Comparing Fixed and Individualized Channel Interaction Coefficients for Speech Perception With Dynamic Focusing Cochlear Implant Strategies

Abstract

Dynamic focusing cochlear implant strategies aim to emulate normal cochlear excitation patterns by varying the degree of current focusing as a function of input level. Results on the speech perception benefits of these strategies have been mixed. In previous studies, channel interaction coefficients (K), which mediate the relationship between current level and degree of focusing, were fixed across channels and participants. Fixing K without accounting for channel interaction and the current required to accurately stimulate target neurons may elicit suboptimal loudness growth and speech perception. This study tested whether individualizing K improved speech perception relative to fixed-K and monopolar strategies. Fourteen ears of implanted adults were programmed with 14-channel strategies matched on pulse duration, pulse rate, filtering, and loudness. Sentence recognition and vowel identification was measured at 60 dB SPL equivalent in quiet and four-talker babble. On the group level, speech recognition in quiet and noise was similar between strategies. On the individual level, there were participants who benefitted with dynamic focusing strategies for speech perception in noise. Patterns of benefit were generally unclear, beyond associations between focused thresholds, duration of hearing loss, and individual-K benefit. Participants rated dynamic focusing like monopolar in clarity and ease of listening. Almost all participants expressed their willingness to use the strategies in a take-home trial. These results suggest that while individualizing K does not benefit all, there are individuals who benefit, for which the electrode-neuron interface may play a role. Future studies will evaluate acclimatization of dynamic focusing strategies using take-home trials.

Keywords: cochlear implant; current focusing; stimulation strategy; vowel identification.

Figure 1.

The effect of channel interaction (K) and focusing (σ) coefficients on current levels required for neural stimulation. Electrodes are shown as rounded rectangles, and neurons as circles. The edge of the osseous spiral lamina is indicated by the dashed line, and the lateral wall by the solid line. The spatial extent of current required to activate neurons are indicated by the shaded areas. Active electrodes and excited neurons are filled. The active electrodes are shown with dotted arrows indicating the direction of current flow toward the ground electrodes. Schematic A (top panel) shows a short electrode-to-neuron distance (d) where K  =  0.1. The current level is relatively constant and increases only very slightly with an increase in focusing, as indicated by the similar, light shading for stimulation at σ  =  0.5 (left) and σ  =  0.8 (right). Schematic B (bottom panel) shows a large electrode-to-neuron distance where K  =  0.9. Here, the current levels are greater than for K  =  0.1, represented by the darker shading. The current level is highly dependent on changes in focusing, and increases a large amount with an increase in focusing, as indicated by the much darker shading for σ  =  0.8 (right) compared to σ  =  0.5 (left).

Figure 2.

Line plots of the electrical dynamic range display the mean detection thresholds and most-comfortable loudness levels for two electrode configurations: monopolar (left panel) and steered quadrupolar with σ  =  0.8 (right panel). The mean detection thresholds are represented by upward pointing triangles on thick black lines, while the most-comfortable loudness levels are represented by downward pointing triangles on thick black lines. Error bars are ± 1 standard deviation. Individual data are shown as thin gray lines.

Figure 3.

Bar plots showing speech recognition in quiet (RAU) with the monopolar (red), fixed-K (blue) and individual-K (green) strategies. Bars are ordered according to keyword recognition in quiet performance with the monopolar strategy. Error bars on the average (end; “AVG”) bars denote 95% confidence intervals of the means. RAU=rationalized arcsine unit.

Figure 4.

Bar plots showing speech recognition in noise (RAU) with the monopolar (red), fixed-K (blue) and individual-K (green) strategies. Bars are ordered according to keyword recognition in quiet performance with the monopolar strategy. Error bars on the average (end; “AVG”) bars denote 95% confidence intervals of the means. RAU=rationalized arcsine unit.

Figure 5.

Scatterplots comparing monopolar, fixed-K and individual-K performance for keyword recognition in noise (circle markers) and vowel identification in noise (diamond markers). Filled markers denote performance greater than a critical difference for that particular comparison of strategies. Markers filled red denote better performance with monopolar, markers filled blue denote better performance with fixed-K, while markers filled green denote better performance with individual-K.

Figure 6.

The distribution of sound quality ratings for the monopolar (red), fixed-K (blue) and individual-K (green) strategies on clarity and ease of listening in quiet and noise, compared to everyday strategies. Ratings of “much less” to “slightly less” clear or easy are grouped as “less clear” or "less easy", while “slightly” to “much” clearer or easier ratings are grouped as “clearer” or "easier".