Cortical encoding of pitch contour changes in cochlear implant users: a mismatch negativity study

Abstract

A better understanding of melodic pitch perception in cochlear implants (CIs) may guide signal processing and/or rehabilitation techniques to improve music perception and appreciation in CI patients. In this study, the mismatch negativity (MMN) in response to infrequent changes in 5-tone pitch contours was obtained in CI users and normal-hearing (NH) listeners. Melodic contour identification (MCI) was also measured. Results showed that MCI performance was poorer in CI than in NH subjects; the MMNs were missing in all CI subjects for the 1-semitone contours. The MMNs with the 5-semitone contours were observed in a smaller proportion of CI than NH subjects. Results suggest that encoding of pitch contour changes in CI users appears to be degraded, most likely due to the limited pitch cues provided by the CI and deafness-related compromise of brain substrates.

Figure 1

The five-note pitch contours used for MCI test. The frequency interval between successive notes in the pitch contours was 1 or 5 semitones. The shaded notes represent the root note in each contour.

Figure 2

The five-note pitch contours used for EEG recording. The frequency interval between successive notes in the pitch contours was 1 or 5 semitones.

Figure 3

Behavioral MCI performance for NH and CI subjects. The error bar indicates one standard error.

Figure 4

Grand average LAEPs for good (n=6) and poor NH performers (n=4) whose MMNs were judged to be present. The STANDARD waveforms, the DEVIANT waveforms, and the difference waveforms (DEVIANT–STANDARD) are shown. Time zero represents the onset of the 4^th note. Data from nine electrodes in the fronto-central area were averaged for each waveform. MMN markers including the onset time, offset time, and duration are shown in the top left plot.

Figure 5

Similar plot as Figure 3, but for CI users whose MMNs were judged to be present for 5-semitone spacing (n = 3 for Falling/Falling-flat condition and n = 6 for Rising/Rising-flat condition).

Figure 6

Subplot A shows LAEPs with 1-semitone and 5-semitone spacing in one good NH subject (SnhBB) and the top-performing CI subject (Sci06). Subplot B displays the mean data for NH performers whose MMNs were judged to be present (n=8) for better visual comparison.

Figure 7

The MMN measures (peak amplitude, peak latency, and duration) as the function of behavioral performance in CI users (left subplots) and NH listeners (right subplots) for 5-semitone spacing. The correlation between MMN peak amplitude and percent correct in CI users can be described by a linear regression model with a R² at 0.84 (top left subplot, p<0.05). The line indicates the regression line.