Random Gap Detection Test (RGDT) and RGDT-Expanded Test.

The RGDT [35] is a clinically available test that assesses the auditory gap detection threshold by having the subject identify if one or two signals are heard when signal pairs are separated in time from 0 to 40 ms (0, 2, 5, 10, 15, 20, 25, 30 or 40 ms). The subtest for tonal stimuli includes tones at four frequencies (500, 1000, 2000, and 4000 Hz). Each stimulus in the signal pairs had the same frequency and the same duration (17 ms, including a 1-ms rise-fall time). The order of gap intervals was randomized. Stimulus pairs were separated by a 4.5-second interval to give subjects time to respond. The subject was required to raise 1 or 2 fingers to indicate whether he/she heard 1 or 2 tones. A practice session was presented with tone pairs at 1000 Hz followed by testing with variable gap durations. The lowest gap was detected for each of the frequencies 500–4000 Hz. When a subject failed the RGDT (no threshold was identified using RGDT), the RGDT-Expanded Test was administered in the same manner as the RGDT. The RGDT-Expanded Test is the same as the RGDT except that time intervals include 50, 60, 70, 80, 90, 100, 150, 200, or 300 ms. The stimuli were presented in the sound field through an audiometer using a CD player.

Consonant Nucleus Consonant (CNC) Test.

One list of CNC words were presented in quiet. The subjects were required to repeat the words they heard. The CNC test assesses the perception of isolated monosyllabic words without the syntactic and semantic cues provided in sentence-based tests. This test was used because previous studies have reported a significant correlation between temporal processing ability (gap detection or modulation detection threshold) and phoneme/word recognition [6, 36].

AzBio.

AzBio [37] is a set of sentence lists that has been used to evaluate the speech perception abilities of hearing-impaired listeners and CI users. One list of 20 sentences from the multiple-talker AzBio sentence corpus was presented in quiet. These sentences range in length from 4 to 12 words and the speakers spoke in a casual style. A total of 4 speakers (2 males and 2 females) were presented for each 20-sentence list.

ECAP measurement.

The ECAP was recorded on Electrode 16 within the cochlea using the Neural Response Telemetry (NRT) incorporated into Custom Sound EP software. Although there may be differences in the adaptation across electrodes [38], we primarily selected Electrode 16 for stimulation, because this electrode typically is assigned to a frequency range involving 1000 Hz, which is close to the pulse rate preferred by most CI users [39] and also the same frequency of tone bursts used for LAEP recordings (see LAEP measurement below). If measurements from Electrode 16 were not possible due to electrode dysfunction, the recordings were made from an electrode close to Electrode 16. ECAPs were measured in response to pulses that consisted of cathodic leading, charge-balanced, biphasic current pulses and were presented in monopolar mode. Pulse width was 25 µs/phase and the interphase gap was 7 µs.

Before recording these potentials, the dynamic range (the range of the intensity unit of electrical pulse between threshold and upper limit of comfortable loudness) at the recording electrode was checked in each subject based on their clinical mapping. Then 1000 pulse per second (pps) pulse trains were presented at different intensity levels around the most comfortable listening level and the subject was asked to estimate the loudness of the stimuli. An intensity corresponding to loudness level 7 on a 0-10-point (inaudible to too loud) numerical scale, which has been proven to be the most comfortable listening level [40,41], was used for ECAP recordings.

This project examined ECAP adaptation evoked by a 50-ms pulse train presented at a rate of 1000 pps. Although NRT incorporates an online subtraction procedure that takes advantage of neural refractoriness to remove stimulus artifact for single-pulse ECAP [25,42], it does not allow for measuring ECAP to individual pulses in the pulse train. A previously reported approach [43] was used for this purpose. Figure 1 illustrates the procedure. Specifically, in addition to single pulse recording, the recordings were made when presenting pulse trains with different durations. The NRT uses a forward masking method that consists of recordings under 4 stimulus frames to remove stimulus artifact: probe alone (A), masker-plus-probe (B), masker alone (C), and zero-amplitude pulse (D). From the single pulse recording, the ECAP response derived by using subtraction (A-B+C-D) was used as the response to the 1^st pulse in the train when evaluating adaptation. Also the single pulse recording provides an artifact template that was derived by subtracting the recording frame for the masker alone condition from the recording frame for the masker-plus-probe condition (B-C). In the pulse train recording, the last pulse in the train was treated as the probe and the preceding pulses were treated as the masker pulse train. After the data were collected, the recording from the masker pulse train-alone condition was subtracted from the masker pulse train-plus probe condition (B’-C’) to derive an ECAP to the last pulse in the train along with the stimulus artifact (B’-C’). Then the artifact template from single pulse recording was subtracted from to obtain the ECAP to the last pulse in the train [(B’-C’)-(B-C)]. By changing the duration of the pulse train (thereby the number of pulses), we were able to obtain ECAP evoked by individual pulses in the train (Figure 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. The recording conditions to obtain an ECAP in single pulse recording and in pulse train recording (I) and how the ECAP to an individual pulse in the train was derived (II).

See the text in Methods section for detailed information on how to derive the ECAP to an individual pulse in the train.

https://doi.org/10.1371/journal.pone.0084631.g001

Time constraints prohibited the recording of an ECAP response to each of the pulses in the masker pulse trains. Therefore, 13 different masker pulse-train lengths were chosen to assess the degree of auditory-nerve adaptation. These pulse trains contained 2, 3, 4, 5, 6, 23, 24, 25, 26, 48, 49, 50, and 51 pulses, respectively. In addition, single-pulse ECAP recording was measured 5 times (the first and the last 4 recordings). The average of the 5 recordings was used to provide the ECAP to the first pulse in the train for adaptation measure and the artifact template. Therefore, there were a total of 18 recordings for each subject.

LAEP measurement.

The stimuli were 1 kHz tone bursts (60 ms, 10 ms rise/fall time) generated and presented using STIM² software through the Neuroscan workstation (Compumedics Neuroscan, Inc., Charlotte, NC). The stimulus-train paradigm used in this study was similar to that in previous studies [31, 32, 44, 45]. The tone bursts were presented in 30 trains consisting of 5 tone-bursts each, with inter-stimulus intervals (ISIs) of 0.7 sec and inter-train intervals (ITIs) of 15s. The stimuli were presented at an intensity level corresponding to a loudness level of approximately 7. At least 6 recordings (180 trials for each stimulus in the train) were collected from each subject.

The electroencephalographic (EEG) recordings were performed using a 40-channel system (NuAmps, Compumedics Neuroscan, Inc., Charlotte, NC). Electrode placement was based on the International 10-20 system, using an Electro-Cap placed on the head of the participant, with the contralateral earlobe as the reference. This procedure has been found to reduce the stimulus artifact in some CI users (McNeill et al., 2007). Electroocular activity (EOG) was monitored so that eye movement artifacts could be identified and rejected during the offline analysis. Approximately 1-3 electrodes located directly over or closely surrounding the implant transmission coil were not used. Electrode impedances in the remaining electrodes were kept at or less than 5 kΩ. Continuous EEG recordings were collected from participants using the SCAN software (version 4.3, Compumedics Neuroscan, Inc., Charlotte, NC) with a band-pass filter setting from 0.1 to 100 Hz, and an analog-to-digital-converter (ADC) sampling rate of 1000 Hz.

Behavioral tests.

The GDTs (ms) at 4 different frequencies (0.5, 1, 2, and 4 kHz) were obtained. The composite GDT (CGDT) value was also calculated by averaging the values for the tested frequencies. CNC words and AzBio sentences were evaluated using the percent correct score. The average score of CNC and AzBio outcomes was used for each subject to represent the speech perception performance.

ECAP.

After the ECAP data collection, the data were exported and entered to Sigmaplot software program to do off-line analysis. Further analysis was performed to derive the ECAP for each pulse in the train using the aforementioned pulse-train subtraction technique. The averaged ECAP of the 5 single pulse recordings was used as the response to the 1^st pulse in the train. The negative and positive peaks were selected for each ECAP as illustrated in Figure 1. The amplitude of the ECAP was measured between the negative and positive peaks. The ECAP was compared between the single pulse (1^st) and later stimuli in the train.

LAEP.

EEG data from both CI and NH subjects were processed using SCAN software (Compumedics Neuroscan, Inc., Charlotte, NC). The data were epoched with a 100-ms pre-stimulus time and a 500 ms post-stimulus time, the repeated recordings were combined, and baseline adjustment was performed based on data in the pre-stimulus time of each stimulus in the train. Data was visually checked and the epoches that show excessive noise such as the responses from the muscle artifact were removed.

In order to minimize stimulus artifacts, the epoched data were concatenated and imported to the EEGLAB Toolbox (EEGLAB, San Diego, CA), an online open source toolbox (freely available from http://sccn.ucsd.edu/eeglab), to minimize the stimulus artifact using Independent Component Analysis [46, 47]. The ICA model decomposes the EEG dataset into mutually independent components, including those from artifactual and neurophysiologic sources. Then an iterative process changes the weights and directions of the vectors in a mixing matrix until maximum independence is identified from the higher order statistics. The stimulus artifact components were identified using topographic 2-D maps with the criteria of CI artifact described in a previous study [47]. Simply, the main feature of the artifact components is that the artifact starts slightly after the stimulus onset and ends slightly after the stimulus offset, with a square shape during the stimulus presentation. Another feature of artifact components is that scalp projections of the activity display a centroid on the side of the implanted device. After identifying stimulus artifact components, these components were linearly subtracted from the epoched EEG data. The remaining components were then constructed to form the final epoched EEG recording, which was later filtered (0.1-30 Hz, 12 dB/octave). Finally, averaged responses were derived independently for each tone burst within the train across the total number of train presentations. For each LAEP, the N1 and P2 peaks were labeled and the N1-P2 peak-to-peak amplitude was used for the amplitude of the LAEP. In CI users, the N1 was defined as the most negative peak around 100 ms after stimulus onset and the P2 was the positive peak following the N1. In CI users, the N1 may occur later. Therefore, we looked for the typical morphology of N1-P2 complex before identifying the N1. Because the LAEP amplitude is the largest for electrode Cz and becomes progressively smaller for electrodes distant from Cz, we restricted the later analysis to measures from Cz.

LAEP.

A number of studies have reported adaptation of the LAEP in NH listeners [13, 32, 44, 52-54]. Authors typically have found that the amplitude reduction can be approximately 50% over the first several stimuli before it reaches asymptotic amplitude, which is lower for shorter ISIs than for longer ISIs [31, 55]. LAEP amplitude does not change significantly for ISIs longer than 10 s [55].

Our group was the first to report LAEP adaptation in CI users [56]. We reported that the LAEP adaptive pattern in CI users (mostly stimulated monaurally) was less prominent than in NH listeners (stimulated binaurally) reported in Zhang et al. [57]. The current study first reports a direct comparison between CI and NH subjects with monaural stimulation. The finding of less prominent adaptive pattern in CI users compared to NH listeners in the current study is consistent with that reported in Zhang et al. [56]. One explanation for the reduced amount of adaptation in CI users is that neural degeneration may have altered the adaptation/refractory features of neural responses [58]. This explanation is supported by a significant correlation between AI_LAEP and the age at implantation, which is a factor affecting the degree of neural degeneration. An alternative explanation is related to the reduced amplitude of the LAEP evoked by the 1^st stimulus in the train and the lack of further amplitude decrements in the LAEP evoked by later stimuli. The LAEP evoked by stimuli presented with a long ISI, ie., the 1^st stimulus in a train, is mainly due to the “non-specific” component generated from regions outside of the auditory cortex, while the LAEP evoked by stimuli at a fast and fixed rate is mainly generated from the auditory cortex [28]. These “non-specific” components need a longer time to recover from refractoriness compared to the component in the auditory cortex. The smaller LAEP evoked by the first stimulus in the train in CI users may indicate that the “non-specific” component of the LAEP is more compromised than the component in the auditory cortex due to long-term deafness.

The above speculation that the reduced contribution of the non-specific generator to the LAEP leads to the less prominent adaptation in CI users compared to NH listeners can also be used to explain the less obvious latency reduction in the N1 for the 2^nd stimulus relative to the 1^st stimulus of the train (Figure 3 and 4) in CI users compared to NH listeners. Previous studies reported that N1 is generated from at least two cortical generators that are spatially and temporally distinct [44, 59, 60]. One component has a shorter latency and a shorter refractory period of 1 to 3 s, whereas the other component has a longer latency and a much longer recovery period [28]. These two N1 components can be distinguished on the basis of their distinct refractory properties [28, 61]. The component with a shorter latency can be elicited by stimuli at any rates and the component with a longer latency can be elicited by stimuli at much slower rates [60, 62]. Therefore, the reduction of N1 latency for the 2^nd stimulus relative to the 1^st stimulus in NH listeners is likely to be related to faster recovery of neurons with shorter latencies when neurons with longer latencies are still influenced by refractoriness. In contrast, CI users’ damage of long-latency generator leads to less prominent latency reduction for the 2^nd stimulus relative to the 1^st stimulus in the train.

The combination of the current finding of less adaptation in CI users relative to NH listeners and findings in prior psychophysical and EEG studies provides clearer mechanisms of electric hearing in CI users. Using forward masking paradigm, Chatterjee [63] reported that the psychophysical recovery from forward masking is faster in poorer CI performers than good performers. The recovery functions display a rapid and a slow component and temporal integration influences the slow component more than the rapid component. Their two-exponential model suggests that the shorter the masker duration, the lesser the temporal integration, the less the contribution of the slow component, and the faster the recovery from masker. Therefore, Chatterjee’s study suggested that the temporal integration of CI users is damaged. A previous electrophysiological study in CI users in our lab [56] reported that the recovery function of the LAEP in CI users appeared to be faster in general compared to NH listeners. One possible explanation of this phenomenon is that CI users may have a reduced contribution of the non-specific generator to the LAEP and thus only the fast recovery of the generator in auditory cortex is shown. Taken together, the less prominent adaptive pattern, the faster recovery of LAEP, and the faster psychophysical recovery function in CI users, suggests that the temporal integration ability, which is likely related to the slow recovering neural generators, could be compromised in this population. Much shallower temporal integration functions have been seen in CI users by other researchers [64]. Further research involving both temporal integration measures and EEG source analysis may help confirm our speculation.

Neural adaptation and behavioral measures.

The current study did not find a significant correlation between either ECAP adaptation or LAEP adaptation and gap detection thresholds. Several possible reasons are speculated. First, it is possible that there is an interaction of the effects of stimulus parameters on different measures. Gap-detection thresholds depend on stimulus intensity in both acoustic and electric conditions and the adaptation of both ECAP/LAEP can also be affected by stimulus intensity [57, 65, 66]. Although we presented stimuli at the most comfortable listening level for all tests, the stimulus intensity corresponding to their most comfortable listening levels may be different within the dynamic range for subjects. It is possible that the variability of ECAP/LAEP adaptation and gap detection thresholds resulting from different stimulus intensities led to the lack of correlation. The second possible reason is the low sensitivity of measures used for examining the neural-behavioral correlation. For example, the recovery function of neural responses has been reported to indicate encoding of temporal cues [67]; the behavioral measure of modulation detection threshold has been used to assess temporal processing abilities [10, 68]. It is worthwhile to use other measures of neural responses and temporal processing to reassess the neural-behavioral correlation.

Although our group reported a significant correlation between LAEP adaptation index and speech perception performance in Zhang et al. [56], the current study did not find such correlation. This may be related to the difference in the speech materials used and the subjects recruited in the two studies. Zhang et al. [56] used the average of CNC word recognition in quiet, HINT sentence recognition in quiet, and HINT sentence recognition at 10 dB signal-to-noise ratio (SNR) for speech perception score. The current study used CNC and AzBio. The HINT sentences in quiet and in a background of fixed-level noise have been thought to be relatively easy compared to AzBio [69]. AzBio appears to be a more reasonable test for speech perception than HINT because CI users’ performance in AzBio displays a wide range of distribution while the HINT test, especially in quiet, suffers from ceiling effects. Further research needs to be conducted using other speech tests suitable for CI users to determine if the correlation between LAEP adaptation index and speech perception exists. Moreover, the current study involved more binaural CI users for LAEP test (4/11, see Table 1) compared to Zhang et al. [56] (2/10). Because the two ears of binaural CI users were tested separately in both studies, there was heavier distribution of data results from the same individuals in the current study.

Despite the differences between the current study and Zhang et al. [56], the finding that less prominent LAEP adaptive pattern in CI users compared to NH listeners was consistent in both studies. This finding indicates the positive role of normal LAEP adaptation in speech perception. This finding supports a view in prior studies that proper neural adaptation following stimulus repetition is useful for speech perception by sensitizing the auditory system for novel stimulus detection [70].

We observed a rather high variability among individual CI users for ECAP, LAEP, and GDT data. Part of this variability may be related to CI users’ demographic factors such as the age at implantation, the duration of deafness, and the length of CI use etc. The correlation analysis found that there is a significant correlation between the LAEP adaptation index and the age at implantation if one outlier data point representing response enhancement rather than adaptation is not included for the analysis. This suggests that CI users who have a greater LAEP adaptation index tended to be the ones implanted at a younger age. Implantation at a younger age may prevent degeneration of auditory system and therefore results in a more normal LAEP adaptive pattern. The results indirectly indicate the positive role of LAEP adaptive pattern in CI performance.