₎ = 23.5, p < 0.001] and a significant interaction [F₍₂₎ = 17.1, p < 0.001]. Post-hoc analysis using the Holm-Sidak statistic showed no difference across the three subject groups for either the better-hearing ear or the best-aided condition (p > 0.11 for all comparisons). There were significant differences across the subject groups for the poorer-hearing ear with all subject groups being significantly different from one another (p < 0.001 in all cases).

With reference to the AzBio sentence recognition at +5 dB SNR (Figure 2B), mean performance (in percent correct) for the better hearing ear was 56.8% for bilateral, 49.2% for bimodal, and 60.1% for hearing preservation subjects. Mean performance for the poorer hearing ear was 36.1% for bilateral, 5.7% for bimodal and 16.7% for hearing preservation. For the best-aided condition, mean performance was 65.5% for bilateral, 58.6% for bimodal, and 65.9% for hearing preservation subjects. As with the BKB-SIN results in the S₀N₀ condition, a two-way ANOVA was completed with listening condition and subject group as the variables. The analysis revealed a significant main effect of listening condition [F₍₂₎ = 116.2, p < 0.001], subject group [F₍₂₎ = 15.9, p < 0.001] and a significant interaction [F₍₂₎ = 4.4, p = 0.002]. Holm-Sidak post-hoc analysis showed no difference across the three groups for either the better-hearing ear or the best-aided condition (p > 0.10 for all comparisons). There were significant differences across the groups for the poorer-hearing ear with all three groups being significantly different from one another (bilateral vs. bimodal, p < 0.001; bilateral vs. hearing preservation, p = 0.001; bimodal vs. hearing preservation, p = 0.04).

Summation

Estimates of summation were gathered from Figures 2A and B as the difference score between the better-hearing ear and best-aided condition. Box and whisker plots are shown in Figure 3. For BKB-SIN (Figure 3A), mean summation was 1.0 dB, 0.9 dB and 1.9 dB for the bilateral, bimodal and hearing preservation participants, respectively. Individual estimates of summation ranged from −3.5 to 7.3 dB for bilateral, −2 to 4 dB for bimodal and −1.5 to 6.0 dB for hearing preservation participants. A one-way ANOVA revealed no difference across the subjects groups in terms of summation effects (F = 1.7, p = 0.18). For AzBio sentence recognition at +5 dB (Figure 3B), mean summation was 8.5, 9.5, and 5.4 percentage points for the bilateral, bimodal and hearing preservation subjects, respectively. Individual estimates of summation ranged from −13 to 52 percentage points for bilateral, −13 to 25 percentage points for bimodal and −12 to 42 percentage points for hearing preservation participants. As with BKB-SIN, a one-way ANOVA revealed no difference across the groups (F = 0.62, p = 0.54).

FIGURE 3.

Box and whisker plots for BKB-SIN (in dB) and AzBio at +5 dB (in percentage points), in panels A and B, respectively. The box extends from the 25th to 75th percentiles with the horizontal line in the middle representing the median. The whiskers extend from the minimum to the maximum value for all individual data thus displaying the range of scores for any given condition. The shaded bars represent bilateral subjects, the unfilled bars represent bimodal subjects and the patterned bars represent hearing preservation subjects.

Speech in noise: spatially separated listening conditions

Figures 4A – C display box and whisker plots for BKB-SIN scores (SNR-50) for the bilateral, bimodal and hearing preservation subjects for the spatially separated listening conditions (S₀N₉₀ and S₀N₂₇₀). Figures 5A – C display the same box and whisker plots as Figure 4 but for AzBio sentence recognition at +5 dB SNR. With the data shown in Figures 2, 4 and 5, we are able to calculate estimates of head shadow, SRM and squelch.

FIGURE 4.

Box and whisker plots for BKB-SIN (SNR-50) for the bilateral (panel A), bimodal (panel B), and hearing preservation subjects (panel C) for the spatially separated listening conditions with speech at 0° and noise at 90° (S₀N₉₀) and speech at 0° degrees and noise at 270° (S₀N₂₇₀).

FIGURE 5.

Box and whisker plots for AzBio sentences at +5 dB (percent correct) for the bilateral (panel A), bimodal (panel B), and hearing preservation subjects (panel C) for the spatially separated listening conditions with speech at 0° and noise at 90° (S₀N₉₀) and speech at 0° degrees and noise at 270° (S₀N₂₇₀).

Head shadow

Estimates of head shadow (HS) were calculated as follows:

Bilateral subjects

• HS for 1^st CI = score for 1^st CI (noise to 2^nd CI) − score for 1^st CI (noise to 1^st CI)

• HS for 2^nd CI = score for 2^nd CI (noise to 1^st CI) − score for 2^nd CI (noise to 2^nd CI)

Bimodal subjects

• HS for CI ear = score for CI ear (noise to HA ear) − score for CI ear (noise to CI ear)

• HS for HA ear = score for HA ear (noise to CI ear) − score for HA ear (noise to HA ear)

Hearing preservation subjects

• HS for CI ear = score for CI ear (noise to non-CI ear) − score for CI ear (noise to CI ear)

• HS for non-CI ear = score for non-CI ear (noise to CI ear) − score for non-CI ear (noise to non-CI ear)

Estimates of head shadow are shown in Figure 6A and B for BKB-SIN and AzBio at +5 dB SNR, respectively. For BKB-SIN, mean HS for the bilateral CI recipients was 7.6 and 5.3 dB for the 1^st and 2^nd implanted ears, respectively. Mean HS for the bimodal listeners was 7.6 and 3.3 dB for the CI and HA ears, respectively. Mean HS for the hearing preservation subjects was 5.0 and 4.8 dB for the CI and non-CI ears, respectively. A two-way ANOVA with subject group and listening condition (poorer vs. better ear) revealed no effect of subject group [F₍₂₎ = 1.2, p = 0.30], a significant effect of listening condition [F₍₁₎ = 7.7, p = 0.006], and no interaction [F_(2,1) = 2.4, p = 0.09]. Thus there was no difference in HS across the subject groups but estimates were significantly greater for the better hearing ear.

FIGURE 6.

Box and whisker plots displaying estimates of head shadow for BKB-SIN (in dB, panel A) and AzBio at +5 dB SNR (in percentage points, panel B) for the poorer ear, better ear, and best-aided conditions.

For AzBio at +5 dB SNR, mean HS for the bilateral CI recipients was 34.7- and 32.9-percentage points for the 1^st and 2^nd implanted ears, respectively. Mean HS for the bimodal listeners was 36.6- and 7.1-percentage points for the CI and HA ears, respectively. Mean HS for the hearing preservation subjects was 22.0- and 11.8-percentage points for the CI and non-CI ears, respectively. A two-way ANOVA completed with subject group and listening condition revealed a significant effect of subject group [F₍₂₎ = 8.7, p < 0.001], a significant effect of listening condition [F₍₁₎ = 14.0, p < 0.001], and a significant interaction [F_(2,1) = 8.2, p < 0.001]. Holm-Sidak post-hoc analysis revealed that the HS estimates were significantly greater for the bilateral implant recipients as compared to both the bimodal (t = 3.4, p = 0.001) and the hearing preservation (t = 3.8, p = 0.0002) participants. There was no significant difference across the three subject groups for the better-hearing ear (p > 0.05 for all comparisons). There were significant differences across the subject groups for the poorer hearing ear with all three groups being significantly different from one another (bilateral vs. bimodal, p = 0.03; bilateral vs. hearing preservation, p = 0.03; bimodal vs. hearing preservation, p = 0.01).

Spatial release from masking (SRM)

Estimates of SRM were calculated as follows:

Bilateral subjects

• SRM for 1^st CI = score for 1^st CI (noise to 2^nd CI) − score for 1^st CI (noise at 0°)

• SRM for 2^nd CI = score for 2^nd CI (noise to 1^st CI) − score for 2^nd CI (noise at 0°)

Bimodal subjects

• SRM for CI ear = score for CI ear (noise to HA ear) − score for CI ear (noise at 0°)

• SRM for HA ear = score for HA ear (noise to CI ear) − score for HA ear (noise at 0°)

Hearing preservation subjects

• SRM for CI ear = score for CI ear (noise to non-CI ear) − score for CI ear (noise at 0°)

• SRM for non-CI ear = score for non-CI ear (noise to CI ear) − score for non-CI ear (noise at 0°)

For all subject groups, SRM in the best-aided condition

• SRM for best aided = score for best-aided condition (noise to poorer ear) − score for best-aided condition (noise at 0°)

Estimates of SRM in the best-aided condition are shown in Figure 7A and B for BKB-SIN and AzBio at +5 dB SNR, respectively. For BKB-SIN, mean SRM for the bilateral CI recipients was 3.5 and 3.8 dB for the 1^st and 2^nd implanted ears, respectively, and 5.1 dB for the bilateral best-aided condition. Mean SRM for the bimodal listeners was 5.7 and 1.5 dB for the CI and HA ears, respectively, and 4.9 dB for the bimodal best-aided condition. Mean SRM for the hearing preservation subjects was 4.0 and 2.1 dB for the CI and non-CI ears, respectively, and 3.8 dB for the best-aided condition (CI + bilateral HA). A one-way ANOVA completed for SRM observed revealed no effect of subject group for the better hearing ear [F₍₂₎ = 1.4, p = 0.25], nor for the best-aided condition [F₍₂₎ = 1.3, p = 0.29] but a significant effect of group for the poorer hearing ear [F₍₂₎ = 11.1, p < 0.001]. For the poorer hearing ear, post hoc testing (Holm-Sidak) revealed that bimodal subjects’ SRM was significantly poorer than both bilateral (t = 4.5, p = 0.00002) and hearing preservation (t = 3.0, p = 0.004) subjects, but that bilateral and hearing preservation were not different (p = 0.49).

FIGURE 7.

Box and whisker plots displaying estimates of spatial release from masking (SRM) for BKB-SIN (in dB, panel A) and AzBio at +5 dB SNR (in percentage points, panel B) for the best-aided condition.

For AzBio at +5 dB SNR, mean SRM for the bilateral CI recipients was 20.9- and 16.3-percentage points for the 1^st and 2^nd implanted ears, respectively, and 20.4-percentage points for the best-aided condition. Mean SRM for the bimodal listeners was 19.9- and 2.7-percentage points for the CI and HA ears, respectively, and 14.1-percentage points for the best-aided condition. Mean SRM for the hearing preservation subjects was 8.4- and 11.5-percentage points for the CI and non-CI ears, respectively, and 10.0-percentage points for the best-aided condition. A one-way ANOVA completed for SRM observed revealed no effect of subject group for the better hearing ear [F₍₂₎ = 2.2, p = 0.12], nor for the best-aided condition [F₍₂₎ = 0.14, p = 0.87] but a significant effect of group for the poorer hearing ear [F₍₂₎ = 13.3, p < 0.001]. For the poorer hearing ear, Holm-Sidak post-hoc analysis revealed that the SRM estimates were significantly greater for the bilateral implant recipients as compared to both the bimodal (t = 5.2, p < 0.0001) and the hearing preservation (t = 2.2, p = 0.049) participants.

Squelch

Estimates of squelch were calculated as follows:

Bilateral subjects

• Squelch for 1^st CI = score for bilateral CI (noise to 2^nd CI) − score for 1^st CI (noise to 2^nd CI)

• Squelch for 2^nd CI = score for bilateral CI (noise to 1^st CI) − score for 2^nd CI (noise to 1^st CI)

Bimodal subjects

• Squelch for CI ear = score for bimodal (noise to HA ear) − score for CI ear (noise to HA ear)

• Squelch for HA ear = score for bimodal (noise to CI ear) − score for HA ear (noise to CI ear)

Hearing preservation subjects

• Squelch for CI ear = score for best aided (noise to non-CI ear) − score for CI ear (noise to non-CI ear)

• Squelch for non-CI ear = score for best aided (noise to CI ear) − score for non-CI ear (noise to CI ear)

For BKB-SIN, because a lower score represents better performance, squelch estimates obtained via the above-referenced equations yielded negative numbers for positive squelch and positive numbers for negative squelch. For ease of reporting and interpretation, all squelch estimates for the BKB-SIN metric have been inverted such that positive squelch will be reported as positive numbers and vice versa. For BKB-SIN, mean squelch for the bilateral CI recipients was 0.9 and 2.3 dB for the 1^st and 2^nd implanted ears, respectively. Mean squelch for the bimodal listeners was −0.7 and 10.4 dB for the CI and HA ears, respectively. Mean squelch for the hearing preservation subjects was 0.5 and 7.2 dB for the CI and non-CI ears, respectively. A two-way ANOVA completed with subject group and listening condition revealed an effect of subject group [F₍₂₎ = 12.3, p < 0.001], listening condition [F₍₁₎ = 99.9, p < 0.001], and a significant interaction [F_(2,1) = 26.7, p < 0.001]. Holm-Sidak post-hoc analysis revealed that the squelch estimates were significantly lower for the bilateral implant recipients as compared to both the bimodal (t = 3.4, p = 0.001) and the hearing preservation (t = 3.8, p = 0.0002) participants—an effect that was driven primarily by the non-CI ears for the bimodal and hearing preservation participants. There was no significant difference across the three subject groups for the better-hearing ear (p > 0.006 for all comparisons). There were significant differences across the subject groups for the poorer-hearing ear with all three groups being significantly different from one another (bilateral vs. bimodal, p < 0.0001; bilateral vs. hearing preservation, p < 0.001; bimodal vs. hearing preservation, p = 0.006).

For AzBio sentence recognition at +5 dB SNR, mean squelch estimates for the bilateral CI recipients were 6.1- and 7.2-percentage points for the 1^st and 2^nd implanted ears, respectively. Mean squelch for the bimodal participants was 0.2- and 32.7-percentage points for the CI and HA ears, respectively. Mean squelch for the hearing preservation subjects was 1.7- and 42.1-percentage points for the CI and non-CI ears, respectively. Statistical analysis revealed a significant effect of subject group [F₍₂₎ = 5.7, p = 0.004], a significant effect of listening condition [F₍₁₎ = 47.2, p < 0.001], and a significant interaction [F_(2,1) = 11.9, p < 0.001]. Post-hoc analysis using the Holm-Sidak test revealed that squelch was significantly lower for the bilateral implant recipients as compared to both the bimodal (t = 2.8, p = 0.0017) and the hearing preservation (t = 3.0, p = 0.003) participants—as discussed above with BKB-SIN, this outcome was primarily driven by the non-CI ears for the bimodal and hearing preservation participants. There was no difference across the three subject groups for the better-hearing ear (p > 0.25 for all comparisons). There was a significant difference across the subject groups for the poorer-hearing ear with the bilateral group demonstrating significantly less squelch than both bimodal listeners (p < 0.001) and hearing preservation (p <0.001) participants.

DISCUSSION

The results of the current study were largely consistent with previous reports in the literature examining availability of binaural cues for bilaterally implanted adults. Mean estimates of HS for the implanted ears were consistent with previous reports in the literature ranging from 5.0 to 7.6 dB for the pseudo-adaptive BKB-SIN (e.g., [Gantz et al., 2002; Litovsky et al., 2006; Muller et al., 2002; Schleich et al., 2004; Schon et al., 2002]) and 22.0- to 36.6-percentage points for AzBio sentences at +5 dB (e.g., [Buss et al., 2008; Eapen et al., 2009]). Mean estimates of squelch for the bilaterally implanted patients were also consistent with previous reports ranging from 0.9 to 2.3 dB for the BKB-SIN measure [Litovsky et al., 2006; Schleich et al., 2004] and 6.1 to 7.2-percentage points for AzBio sentence recognition at +5 dB SNR [Buss et al., 2008; Eapen et al., 2009]. Summation was also roughly consistent with previous reports for bilaterally implanted adults, though slightly lower in the current study, with mean summation being 1.0 dB for BKB-SIN [Litovsky et al., 2006; Schleich et al., 2004; Zeitler et al., 2008] and 8.5-percentage points for AzBio sentences at +5 dB SNR [Buss et al., 2008; Eapen et al., 2009]. Thus the results of the current study with simultaneous and sequential bilateral implant recipients has replicated the findings of previous reports—most of which examined bilateral recipients who had received their implants in the same surgery.

Considering these effects for the unilaterally implanted participants, neither estimates of HS, squelch, nor SRM were significantly different from that observed for the bilateral implant recipients—when considering the implanted ear. Estimates of summation were also not different for the bimodal and hearing preservation subjects with acoustic hearing as compared to the bilateral implant users.

Considering the non-implanted ear for bimodal and hearing preservation participants, estimates of squelch were significantly greater than that observed for any of the implanted ears. The reason for this seemingly paradoxical finding is related to the underlying assumption of symmetry in performance across ears. That is, squelch provides information regarding improvement in speech understanding provided by adding an ear with a poorer SNR. For the bimodal and hearing preservation patients, the improvement gained by adding the implanted ear was disproportionally greater than the improvement gain for bilateral implant users given that the CI ear yields significantly higher speech understanding than the non-implanted ears. Thus the current squelch estimates for the bimodal and hearing preservation listeners simply do not provide an accurate description as the ear with the better SNR was neither the better performing ear nor equivalent in performance to the ear with the poorer SNR. For this reason, Morera et al. (2012) suggested the use of a modified equation for describing bimodal squelch effects as follows:

$$\text{squelch}=\text{Bimodal}\text{score}(\text{noise}\text{to}\text{non}-\text{CI}\text{ear})-\text{CI}\text{score}(\text{noise}\text{to}\text{non}-\text{CI}\text{ear})$$

This is the equation we used to describe squelch for the implanted ears of bimodal and hearing preservation patients for which mean estimates were −0.7 and 0.5 dB for BKB-SIN and 0.2- and 1.7-percentage points for AzBio sentences at +5 dB.

Given that some hearing preservation patients have been shown to exhibit some preserved sensitivity to ITDs for low-frequency stimuli [Gifford et al., 2013], it is reasonable to ask why the current hearing preservation patients failed to exhibit summation and, more importantly, squelch. It is likely that there are a number of potential explanations for this finding. One potential explanation may relate to the degree of summation effects that could be used in the listening paradigm used here. As seen in Figures 3A and B, the range of measurable summation for the hearing preservation patients was quite variable with some exhibiting summation effects of 6 dB for BKB-SIN and 42-percentage points for AzBio at +5 dB. By completing correlation analyses for summation and squelch for all listener groups in the current study, there was a significant positive correlation observed between summation and squelch for the hearing preservation patients (BKB-SIN r = 0.54, p = 0.04; AzBio r = 0.71, p = 0.01) as displayed in Figure 9. The correlation analyses for the bilateral and bimodal participants, however, were not significant for either speech metric. One could contend that these data support the current hypothesis that implant recipients with preserved acoustic hearing in the implanted ear should have greater availability to binaural cues. In other words, those implant recipients with bilateral acoustic hearing who exhibit summation effects in the current experimental paradigm were also more likely to benefit from binaural squelch than both bilateral implant recipients and bimodal listeners.

FIGURE 9.

Hearing preservation patients’ individual squelch estimates are plotted as a function of summation for BKB-SIN (dB, panel A) and AzBio +5 dB (percentage points, panel B). Diagonal dashed lines represent a linear regression function representing the Pearson correlation coefficient shown in each panel.

Another possible reason for the lack of binaural effects for the hearing preservation patients is the chosen test environment. The current study incorporated the classic experimental design for calculating HS and squelch including the following: S₀N₀, S₀N₉₀ and S₀N₂₇₀. Presenting noise directly to the side of the listener, as in S₀N₉₀ and S₀N₂₇₀, is not the best choice of conditions given the microphone port location for CI processors and behind-the-ear (BTE) hearing aids. Festen and Plomp (1986) demonstrated that the BTE mic location negatively affected SNR by 2 dB when speech was presented at 0° as compared to +/− 90°. In other words, the physical SNR at the mic will be lower (i.e. poorer) with noise originating at 90° or 270° as compared to having both speech and noise at 0°. Since 14 of the 16 hearing preservation patients were Nucleus implant recipients, the premise of microphone location affecting the outcomes in the chosen test environment may appear flawed given that directional microphones have been used in Cochlear processors since the introduction of the Freedom. In a description of the polar patterns of the Freedom processor, Patrick et al. (2006) showed that with speech at 0° and noise at 90° that the signal at 90o was still approximately 2 dB higher than the signal at 0° azimuth. Thus the experimental setup had the potential to artificially inflate estimates of HS and negatively impact estimates of squelch for all subject groups in the current study, as well as all previous studies for which noise was presented directly to the side of the listener. In a more diffuse noise condition—as experienced in everyday listening environments—it may be the case that hearing preservation patients would benefit from binaural squelch.

Finally, another possible reason for the lack of binaural effects for the hearing preservation patients was the heterogeneity of electrode arrays and range of hearing thresholds in the implanted ear. As mentioned in the Subjects section, not all of the implanted electrodes for the hearing preservation were specifically designed for hearing preservation purposes. Neither the Nucleus perimodiolar arrays nor the Sonata H (i.e. standard) electrodes were designed with hearing preservation in mind despite having met inclusion criteria for the current study. Further investigation will continue to determine whether the purportedly atraumatic electrodes yield binaural benefit, namely summation and squelch, for implant recipients with preserved hearing.

CONCLUSION

The hypothesis for the current study was that preserved acoustic hearing in the implanted ear—affording bilateral acoustic hearing—would yield greater access to binaural cues for the hearing preservation participants. The current findings could not reject the null hypothesis. Although the current study did not provide evidence for the availability of binaural cues to patients with hearing preservation, the findings may have been confounded by the test conditions. Specifically, the spatially separated noise conditions directed noise toward the microphone port of the behind-the-ear hearing aid and implant processor which negatively affects the SNR as compared to the S₀N₀ condition [Festen and Plomp, 1986]. Given that 12 of the 16 hearing preservation listeners exhibited some binaural summation, it is possible that greater access to binaural cues may be present in more realistic listening environments.

FIGURE 8.

Box and whisker plots displaying estimates of spatial release from masking (SRM) for BKB-SIN (in dB, panel A) and AzBio at +5 dB SNR (in percentage points, panel B) for the best-aided condition.