Frequency-to-Place Mismatch: Characterizing Variability and the Influence on Speech Perception Outcomes in Cochlear Implant Recipients

Abstract

Objectives: The spatial position of a cochlear implant (CI) electrode array affects the spectral cues provided to the recipient. Differences in cochlear size and array length lead to substantial variability in angular insertion depth (AID) across and within array types. For CI-alone users, the variability in AID results in varying degrees of frequency-to-place mismatch between the default electric frequency filters and cochlear place of stimulation. For electric-acoustic stimulation (EAS) users, default electric frequency filters also vary as a function of residual acoustic hearing in the implanted ear. The present study aimed to (1) investigate variability in AID associated with lateral wall arrays, (2) determine the subsequent frequency-to-place mismatch for CI-alone and EAS users mapped with default frequency filters, and (3) examine the relationship between early speech perception for CI-alone users and two aspects of electrode position: frequency-to-place mismatch and angular separation between neighboring contacts, a metric associated with spectral selectivity at the periphery.

Design: One hundred one adult CI recipients (111 ears) with MED-EL Flex24 (24 mm), Flex28 (28 mm), and FlexSOFT/Standard (31.5 mm) arrays underwent postoperative computed tomography to determine AID. A subsequent comparison was made between AID, predicted spiral ganglion place frequencies, and the default frequency filters for CI-alone (n = 84) and EAS users (n = 27). For CI-alone users with complete insertions who listened with maps fit with the default frequency filters (n = 48), frequency-to-place mismatch was quantified at 1500 Hz and angular separation between neighboring contacts was determined for electrodes in the 1 to 2 kHz region. Multiple linear regression was used to examine how frequency-to-place mismatch and angular separation of contacts influence consonant-nucleus-consonant (CNC) scores through 6 months postactivation.

Results: For CI recipients with complete insertions (n = 106, 95.5%), the AID (mean ± standard deviation) of the most apical contact was 428° ± 34.3° for Flex24 (n = 11), 558° ± 65.4° for Flex28 (n = 48), and 636° ± 42.9° for FlexSOFT/Standard (n = 47) arrays. For CI-alone users, default frequency filters aligned closely with the spiral ganglion map for deeply inserted lateral wall arrays. For EAS users, default frequency filters produced a range of mismatches; absolute deviations of ≤ 6 semitones occurred in only 37% of cases. Participants with shallow insertions and minimal or no residual hearing experienced the greatest mismatch. For CI-alone users, both smaller frequency-to-place mismatch and greater angular separation between contacts were associated with better CNC scores during the initial 6 months of device use.

Conclusions: There is significant variability in frequency-to-place mismatch among CI-alone and EAS users with default frequency filters, even between individuals implanted with the same array. When using default frequency filters, mismatch can be minimized with longer lateral wall arrays and insertion depths that meet the edge frequency associated with residual hearing for CI-alone and EAS users, respectively. Smaller degrees of frequency-to-place mismatch and decreased peripheral masking due to more widely spaced contacts may independently support better speech perception with longer lateral wall arrays in CI-alone users.

Figures

Fig. 1.

Mean angular insertion depth of the most apical electrode contact with standard deviations for complete insertions of Flex24, Flex28, and FlexSOFT/Standard arrays. The predicted cochlear place frequency determined by the spiral ganglion map equation is displayed on the right axis.

Fig. 2.

Correlation between cochlear duct length and angular insertion depth of the apical electrode contact for complete insertions of Flex24, Flex28, and FlexSOFT/Standard arrays. Electrode array types are shown with different colored shapes as defined in the legend.

Fig. 3.

(A) Mean angular insertion depth of each electrode contact for complete insertions of Flex24 (green squares), Flex28 (blue triangles), and FlexSOFT/Standard (red circles) arrays as a function of the manufacturer’s recommended default center frequency in CI-alone users. Individual differences by electrode array are demonstrated for (B) Flex24, (C) Flex28, and (D) FlexSOFT/Standard arrays. The estimated cochlear place frequency with the spiral ganglion map is represented by the solid black line. Dotted black lines denote data for individuals with partial insertions.

Fig. 4.

(A) Mean frequency-to-place mismatch in CI-alone users with complete insertions and default frequency filters represented by deviation in semitones of the electric place of stimulation from the estimated spiral ganglion frequency, and (B-D) degree of mismatch for individuals implanted with the same array. Dotted black lines denote data for individuals with partial insertions. Electrode array types are shown with different colored shapes as defined in the legend.

Fig. 5.

(A) Angular insertion depth of each electrode contact as a function of the default electric frequency filters for individual EAS device users with complete insertions and (B-D) data by array type. The estimated spiral ganglion cochlear place frequency is represented by the solid black line. Dotted black lines denote data for individuals with partial insertions.

Fig. 6.

(A) Frequency-to-place mismatch in electric-acoustic stimulation device users with default frequency filters, represented by deviation in semitones of the electric place of stimulation from the estimated spiral ganglion frequency and (B-D) additionally separated by array type. Dotted black lines denote data for individuals with partial insertions.

Fig. 7.

Relationship between absolute frequency-to-place mismatch at 1500 Hz (approximate spectral center of important speech information) and array type for CI-alone users with complete insertions. ns, not statistically significant; **, p < 0.01; ***, p < 0.001.

Fig. 8.

CNC word scores in RAU plotted as a function of absolute frequency-to-place mismatch at 1500 Hz (top row), and angular separation between neighboring electrode contacts (bottom row). Data in separate columns indicate results at 1, 3, and 6 months post-activation for CI-alone users with complete insertions (n = 48). Symbol color and shape reflects the electrode array type, as defined in the legend. Text at the right of each panel indicates the correlation illustrated with line fits. CNC, consonant-nucleus-consonant; RAU, rationalized arcsine unit.

Fig. 9.

Four schematics illustrating the relationship between frequency-to-place mismatch and angular separation between neighboring electrode contacts. Electrode contacts located in the approximate 1-2 kHz region on the spiral ganglion map (224° to 333°) are highlighted in green or red stripes to depict relatively smaller or larger frequency-to-place mismatches, respectively. Cochlear duct length (Panel A vs. B), basal insertion depth (Panel B vs. C), and differences in electrode contact spacing between array types (Panels A-C vs. D) all contribute to varying degrees of both mismatch and angular separation between electrode contacts across individuals.