INTRODUCTION
Cochlear implants (CIs) have successfully restored hearing to more than 300,000 patients worldwide. Although many listeners achieve excellent open-set speech perception, even in the absence of visual information, performance varies widely across listeners. The primary source of this variability remains a matter of debate, but one hypothesis is that the within-listener variability in perception for stimulation from individual CI channels reflects an underlying pathology that impairs speech perception.
Psychophysical measures obtained using focused electrical stimulation vary substantially across CI listeners and across electrodes within listeners. Previous studies have shown that channels with detection thresholds higher than the average for that listener exhibit poorer spectral/spatial resolution (Bierer and Faulkner
2010; Long et al.
2014) and smaller dynamic ranges (Bierer and Nye
2014). Further, high variability in threshold from channel to channel within individuals measured with monopolar (MP) (Pfingst and Xu
2004) and/or with focused stimulation (Pfingst and Xu
2004; Bierer
2007; Long et al.
2014) is associated with poor speech perception scores. Those findings suggest that speech perception may be degraded by the poor transmission of information by a subset of electrodes and that detection thresholds may provide some indication of the presence and identity of those channels. However, it is possible that more than one factor influences the detection threshold for stimulation of a given electrode and that these factors differ in the extent to which they influence the transmission of speech information. Specifically, detection thresholds may be affected by both degeneration of neurons close to the stimulating electrode and/or by the distance of that electrode from the modiolus (Bierer
2007; Long et al.
2014). Although both of these factors would lead to broader excitation patterns, it is possible that this could be alleviated by the use of more focused stimulation in the case of large modiolar distance, but not when local neural survival is poor.
In everyday use, loudness is, to a first approximation, equated across electrodes in a patient’s clinical map by the assignment of different thresholds (T) and most comfortable levels (MCLs). However, substantial across-electrode differences in suprathreshold tasks may remain (Kong et al.
2009; Pfingst
2011; Garadat et al.
2012). The present study investigates the across-electrode variation in temporal information conveyed by CI electrodes. We measured gap-detection thresholds (GDTs), which we believe represent a relatively pure measure of temporal sensitivity. GDTs have been modeled using the concept of a sliding temporal window followed by the detection of changes in the level of the output of that window (Plack and Moore
1990). In normal hearing, the equivalent rectangular duration of that window is less than 10 ms, and a similar value has been used to model data from CI users, albeit with a different experimental paradigm (McKay and McDermott
1998). Detection of such short gaps will be affected by both differences in the duration of the temporal window and/or by differences in the fidelity of the auditory nerve response that, presumably, forms the input of that window. Another task that has been proposed as a measure of temporal sensitivity is the detection of amplitude modulation (modulation detection thresholds (MDTs)). Those tasks typically use low modulation rates, for example, 10 Hz, for which the corresponding period is much longer than 10 ms (Garadat et al.
2012). We believe that those tasks may be more sensitive to the encoding of amplitude differences than to differences in the duration of the temporal window, and, as the modulation period gets longer, the task becomes more and more like one of level discrimination. Yet another task, rate discrimination, does involve the processing of short temporal intervals, but may additionally involve pitch processing.
We hypothesize that gap detection might be impaired in some channels by the altered temporal discharge patterns associated with neural degeneration. Shepherd and Javel (
1997) reported that the deafened auditory nerve could produce a “bursting” response to high-rate electrical stimulation and showed that this occurred for cats that had been deaf for 2 months, but not in a cat that was deafened immediately prior to testing. If elevated detection thresholds, particularly in focused stimulation mode, are largely indicative of poor neural survival, and if the effects of that poor survival are similar for stimulus detection and gap detection, then we would expect performance on the two tasks to correlate across electrodes. On the other hand, it is possible that no such correlation will be found, as has previously been reported for MDTs (Pfingst
2011). This could occur if elevated detection thresholds were primarily due to larger electrode-modiolar distances rather than poor local neuron survival. Alternatively, no correlation would be observed if, for example, some electrodes excited regions where AN fibers were sparse but responded with good temporal fidelity, whereas others excited a region where fibers were more numerous but that produced a temporally degraded response.
We measured both detection thresholds and gap-detection thresholds (GDTs) using MP and a focused partial-tripolar (pTP) mode of stimulation, across a subset of electrodes in ten implanted ears from nine CI users. Our results show that both measures could vary substantially across electrodes within a given subject and that, for a given measure, there was a highly significant correlation between the two modes of stimulation. However, there was no correlation between the two tasks, suggesting that GDTs are influenced by factors separate from, or additional to, those reflected in detection thresholds.
A second goal was to evaluate across-electrode variation in MP mode. Although large across-electrode variations in detection threshold have been obtained with focused stimulation, smaller variations are typically obtained in MP mode (Pfingst and Xu
2004; Bierer
2007; Long et al.
2014). However, substantial across-electrode differences have been observed in MP mode with suprathreshold tasks such as rate discrimination (Kong et al.
2009) and modulation detection (Pfingst
2011; Garadat et al.
2012). Indeed, Garadat and Pfingst (
2011) have reported across-electrode variations in MP GDTs, when each electrode was stimulated at the same percentage of its dynamic range, although differences were greatly reduced when electrodes were compared at the same loudness. To minimize such loudness effects, we measured GDTs for stimuli presented at the same loudness, namely the MCL. We compared the amount of across-electrode variation in MP and focused (pTP) stimulation. To do so, we computed a metric that controlled for any possible differences in the absolute size of the GDTs between the two modes. The metric was defined as the ratio between two values: the
between-
electrode standard deviation, calculated from the standard deviation of the mean GDTs from the four adaptive runs used for each electrode, and the
within-
electrode standard deviation, obtained by calculating the standard deviation across adaptive runs for each electrode separately and then averaging these standard deviations across electrodes. This dimensionless metric, termed the standard deviation ratio (“SDR”), could also be modified to describe the across-electrode variation in stimulus-detection thresholds, or to compare the across-electrode variation between different tasks, such as detecting a pulse train and detecting a gap. In addition, it provides an arguably more valid method than the simple across-electrode standard deviation, when comparing the across-electrode variation in detection thresholds between the two modes. This is because the same across-electrode standard deviation may not correspond to the same amount of variation in sensitivity in the two modes, if the slopes of the underlying psychometric functions are different. A difference in slope should, however, affect both the within- and between-electrode standard deviation, and so the ratio between these two values may be more appropriate than the across-electrode standard deviation alone.
METHODS
Subjects
Nine postlingually deafened adults wearing the Advanced Bionics HiRes 90K CI participated; their details are shown in Table
1. One subject, who was bilaterally implanted, was tested in each ear and is listed as S30L and S39R in the table. This subject’s two ears were treated as completely separate, and therefore, for the purposes of analysis and for discussion in the remainder of this article, there were ten “subjects.” Five of the subjects were implanted and tested in Cambridge, England, UK, whereas the other half was implanted and tested in Seattle, WA, USA. All procedures were approved by the respective Human Subjects Review Boards.
TABLE 1
Details of the subjects who took part in the experiments. Subject codes starting with the letter “S” refer to subjects implanted and tested in Seattle (WA, USA), whereas those starting with “C” were implanted and tested in Cambridge, England.
S22 | 72 | 55 | 4 years 9 months | Hereditary |
S27 | 83 | 55–60 | 4 years 7 months | Unknown |
S28 | 74 | 26 | 4 years 5 months | Hereditary |
S30L | 49 | 16 | 9 years | Hereditary |
S39R | 49 | 16 | 19 years (1) 8 years (2) 2 years (3) | Hereditary |
C1 | 67 | 32 | 3 years | Unknown |
C2 | 31 | 7 | 2 years | Unknown |
C3 | 69 | 50 | 2 years | Otosclerosis |
C4 | 66 | 37 | 4 years | Otosclerosis |
C5 | 53 | Prog since 1992 | 4 years | Unknown |
Stimuli
Biphasic, charge-balanced, cathodic-phase-first pulses were used. Phase durations were either 97 or 194 μs, and the pulse rate was 1031 pps. The longer phase duration was used when MCL could not be reached within the compliance limits for the shorter phase duration. MCLs were obtained with individual pulse train presentations. Following each presentation, the subject indicated the loudness of the train using the Advanced Bionics loudness rating scale for which a “6” is “most comfortable” and a “7” is “loud but comfortable.” The level was increased in 0.5 or 0.1 dB steps until the rating was “7” and then reduced until the listener reported a “6” again. MCLs were not loudness balanced. Pulse train durations were 200 ms for signal detection and 400 ms for gap detection. Stimuli were either presented in the MP or the pTP configuration with a return current fraction (σ) of 0.75. All stimuli were presented and controlled using research hardware and software (“BEDCS”) provided by the Advanced Bionics company. Programs were written using the MATLAB programming environment, which controlled low-level BEDCS routines. Stimuli were checked using a test implant and digital storage oscilloscope. The same software and identical hardware were used in both testing sites (Cambridge and Seattle).
Signal detection thresholds were measured using a three-down, one-up, two-interval forced-choice adaptive procedure that converged on 79 % correct. Step size was 1 dB for the first two turnpoints and 0.25 dB thereafter. The mean of the last four of six reversals was used to estimate threshold. Four or five repetitions were performed for each measurement. Thresholds were measured in pTP mode for all available channels (usually 2 through 15). Subjects were asked “Which interval contained the sound?” and responded using a computer mouse. Note that subject C3 was not tested on electrodes 7, 8, and 9 because those channels were deactivated from that subject’s everyday program. The same is true for subject C5 for electrode 15. Because of time constraints and health issues with subject S27, only a subset of electrodes was tested. We had previously tested all 14 electrodes for S27 but with a different pTP fraction of 0.9, so the highest and lowest threshold channels from these earlier measures were purposely included.
For each subject, four or five electrodes were selected for further testing, such that at least two had low pTP thresholds and at least two had high pTP thresholds. The general rule was to select the two highest and two lowest thresholds, unless this involved two adjacent “high” or “low” electrodes. Detection thresholds for these additional electrodes were obtained in MP mode, using the same method as described above for pTP stimulation. GDTs were obtained in both modes using an adaptive procedure similar to that used for the threshold measurements. Each run started with an easily discriminable gap size that was then adjusted in steps of 40 and 10 % of the gap durations for the first two and last four turnpoints, respectively. The threshold for each run was determined from the arithmetic mean of the last four turnpoints. Four or five runs were performed for each measurement, and the threshold for that measurement was calculated from the arithmetic mean of those runs. Subjects were asked to answer, “Which interval contained the gapped sound?” For the gap-detection task, the pulse train duration was roved by +/− 10 %, in order to reduce the usefulness of duration as a cue. The stimulus level was not roved. For both the GDT and detection threshold measures, correct-answer feedback was provided after every trial.
Loudness Balancing
Gap stimuli were presented at each individual’s MCL with both the MP and pTP electrode configurations, when possible. In one case, the most comfortable level could not be reached because of the compliance limits of the system, and so a “soft” level was used in both modes and loudness balancing of stimuli across channels was performed. For this one subject (C2), the procedure was first to determine the highest level reachable within the compliance limits and comfort levels for all test electrodes. The lowest of those current levels was then used for loudness balancing. The subject set the level of all test electrodes to be at the same subjective level on the loudness rating scale. Then one of the low-threshold electrodes was selected to be the reference, and all of the other electrodes were adjusted to match the perceived loudness of the reference. The subject had manual control over the level of the second sound with either 0.25 (“+” or “−”), 0.5 (“++” or “− −”), or 1 dB (“+++” or “− − −”) step sizes. When the listener believed the two sounds to be equally loud, they clicked on a “Done” button. For half the matches, the starting level was well below the loudness-matched level predicted from loudness ratings; for the other half, the starting level was above this estimated value. Following four repetitions of loudness balancing for each test electrode, the reference electrode became the test electrode. The procedure was repeated four times for each reference electrode.