Cochlear implantation in adults with acquired single-sided deafness improves cortical processing and comprehension of speech presented to the non-implanted ears: a longitudinal EEG study

Abstract

Former studies have established that individuals with a cochlear implant (CI) for treating single-sided deafness experience improved speech processing after implantation. However, it is not clear how each ear contributes separately to improve speech perception over time at the behavioural and neural level. In this longitudinal EEG study with four different time points, we measured neural activity in response to various temporally and spectrally degraded spoken words presented monaurally to the CI and non-CI ears (5 left and 5 right ears) in 10 single-sided CI users and 10 age- and sex-matched individuals with normal hearing. Subjective comprehension ratings for each word were also recorded. Data from single-sided CI participants were collected pre-CI implantation, and at 3, 6 and 12 months after implantation. We conducted a time-resolved representational similarity analysis on the EEG data to quantify whether and how neural patterns became more similar to those of normal hearing individuals. At 6 months after implantation, the speech comprehension ratings for the degraded words improved in both ears. Notably, the improvement was more pronounced for the non-CI ears than the CI ears. Furthermore, the enhancement in the non-CI ears was paralleled by increased similarity to neural representational patterns of the normal hearing control group. The maximum of this effect coincided with peak decoding accuracy for spoken-word comprehension (600-1200 ms after stimulus onset). The present data demonstrate that cortical processing gradually normalizes within months after CI implantation for speech presented to the non-CI ear. CI enables the deaf ear to provide afferent input, which, according to our results, complements the input of the non-CI ear, gradually improving its function. These novel findings underscore the feasibility of tracking neural recovery after auditory input restoration using advanced multivariate analysis methods, such as representational similarity analysis.

Keywords: EEG; cochlear implant; representational similarity analysis; single-sided deafness; speech comprehension.

Graphical Abstract

Figure 1

Time-resolved multivariate analysis on EEG data. (A) General overview of the RSA. First, we extracted condition-specific EEG sensor values for every time point in every epoch (trial) and formed them into response vectors. Then, using a leave-one-out cross-validation scheme, we trained and tested a SVM to classify speech comprehension from the response vectors. The decoding result (pairwise decoding accuracy, e.g. the very easy condition versus the very difficult condition, 50% chance level) for every time point was aggregated into a RDM of size 4 × 4. Here, decoding accuracy is used as a dissimilarity measure. The matrix is symmetric along the diagonal, which is undefined. Averaging the lower triangular part of the matrix (accuracies within the marked area) resulted in grand average decoding accuracy as an index of how well neural representations distinguish speech comprehension at each time point. Grand average decoding accuracies for each time point were formed into the decoding accuracy curve which depicted how the brain discriminates speech comprehension over time. (B) General overview of the shared representation analysis. We employed RDMs to relate speech degradation representations among CI users and normal hearing controls. We computed Spearman’s R to correlate RDMs between CI ears and normal hearing controls as well as between non-CI ears and normal hearing controls for all time point combinations (t_x, t_y).

Figure 2

Results of comprehension ratings. (A, B) The subjective comprehension rating increased over time not only in the CI ear but also in the non-CI ear. The asterisks on the bottom of the bars in the non-CI ear and CI ear denote whether there is a significant difference between those and normal hearing controls. (A, C, D) Significant differences between all pairs of conditions were found in the normal hearing controls and the non-CI ear (N = 10 for each group; all P < 0.001, CLMMs, pairwise comparisons corrected by Tukey test). Significance levels: *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 3

Average decoding accuracy over time for the non-CI ear (left panel) and the CI ear (right panel) for all conditions. Results from post-operative non-CI ears along with normal hearing controls showed greater-than-chance decoding accuracy (left panel). On the other hand, CI ears did not show greater-than-chance decoding accuracy in any session (right panel). Rows of dots indicate time points with significantly above-chance decoding accuracy (N = 10 for each group; right-tailed sign permutation test, corrected significance level P < 0.05).

Figure 4

Quantification of shared degraded speech representation between CI users and normal hearing controls. (A) Average time-generalized Spearman’s R matrices relating CI ears, non-CI ears and normal hearing controls for each session. Significant correlations are highlighted with black lines (N = 10 for each group, right-tailed sign permutation tests, cluster-corrected significance level P < 0.05). For the result of relating CI users’ CI ear and non-CI ear, see Supplementary Fig. S5. (B) Average all of the Fisher Z-scores of correlation coefficients within 600 and 1200 ms (white dashed squares in Fig. 4A) of each session and compared them across sessions. The result parallels the findings of decoding accuracy. It shows that the correlation of non-CI ears with normal hearing controls improves at 6- and 12-month post-operation (left panel). On the other hand, CI ears did not show improvement over time (right panel). Bars represent 95% confidence intervals. Significance levels: *P_Bonferroni < α (see ‘Statistical analysis’; N = 10 for each group, permutation paired t-tests with Bonferroni correction).