Evaluation of a Transimpedance Matrix Algorithm to Detect Anomalous Cochlear Implant Electrode Position

Introduction: Transimpedance measurements from cochlear implant electrodes have the potential to identify anomalous electrode array placement, such as tip fold-over (TFO) or fold-back, basal electrode kinking, or buckling. Analysing transimpedance may thus replace intraoperative or post-operative radiological imaging to detect any potential misplacements. A transimpedance algorithm was previously developed to detect deviations from a normal electrode position with the aim of intraoperatively detecting TFO. The algorithm had been calibrated on 35 forced, tip folded electrode arrays in six temporal bones to determine the threshold criterion required to achieve a sensitivity of 100%. Our primary objective here was to estimate the specificity of this TFO algorithm in patients, in a prospective study, for a series of electrode arrays shown to be normally inserted by post-operative imaging. Methods: Intracochlear voltages were intraoperatively recorded for 157 ears, using Cochlear’s Custom Sound™ EP 5 electrophysiological software (Cochlear Ltd., Sydney, NSW, Australia), for both Nucleus® CI512 and CI532 electrode arrays. The algorithm analysed the recorded 22 × 22 transimpedance matrix (TIM) and results were displayed as a heatmap intraoperatively, only visible to the technician in the operating theatre. After all clinical data were collected, the algorithm was evaluated on the bench. The algorithm measures the transimpedance gradients and corresponding phase angles (θ) throughout the TIM and calculates the gradient phase range. If this was greater than the predetermined threshold, the algorithm classified the electrode array insertion as having a TFO. Results: Five ears had no intraoperative TIM and four anomalous matrices were identified from heatmaps and removed from the specificity analysis. Using the 148 remaining data sets (n = 103 CI532 and n = 45 CI512), the algorithm had an average specificity of 98.6% (95.80%–99.75%). Conclusion: The algorithm was found to be an effective screening tool for the identification of TFOs. Its specificity was within acceptable levels and resulted in a positive predictive value of 76%, with an estimated incidence of fold-over of 4% in perimodiolar arrays. This would mean 3 out of 4 cases flagged as a fold-over would be correctly identified by the algorithm, with the other being a false positive. The measurements were applied easily in theatre allowing it to be used as a routine clinical tool for confirming correct electrode placement.