Cochlear implants (CI) utilize default frequency allocation tables (“pitch maps”) to distribute the frequency range important for speech perception across their electrode array. Default pitch maps do not address the significant pitch-place mismatch that is inherent in cochlear implantation, nor the variability between subjects or array lengths. Recent research has utilized postoperative high-resolution flat-panel computed tomography (CT) imaging to measure the precise location of electrode contacts within an individual's cochlea, to generate a custom pitch map and decrease the pitch-place mismatch. The objective was to determine whether chronic use of CT pitch maps would improve CI user performance in the areas of speech and music perception, as compared to the non-custom default pitch maps provided by the CI manufacturer. A cohort of 10 experienced CI users (14 CI ears) were recruited to receive CT scans and then use a custom CT pitch map for 1 month. The efficacy of these maps was measured using a battery of speech and music tests. No change was found at the group level; however, large inter-subject variability of the benefit of the CT maps was correlated to CI electrode array placement. This application of a custom, strict CT mapping is not beneficial for all CI users. Results may be limited by long acclimation periods to default pitch maps before CT map usage.

Subjects

Study subject demographics, describe the 10 MED-EL cochlear implant (CI) participants (14 CI ears). Four men and six women participated, with an average age of 64.3 ± 12.2 years. All participants were adults with at least 6 months of CI listening experience, implanted with a MED-EL CI, postlingually deafened, native speakers of American English, and used oral/aural communication as their primary modality of communication. The mean length of CI use was 3.6 ± 1.8 years and the mean duration of severe-profound hearing loss before implantation was 6.9 ± 9.6 years. The CI users utilized a variety of sound processor models and processing strategies. Subjects were excluded if they had intracochlear electrodes with open or short circuits, 4 or more extracochlear electrodes (≥1/3 of the array), or documented concomitant conditions that may affect performance (e.g., cognitive impairment). The Institutional Review Board at the University of California, San Francisco (UCSF) approved this study and informed consent was obtained from all participants.

Study Design

Experienced CI recipients were asked to use a CT experimental pitch map, full-time, for one month. Subjects completed a battery of speech and music tasks at the beginning and end of month on both their clinical (Clin) and experimental (Exp) pitch maps. CI users entered the study with chronic exposure (6+ months) to their clinical pitch maps. Our primary hypothesis, namely that CT mapping would improve speech and music perception, was thus best revealed by comparing performance following chronic use of each map, specifically the clinical map at the first test session and the experimental CT map at the second test session.

CT Experimental Map

The CT map experimental was intended to create a strict tonotopic pitch map while utilizing the full 70-8500Hz bandwidth available in the CI software. To this end, the characteristic frequency (CF) of each electrode was calculated and, in the CI programming software, the center of each channel band was matched to a corresponding electrode CF. This approach created a frequency allocation table that minimized pitch-place mismatch, such that frequencies of incoming sounds were directed to the most anatomically correct location in the cochlea. If there were electrodes whose CFs fell too far outside the programmable frequency range (>4 semitones from a channel center), then they were deactivated. The only fitting aspect that was explicitly changed for the CT maps was the frequency allocation table. Other fitting parameters, such as upper and lower electrical stimulation levels, for example, were held constant between the clinical and experimental maps.

Test Conditions

All study participants used their sound processors for the chronic trial period(s) and any corresponding speech and music testing. Speech and music stimuli were presented at 65 dBA and the non-test ear was masked. All CI users were asked if they could hear the test stimuli with the non-test ear while the CI on their test ear was turned off; all subjects confirmed being unable to hear the test stimuli with the non-test ear.

During speech and music testing, any additional sound processing (i.e., directional microphones, steady-state noise reduction algorithms) were deactivated, and the volume and sensitivity settings were fixed to 100% and 75%, respectively. Upper stimulation levels (MCLs) for bilateral users were globally adjusted as needed to ensure a comfortable listening level in both ears; any global MCL adjustments needed for testing were replicated in both control and experimental fittings.

Participants completed the experiment using a tablet computer (Microsoft Surface Pro 3) and were given the option of using a touchscreen or mouse. The computer ran the speech and music stimuli with the following software programs: Windows Media Player (v12, 2009), MATLAB (vR2012b, Mathworks, Natick, MA, USA), and LabView (v11.0, National Instruments, Austin, TX, USA).

Test Metrics

CNC Words and Phonemes

The Consonant-Nucleus-Consonant (CNC) Monosyllabic Word Test (Peterson and Lehiste, 1962),used to test recognition of open-set monosyllabic words in quiet, was sourced from the Minimum Speech Test Battery for Adult CI Users (Luxford et al., 2001; MTSB, 2011). 

Sentences in Noise

The Bamford-Kowal-Bamford Speech-in-Noise (BKB-SIN) Test (Bench, Kowal and Bamford, 1979; Etymōtic Research, 2005; Luxford et al., 2001) employs a modified adaptive approach wherein sentences are presented at a fixed level and four-talker babble (Auditec of St. Louis, 1971) is presented at increasingly more difficult signal-to-noise ratios (SNRs). An SNR-50 score was calculated for each sentence list-pair, which represents the level of SNR (in dB) at which the subject correctly recognized exactly half of the keywords throughout. 

Synthetic Vowels

The Iowa Medial Vowel Test (Tyler et al.,1986) utilizes eight synthetic vowel stimuli presented in an ‘‘/h/vowel/d/’’ context (e.g., had, hid, heed, etc.). After familiarization with a practice module with the test interface and all eight stimuli, the subject then completed a closed-set test with five blocks of eight tokens each, where the token stimulus order was randomized. The test was administered twice, yielding 80 total trials. 

Timbre Discrimination

Musical timbre is conveyed by spectrotemporal features, beyond fundamental frequency and amplitude, that give a particular sound its characteristic tone color. Timbre perception is typically examined using musical instrument identification tasks that generally reveal poor performance in cochlear implant (CI) users (Kang et al., 2009; Gfeller et al., 2002; Spitzer et al., 2008). Relatively little is known, however, regarding timbre discrimination—the ability to differentiate between two notes of different timbres—a task which may provide greater insight into CI-mediated listening than whether or not instruments can be correctly identified.

A task developed by the present investigators was used to measure timbre discrimination through acoustic blends of source instruments (clarinet, flute, French horn, tuba, violin, and cello) which were combined in varying amplitude ratios while maintaining constant overall volume (Gilbert et al., 2019a). Higher performance on the timbre discrimination task indicates better ability to distinguish between musical instrument stimuli using timbre-related cues only. 

Pitch Ranking

Our lab designed a test paradigm to characterize pitch height (for pure tones) across the entire stimulable frequency range of the MED-EL CI (70-8500Hz) (Jiam et al., 2019). Better performance on this task is indicated by a lower number of errors. During the analysis of this test, we dropped/didn’t count the errors/responses that corresponded to areas of the frequency range where the Clinical and CT Maps were overlapping. Overlapping was defined as the upper ends of the channels and the lower ends of the channels were each within 3 ST of each other. 

Statistical Analysis

All statistical analyses were completed in JASP software (JASP, Version 0.18.3; Amsterdam, Netherlands) using p<0.05 to determine statistical significance. The primary comparison of interest was between performance with the incoming Clinical Map and following one month of chronic use of the Experimental Map. Shapiro-Wilk tests were used to test normal distribution of the data. All the performance test results were normally distributed, and so paired sample student t-tests assessed whether mean performance on the clinical and experimental maps were different. Linear correlations between test results and subject- or ear-specific factors (i.e., CI array alignment, demographic data, and hearing history) were explored with either Pearson’s r (for parametric) or Spearman’s rho (for non-parametric) tests.