Reducing Channel Interaction Through Cochlear Implant Programming May Improve Speech Perception: Current Focusing and Channel Deactivation

Abstract

Speech perception among cochlear implant (CI) listeners is highly variable. High degrees of channel interaction are associated with poorer speech understanding. Two methods for reducing channel interaction, focusing electrical fields, and deactivating subsets of channels were assessed by the change in vowel and consonant identification scores with different program settings. The main hypotheses were that (a) focused stimulation will improve phoneme recognition and (b) speech perception will improve when channels with high thresholds are deactivated. To select high-threshold channels for deactivation, subjects' threshold profiles were processed to enhance the peaks and troughs, and then an exclusion or inclusion criterion based on the mean and standard deviation was used. Low-threshold channels were selected manually and matched in number and apex-to-base distribution. Nine ears in eight adult CI listeners with Advanced Bionics HiRes90k devices were tested with six experimental programs. Two, all-channel programs, (a) 14-channel partial tripolar (pTP) and (b) 14-channel monopolar (MP), and four variable-channel programs, derived from these two base programs, (c) pTP with high- and (d) low-threshold channels deactivated, and (e) MP with high- and (f) low-threshold channels deactivated, were created. Across subjects, performance was similar with pTP and MP programs. However, poorer performing subjects (scoring < 62% correct on vowel identification) tended to perform better with the all-channel pTP than with the MP program (1 > 2). These same subjects showed slightly more benefit with the reduced channel MP programs (5 and 6). Subjective ratings were consistent with performance. These finding suggest that reducing channel interaction may benefit poorer performing CI listeners.

Keywords: channel selection; cochlear implant; electrode configuration; phoneme perception; speech perception.

© The Author(s) 2016.

Figures

Figure 1.

Each panel shows the pTP detection threshold (dB re 1 mA) as a function of active electrode number (from apical to basal) for each subject. Subjects are organized by performance on medial vowel identification using their everyday listening program (indicated in the bottom left of each panel). Filled red and gray circles indicate channels selected for deactivation from high- and low-off programs, respectively. The solid black and blue dashed lines represent the mean and standard deviation of thresholds for each subject, respectively.

Figure 2.

The bars represent performance (converted from percent correct) for medial consonant (top), medial vowel (middle), and the average of consonants and vowels (bottom). The color of the bars represent the electrode configuration used in these all channel programs. The lighter color bars indicated that the testing was performed in the presence of four-talker babble noise at a signal-to-noise ratio threshold of + 10 dB. Error bars on the averaged data to the right in each panel represent 1 standard deviation of the mean.

Figure 3.

Bars represent the subjective quality (top) and clarity (bottom) ratings on a scale from 1 to 10 for subjects listening to either pTP (red) or MP (blue) experimental programs. Conventions as in previous figure.

Figure 4.

Circles represent data for each subject for the difference in performance between the all channel pTP and MP programs (y-axis) and with the subjects’ everyday listening program (x-axis) for medial consonants (left), vowels (middle), and the average of consonants and vowels (right). The stars represent the average data for poorer performers (filled) and better performers (open). Performance was categorized based on medial vowel performance, indicated by the vertical dashed line at 62%.

Figure 5.

Bars indicate the difference in performance between the pTP all channel program and the programs with either high (dark pink in quiet and gray in noise) or low (light pink in quiet and white in noise) channels deactivated. Upward going bars indicate that performance was better with channels deactivated.

Figure 6.

Conventions as in Figure 5 except the data were obtained with the MP configuration. Color indicates programs with either high (dark blue in quiet and gray in noise) or low (light blue in quiet and white in noise) channels deactivated. Note that because of time constraints S30 did not participate in the MP channel deactivation portion of the experiment and therefore the data are missing from the figure.

Figure 7.

Conventions as in Figure 4. The top and bottom rows of panels represent the difference between the pTP and MP all channel programs and the average of the high- and low-channels deactivated programs (y-axis). Positive numbers indicate the subject performed better with channels deactivated compared with using all channels.

Figure 8.

Schematic showing electrical fields generated by the Goldwyn et al. (2010) cylinder model for MP (top) and pTP (bottom). Rectangles represent electrode contacts and pink ovals represent spiral ganglion neurons. The left panels show the full complement of active channels. The right panels show the distribution of voltage that would occur if the electrodes with x’s were not activated in a program.