Intracochlear electric potential measurements for estimating electrode array position in cochlear implantation: The in vivo utility of an ex vivo model.

INTRODUCTION

Stimulation of cochlear implant electrodes generates intracochlear electric potentials. The local electric potentials can be assessed using e.g. transimpedance matrix (TIM) and four-point impedance (Z). Both of these measurements are dependent on the cochlear dimensions and the distance between the electrode and the medial wall of the scala tympani (d). In a recent temporal bone study, a model based on electric potential measurements gave good predictions of scalar cross-sectional area (A) and d. The purpose of this study was to further improve this model and evaluate its clinical usefulness. To this end, the intraoperative TIM and Z measurements from cochlear implant patients were used as independent variables in the model to predict their A and d at each electrode contact, which were then compared to those measured from postoperative cone-beam computed tomography (CBCT) images.

METHODS

In an earlier study, six cadaveric temporal bones were sequentially implanted with three different electrode arrays: a lateral-wall electrode array, Slim Straight, and two precurved perimodiolar electrode arrays, Contour Advance and Slim Modiolar (Cochlear Ltd, Sydney, Australia). The TIM and Z measurements were performed alongside CBCT imaging, from which the A and d at each electrode contact were measured. From the TIM measurements, the peak amplitudes and decay rate of the electric potentials (EP) were computed. In this follow-up study, the statistical modeling of the ex vivo measurements was refined to better account for individual characteristics by employing mixed-effect models to predict the As and ds. Then, in vivo recordings from thirteen patients, of which six were implanted with the Slim Straight and seven with the Contour Advance electrode arrays, were retrospectively analyzed. The As and ds were measured from their postoperative CBCT images in a similar manner to the temporal bones. To validate the mixed-effects models developed with the temporal bone data, the patients' intraoperative TIM and Z measurements were used as independent parameters in the models to predict their As and ds. Finally, the TIM and Z parameters measured in vivo and ex vivo and the measured and predicted As and ds were compared using t-tests. Also, Pearson's correlation coefficients were computed between the measured and predicted in vivo As and ds.

RESULTS

Both the amplitudes, indicating electric potential peaks, and Zs, reflecting local potential differences, were lower in vivo than ex vivo (790 vs. 1090 Ω and 253 vs. 270 Ω, respectively, p < 0.001 for both), but no differences were detected in the decay of the electric potentials. In addition, the in vivo Zs were lower, and the electric potential decay was slower with the lateral wall (Slim Straight) compared to perimodiolar (Contour Advance) array (234 vs. 284 Ω and -94 vs. -160 Ω/mm, p < 0.001 for both). The mixed effects models with and without random effects explained 73 % and 51 % of the variance, respectively, for A. The mean absolute error between measured and predicted As was 12 %. For ds, the corresponding percentages were 65 %, 50 %, and 51 %. The correlations between the patients' measured and predicted As and ds were r = 0.60 and r = 0.48 (p < 0.001, for both). When compared in the basal, middle, and apical sections, the predicted As differed significantly from the measured values only in the middle section of the array (4.0 ± 0.48 mm vs 3.50 ± 0.36 mm, p < 0.001). For ds, the model gave too large estimates in the apical section of the array (1.04 ± 0.49 mm vs. 1.52 ± 0.48, p < 0.001).

CONCLUSION

The A at each electrode contact can be estimated using the TIM and Z measurements, which may help verify the correct alignment of the electrode array during the surgery. While the measured and predicted ds correlated with each other, there were significant differences between their absolute values. Given the large variation in ds for different array types, electrode-specific ds models could improve the accuracy of the predictions.