Objective: To develop an algorithm, based on the voltage matrix, for detecting regular cochlear implant (CI) electrode position during the implantation procedure, tip fold-over or basal kinking for lateral-wall electrodes. The availability of an algorithm would be valuable in clinical routine, as incorrect positioning of the electrode array can potentially be recognized intraoperatively. Design: In this retrospective study, intraoperative voltage matrix and postoperative digital volume tomography of 525 CI recipients were analyzed. On the basis of these data an algorithm was developed for detecting various kinds of electrode misplacements. Results: Seven incorrect electrode positions, three tip fold (0.57%) and four basal kinking (0.76%) were detected. For detecting correct positioning, a sensitivity of 100%, a specificity of 83.3%, and a positive predictive value (PPV) of 99.8% were found. For detecting tip fold-over, a sensitivity of 100%, a specificity of 100%, and a PPV of 100% were found. For detecting basal kinking, a sensitivity of 66%, a specificity of 100%, and a PPV of 100% were determined. Conclusion: The algorithm was found to be an effective screening tool for detecting tip fold-over or basal kinking.

In cochlear implant (CI) placement, in addition to individual requirements – especially inner-ear anatomy, hearing history, and the preoperative hearing status – precise electrode positioning continues to be of great importance [1‒4]. Alongside atraumatic insertion and sufficient cochlear coverage, it is particularly important to avoid incorrect positioning [4‒8]. Fortunately, extracochlear malpositions such as the vestibule, the Eustachian tube or the internal auditory canal are very rare under normal anatomical conditions [9]. However, intracochlear malpositions – such as tip folds, basal compression, kinking, and incomplete insertion – are also factors that can influence the outcome, as in such cases electrode contacts usually have to be deactivated and surgical revision may be required [10].

Imaging diagnostics are currently regarded as the gold standard for assessing the electrode-array position. Common procedures for this include computer tomography, digital volume tomography (DVT), and conventional X-ray imaging [11‒14]. These procedures differ in value owing to artefacts and summation effects [15, 16]. Last but not least, many years of experience in the assessment of radiological imaging are often required in order to identify various types of malposition. Postoperative imaging is still the most common procedure for checking the position of the electrode array, but this means that surgical correction is only possible in a follow-up surgical procedure. The ability to perform high-quality intraoperative imaging is restricted to a small group of surgical centers, as this involves a high logistical, time, and economic effort.

The extent to which postoperative imaging is still appropriate at all is also still under discussion, on account of the very low rate of malpositions in lateral-wall electrode arrays. For example, the overall incidence for tip fold has been stated to be 3–5% across all studies, and 0.8–1% when considering lateral-wall electrode arrays [10, 17]. This discussion is supported by current efforts to determine the correct position of the electrode array intraoperatively using electrophysiological measurements [11, 18, 19].

There has been a steady increase in methods and evaluation algorithms for electrophysiological measurements to monitor the function of CI systems. In recent years, it has been shown that individual methods can also be used with a high degree of certainty for some CI systems to check the position of the electrode array. A method for the electrophysiological detection of malpositions should have the following properties: high sensitivity and specificity, short measurement time, simple and rapid assessment, and a short time interval after electrode insertion. Methods with the above characteristics can enable the surgeon to correct the electrode array in real time. For example, it has been shown on the basis of electrically evoked compound action potential (ECAP) measurements that the spread of excitation (SOE) profiles can be used to detect malpositions with a high degree of sensitivity and specificity [11, 19, 20]. Hans et al. [11] report a sensitivity of 100% and a specificity of 98.93% for SOE profile determination for the detection of tip folds in Nucleus implants. Müller et al. [19] were able to confirm a sensitivity of 100% for SOE screening for the detection of tip folds for Nucleus implants; the specificity was 95.1%.

Other electrophysiological methods use the measurement of electrical impedances and voltage distributions instead of the neuronal response by means of telemetry functions already integrated into the system to monitor electrode array function. Depending on the manufacturer, these telemetry functions are referred to as electric field imaging (EFI) [21], transimpedance matrices (TIM) [18], or voltage matrices [22]. They can be used to display an electrical potential profile of the cochlea. During stimulation of an intracochlear electrode contact, the voltage difference between unstimulated intracochlear contacts and the extracochlear ground contact at the implant housing can be measured. Typically, there is a decay with increasing distance between stimulating and recording contacts along the electrode array. The voltage’s spatial distribution can be color-coded in a so-called heatmap [18, 21]. This shows a characteristic pattern for regular electrode array positions with a typical drop in voltage with increasing physical distance between the stimulating and recording electrodes. It is therefore to be expected that the classical heatmap will change in the event of an incorrect position of the electrode supports, such as a tip fold or kinking, as this will result in a closer physical distance between the electrodes involved [11, 18, 21, 23].

For Nucleus implants, Hoppe et al. [18] were able to determine a sensitivity of 100% and a specificity of 98.6% for this method; the positive predictive value (PPV) was 76% and the negative predictive value was 99.6–100%. Hans et al. [11] obtained similar results (sensitivity 100%, specificity 97.89%) for Nucleus implants (Hans et al. [11]). For the manufacturer MED-EL, there are currently no data on the measurement of SOE or voltage matrix detection of incorrect electrode placements. The present study was conducted to investigate the method of voltage matrix measurements for detection of tip fold-over or other non-regular electrode positions for the manufacturer MED-EL, despite the fact that the incidence of such incorrect positioning appears very low [17].

The objective of this study was to develop and optimize an algorithm for detecting tip fold-over and basal kinking for lateral-wall electrodes based on the retrospective analysis of more than 500 intraoperatively measured voltage matrices and corresponding imaging. The availability of such an algorithm would be valuable in clinical routine, as incorrect positions – even if they occur very rarely – could then be recognized and corrected during the initial operation.

Study Design and Subjects

In this retrospective analysis, we reviewed all patients who received a MED-EL CI at University Hospital Dresden between January 2010 and December 2023. We analyzed intraoperative voltage matrices and radiological images obtained postoperatively. The sole inclusion criterion for this study was the availability of a postoperative image by means of DVT.

Imaging

DVT examinations were carried out on the first postoperative day using a Flat Panel Computer Tomograph 3D Accuitomo 80 (J. Morita MFG. CORP., Kyoto, Japan). The following imaging parameters were used: tube current 8 mA, tube voltage 90 kV with a 360° rotation within 18.5 s. The raw projection images were reconstructed by using the software i-dixel (J. Morita MFG. CORP.) resulting in cylinder-shaped volumes 60 mm high with a voxel size of 125 μm.

The DVT images were analyzed with the software i-dixel (J. Morita MFG. CORP., Kyoto, Japan). To assess the intracochlear position of the electrode array the cochlear view was used; this view is defined as the plane that passes through the basal turn of the cochlea and is perpendicular to the modiolus according to the consensus paper [24]. This analysis was performed to allow judgment of whether the electrode array was positioned correctly or incorrectly. In this study, the term “correct position” refers to an electrode array whose contacts follow linearly the progression of the basilar membrane, without any kinking or tip fold-over. Thus, the analysis of each image resulted in one of the three classifications “correct position,” “tip fold-over,” or “basal kinking.”

Voltage Matrix Measurement

The intraoperative voltage matrix was obtained by using the impedance field telemetry (IFT) function of the clinical software MAESTRO (MED-EL, Innsbruck, Austria) after electrode array insertion and before the wound was sewn up. The voltages were measured by using a biphasic pulse with a phase duration of 24.17 µs, an interphase gap of 2.1 µs and a stimulation amplitude of 302.4 cu (current units), where 1 cu corresponds approximately to 1 μA. Note that the clinical software presents the results as impedances, which are (by Ohm’s law) directly proportional to the measured voltage owing to the constant amplitude of stimulation. The duration of the total IFT recording is approximately 10 seconds. Implants from the manufacturer MED-EL have 12 intracochlear contacts, resulting in a 12 × 12 voltage matrix when the voltage is measured for each stimulation electrode on all contacts.

As an example, Figure 1 shows a typical intraoperative voltage matrix, color-coded into a heatmap according to the magnitude of the voltage. The heatmap shows the impedances measured at all 132 non-stimulated contacts.

Fig. 1.

Example of a voltage matrix with correct cochlear placement, measured intraoperatively shown as a heatmap. The clinical software MAESTRO gives the result in impedance, which in this context is proportional to voltage (see text).

Fig. 1.

Example of a voltage matrix with correct cochlear placement, measured intraoperatively shown as a heatmap. The clinical software MAESTRO gives the result in impedance, which in this context is proportional to voltage (see text).

Close modal

Analysis of Voltage Matrix

The intraoperative IFT measurement was analyzed by using the software Python [25]. If more than one intraoperative IFT recording were performed, then the last measurement during surgery was selected for further analysis.

The voltages would normally be expected to decrease along the array as the distance between stimulating and recording contacts increases. However, where the electrode position is incorrect different behavior is seen. In tip fold-over, contacts located in the tip of the array are located anomalously close to medial contacts, resulting in abnormally high voltages measured over these contacts. In basal kinking, contacts intended to be located near the round window are in fact located near medial electrodes, again resulting in abnormally high voltages at these contacts. Thus, the pattern of the color-coded voltage matrix can reveal whether the electrodes are correctly placed or deviate from this (see Fig. 2, 5). To automate the process of distinguishing between normal and abnormal patterns, we developed an algorithm based on the analysis of cross-diagonals, summarized in Figure 2a–c, as follows.

Fig. 2.

Example of a case with correct electrode placement. a DVT image. b Heatmap of the voltage matrix with indication of the cross-diagonals (see text; note that all cross-diagonals were used, though here for illustration only five are shown). c Plot of the normalized squared sums of the odd and even cross-diagonals (details in text). The dotted black line indicates the middle of the electrode array, shown as the solid red diagonal in b. The y-axis in c represents the number of the diagonal offset relative to the middle of the array (shown for five diagonals). The baseline in c, shown as a green dashed line, is defined as the smallest value among the cross-diagonals squared sum of the representing the background voltage level.

Fig. 2.

Example of a case with correct electrode placement. a DVT image. b Heatmap of the voltage matrix with indication of the cross-diagonals (see text; note that all cross-diagonals were used, though here for illustration only five are shown). c Plot of the normalized squared sums of the odd and even cross-diagonals (details in text). The dotted black line indicates the middle of the electrode array, shown as the solid red diagonal in b. The y-axis in c represents the number of the diagonal offset relative to the middle of the array (shown for five diagonals). The baseline in c, shown as a green dashed line, is defined as the smallest value among the cross-diagonals squared sum of the representing the background voltage level.

Close modal

Algorithm for Detection of Normal Placement, Tip Fold-Over, or Basal Kinking

The voltage distribution of the voltage matrix was analyzed by using a quadratic measure, the normalized squared sum of the cross-diagonals, as shown in Figure 2b. First, all elements of the voltage matrix were normalized to the largest observed voltage. Then, for each cross-diagonal, the squared normalized voltages were summed and divided by the number of summed elements. The middle of the array is represented by the main cross-diagonal (solid red line in Fig. 2b). The diagonal offset represents the position of the cross-diagonal relative to the main cross-diagonal, with negative values (dotted lines in Fig. 2b) being mapped to the apical and positive values (dashed lines in Fig. 2b) to the basal side of the electrode array. The cross-diagonals separate into an even (red lines in Fig. 2b) and an odd (blue lines in Fig. 2b) group, due to respective different distances of its elements to the stimulated channel. For correct electrode placement a continuous trend was observed, as shown in Figure 2c, while for misplacements, such as tip fold-over or basal kinking, the incorrect proximities between the electrodes led to clear-cut peaks in the plots (see Results for details), thus allowing a distinction between correct positioning and misplacement.

Such peaks were detected automatically by using the “find_peaks” function in the Python SciPy package [26]. If no peak was detected, the case was assigned to a regular electrode position. If a peak was detected it was examined to identify the region in which it occurred. If a peak occurred in the apical region (i.e., diagonal offset below zero), this was interpreted as tip fold-over. If a peak occurred in the basal region (i.e., diagonal offset above zero), this was interpreted as basal kinking. Details and findings are set out in the Results section.

Exclusion of Voltage Matrices with Isolated Contacts

When the electrode array is inserted into the perilymph, air bubbles can form. Such contacts are electrically isolated and cannot measure the electric field in the cochlea. The heatmap shown in Figure 3a shows an example with contact 8 showing high impedances during the intraoperative IFT measurement. The corresponding postoperative IFT measurement did not show high impedances for electrode 8 (Fig. 3b), and it can be assumed that an air bubble caused the high intraoperative impedance of electrode 8 and that the bubble dissolved over time.

Fig. 3.

Example of a voltage matrix with an isolated contact: (a) measured intraoperatively, showing that electrode No. 8 is isolated (the anomalously low voltages are highlighted); (b) measured postoperatively, showing the attainment of normality and suggesting that the anomaly was due to an air bubble (see text).

Fig. 3.

Example of a voltage matrix with an isolated contact: (a) measured intraoperatively, showing that electrode No. 8 is isolated (the anomalously low voltages are highlighted); (b) measured postoperatively, showing the attainment of normality and suggesting that the anomaly was due to an air bubble (see text).

Close modal

Since IFT recordings with isolated contacts (e.g., Fig. 3a) may increase the likelihood of false-positive or false-negative results, such recordings were excluded. For this purpose, the following procedure was applied: A contact was automatically marked as isolated when the sum of the row and column non-diagonal voltages was less than 30% of the voltage in the main diagonal element, testing thereby whether the current at the contact could flow into the cochlea or was kept isolated at the contact surface. IFT recordings with one or more electrodes marked as isolated were excluded from further analysis.

Exclusion of Voltage Matrices with Anomalously High Baseline Levels

To avoid artefacts arising, e.g., from poor contact between the stimulation reference and the tissue, we screened all voltage matrices for anomalously high baseline levels. The background voltage level in the voltage matrix was quantified by computing the minimum value of the summed voltages in the cross-diagonals. This minimum value is referred to as the baseline (see the green dashed line in Fig. 2c). Some measurements contained a background voltage level in the voltage matrix that was too high, indicating an unusually large voltage remaining post-stimulus over the whole cochlea and potentially masking a peak. The distribution of all baseline values is shown in Figure 4. Fitting a Gaussian distribution resulted in a mean baseline value µ = 0.24 with a standard deviation of σ = 0.13. Baselines above the 95th percentile (µ+2σ) we defined as “elevated”, and all measurements with an elevated baseline were excluded from further analysis.

Fig. 4.

Distribution of baseline values (defined in Fig. 2c) for all 525 implants. The mean baseline is shown as a blue line. The blue dotted lines represent mean ± two standard deviations. All cases with baseline values above µ+2σ were excluded from further analysis.

Fig. 4.

Distribution of baseline values (defined in Fig. 2c) for all 525 implants. The mean baseline is shown as a blue line. The blue dotted lines represent mean ± two standard deviations. All cases with baseline values above µ+2σ were excluded from further analysis.

Close modal

Subjects

Of the cochlear implantations between January 2010 and December 2023, a total of 525 cochlear implantations fulfilled the inclusion criterion. Electrode arrays analyzed were: FLEX28 (N = 240), FLEXsoft (204), standard (44), FLEX24 (26), Medium (4), Compressed (4), FLEX20 (2), and FLEX26 (1).

Electrode Position as Determined by DVT

For the 525 implantations, the postoperative DVTs showed 518 correct electrode array placements (98.67%). An example of a DVT image of a correctly implanted electrode is shown above (Fig. 2a). Seven cases were identified by DVT with incorrect electrode positioning. For 3 cases, a tip fold-over (0.57%) and for 4 cases, a basal kinking (0.76%) was found. Details of these cases are given below in the context of their voltage-matrix results.

Excluded Voltage-Matrix Data

Of the 525 cases, 38 recordings were excluded because of isolated contacts due to intraoperative air bubbles as illustrated above (Fig. 1). For all 38 cases of air bubbles, the DVT showed a correct electrode position.

In a further 21 cases, an elevated baseline was identified, as described above; this could for example have been due to a dry reference contact at the stimulator. These measurements suffer from a masking of potential peaks (see, e.g., BK-4 below) and were, therefore, excluded. Among these 21 cases, 20 were found to have a correct electrode position by DVT, while one showed a basal kink.

Therefore, a total of 59 of the 525 voltage matrices (11.2%) were excluded from further analysis. The remaining 466 recordings were analyzed further.

Algorithm Results

Results for a normally, correctly placed implant are shown above (Fig. 2). Corresponding results for electrodes with incorrect positioning are collated in Figure 5.

Fig. 5.

a DVT image, voltage matrix heatmap, and normalized even and odd squared sum of cross-diagonals of the three cases: three with tip fold-over (TFO‑1, TFO‑2, TFO‑3). Note that in TFO‑3 it is difficult to recognize the tip fold-over in the two-dimensional DVT image, as contact No. 1 is projected on top of contact No. 5. b DVT image, voltage matrix heatmap, and normalized even and odd squared sum of cross-diagonals of the four cases with basal kinking (BK‑1, BK‑2, BK‑3, BK‑4). For BK‑4 the results of a postoperative measurement are also shown. For details see text.

Fig. 5.

a DVT image, voltage matrix heatmap, and normalized even and odd squared sum of cross-diagonals of the three cases: three with tip fold-over (TFO‑1, TFO‑2, TFO‑3). Note that in TFO‑3 it is difficult to recognize the tip fold-over in the two-dimensional DVT image, as contact No. 1 is projected on top of contact No. 5. b DVT image, voltage matrix heatmap, and normalized even and odd squared sum of cross-diagonals of the four cases with basal kinking (BK‑1, BK‑2, BK‑3, BK‑4). For BK‑4 the results of a postoperative measurement are also shown. For details see text.

Close modal

In all 3 cases of tip fold-over, the irregular placing was correctly detected in the automated peak-detection, as shown in the respective cross-diagonal analyses.

Among the 4 cases of basal kinking, the elevated baseline in BK-4 is conspicuous (Fig. 5). The software correctly identified the kinking in BK-1 and BK-3; its failure to do so in BK-2 may have been due to the fact that the kinking did not lead to close proximity of electrodes (see BK-2, DVT image). In BK-4, a small peak is discernible in the relevant region of the cross-diagonal plot, but against the background of the high baseline this case had been excluded from analysis. An overall comparison of the seven relevant cases found by DVT to be mispositioned, including information on the patients and their implants, is given in Table 1.

Table 1.

Details of cases with non-regular electrode position

Case No.Patient’s age, yearsImplant modelElectrode typeSideImplantation dateResult
DVTalgorithm
TFO-1 76 Mi1200 FLEX28 Left Sep 2014 Tip fold tip fold 
TFO-2 70 Mi1000 FLEXsoft Right Jan 2011 Tip fold tip fold 
TFO-3 76 Mi1000 FLEXsoft Right May 2016 Tip fold tip fold 
BK-1 70 Mi1000 FLEXsoft Left Mar 2011 Basal kinking basal kinking 
BK-2 54 Mi1000 FLEXsoft Right May 2011 Basal kinking no irregularity 
BK-3 68 Mi1000 FLEXsoft Right Aug 2011 Basal kinking basal kinking 
BK-4 59 Mi1250 FLEX28 Left Jun 2020 Basal kinking no result1 
Case No.Patient’s age, yearsImplant modelElectrode typeSideImplantation dateResult
DVTalgorithm
TFO-1 76 Mi1200 FLEX28 Left Sep 2014 Tip fold tip fold 
TFO-2 70 Mi1000 FLEXsoft Right Jan 2011 Tip fold tip fold 
TFO-3 76 Mi1000 FLEXsoft Right May 2016 Tip fold tip fold 
BK-1 70 Mi1000 FLEXsoft Left Mar 2011 Basal kinking basal kinking 
BK-2 54 Mi1000 FLEXsoft Right May 2011 Basal kinking no irregularity 
BK-3 68 Mi1000 FLEXsoft Right Aug 2011 Basal kinking basal kinking 
BK-4 59 Mi1250 FLEX28 Left Jun 2020 Basal kinking no result1 

1Not analyzed owing to elevated baseline

Table 2 summarizes the results obtained by the algorithm for detecting correctly placed electrodes, tip fold-overs and basal kinking. For each type of finding, the numbers of true- and false-positive and true- and false-negative results are used to calculated the sensitivity, specificity, and positive prediction value (PPV). Note that BK-4 was excluded from the analysis because of an elevated baseline.

Table 2.

Algorithm results for the detection of regular electrode position, tip fold-over, and basal kinking

Finding categoryTrue positiveFalse negativeTrue negativeFalse positiveSensitivitySpecificityPPV
TPFNTNFPTP/(TP + FN), %TN/(TN + FP), %TP/(TP + FP), %
Correct position 460 100 83.3 99.8 
Tip fold-over 463 100 100 100 
Basal kinking 463 66.6 100 100 
Finding categoryTrue positiveFalse negativeTrue negativeFalse positiveSensitivitySpecificityPPV
TPFNTNFPTP/(TP + FN), %TN/(TN + FP), %TP/(TP + FP), %
Correct position 460 100 83.3 99.8 
Tip fold-over 463 100 100 100 
Basal kinking 463 66.6 100 100 

Total N = 466 (exclusion of 38 cases with isolated contact and of 21 with elevated baseline; see text).

“True positive” means the number of positive cases that were identified correctly by the algorithm. “False negative” means the number of positive cases missed by the algorithm. “True negative” indicates the number of negative cases that were identified correctly by the algorithm. “False positive” indicates the number of negative cases that the algorithm incorrectly classified as positive.

PPV, positive predictive value.

Regular Position

460 of the 466 evaluable cases were identified correctly by the algorithm as having the CI in a correct position. Five were identified correctly as wrongly placed (cases TFO-1, TFO-2, TFO-3, BK-1, and BK-3). In one case, the algorithm indicated correct positioning, although imaging showed (slight) basal kinking (case BK-2). This yielded a sensitivity of 100%, a specificity of 83.3%, and a PPV of 99.8%.

Tip fold-over

All three tip fold-overs (cases TFO-1, TFO-2, TFO-3) were detected correctly by the algorithm, yielding a sensitivity of 100%. The algorithm gave no false-positive result, yielding a specificity of 100%. The PPV was accordingly 100%.

Basal Kinking

The basal kinking algorithm detected two of the three instances of basal kinking, yielding a sensitivity of 66.6%. One (slight) basal kinking (BK-2) was not detected, as shown in Figure 5. The algorithm gave no false-positive result, yielding a specificity of 100% and a PPV of 100%.

The placement of the CI electrode can influence the outcome of hearing rehabilitation, as it provides the foundation for future fitting, and misalignments have a negative impact on this from the very beginning [27‒29]. Although considered the gold standard, radiological imaging as a method of position detection is associated with clear disadvantages. Naturally, it comes with an undesirable exposure to radiation, and it also requires a very experienced radiologist to identify the various possible irregularities. Furthermore, in the vast majority of clinical institutions, such imaging is performed postoperatively rather than intraoperatively, so that meaningful insight into possible misalignment is acquired too late to allow immediate revision [30]. In comparison, the use of ECAP-based SOE profiles seems to be very promising [11, 19, 20] when it comes to correctly identifying electrode placement, although it has its own methodological limitations: the measurement takes several minutes, and there is no guarantee that an adequate number of neuronal responses will be recorded [19, 31]. There is therefore a need for alternative methods that are independent of any neuronal contributions.

In recent years, much effort has been put into interpreting solely the measurements of electrical impedances and voltage distributions to determine electrode positioning [11, 18, 23, 32, 33]. Measurement of impedance is mandatory in today’s CI implantation to determine integrity of the implant as well as electrode-to-tissue impedances and is therefore firmly embedded in clinicians’ routine. However, these studies focused primarily on the correct detection of tip folds and not on other possible misalignments [18, 34]. Moreover, they used rather small clinical samples [23] or in temporal bone studies [33]. Hans et al. [11] show for both methods (SOE and TIM) measured in the same study group a high sensitivity (100%) and specificity (98.93% and 97.89%).

Our study demonstrates clear advantages of the algorithm presented, as it can identify several different electrode positions. It was developed on the basis of a meaningfully sized clinical sample. Measuring the IFT takes only a few seconds, while the analysis of the IFT by the algorithm is virtually immediate, resulting in a near real-time application. In this study, retrospective application of this algorithm showed high reliability in detecting several electrode placements (normal, tip fold-over and basal kinking), as validated by diagnostic imaging. After excluding all data sets with problematic intraoperative measurement conditions indicated by isolated contacts (probably caused by air bubbles) or by unusually large post-stimulus voltage over the entire cochlea (probably due to a dry reference contact at the stimulator), 466 cases were analyzed. In CI implantation, the presence of air bubbles or dry reference contacts is not rare. These problems normally vanish over time, as revealed by repeated measurements at the end of the operation. Owing to the retrospective observation in this study, and to the fact that such measurements were not performed routinely, these data sets could not be used further for the detection of positioning, as they completely prevent retrospective position detection based on the algorithm. One of the cases that were excluded from further analysis by the algorithm, because of a conspicuously large remaining post-stimulus voltage, showed basal kinking in imaging. All other cases showed correct electrode positioning. For the case with the basal kinking (BK-4), a postoperative recording (BK-4 [postop]) did show a normal voltage pattern (Fig. 5, below right). It can therefore be assumed that the anomalous intraoperative voltages were due to poor contact during surgery between the stimulator housing reference contact and the tissue. The postoperative voltage matrix analysis detected the basal kinking correctly, as shown in Figure 5.

The intraoperative evaluation of the electrode array position based on the analysis of the voltage matrix was possible in 466 (89%) of 525 intraoperative recordings in our study. An comparably large fraction of evaluable measurements has been reported for the TIM measurement for Nucleus implants (94%) [18]. In contrast, an analysis of the evaluation of the SOE profiles for Nucleus implants showed that, depending on the type of electrode array, these can be used for intraoperative confirmation of the electrode array position in only 80% of cases [19]. This indicates that the measurement of electrical impedances and voltage distributions has an advantage over neuronal response detection in terms of complete and analyzable data-set recording in the intraoperative setting.

Among the 466 analyzable cases, we were able to detect 460 regular electrode placements, with five cases correctly rejected and one false-positive detection. This resulted in a sensitivity of 100%, a specificity of 83.3% and a PPV of 99.8%. The three known tip folds were classified with a sensitivity of 100%, a specificity of 100%, and a PPV of 100%. 463 were correctly rejected, none was missed and no false-positive was recorded. The three known cases of basal kinking that could be included for analysis were classified with a sensitivity of 66.6%, a specificity of 100%, and a PPV of 100%. 463 were correctly rejected, one was missed and no false-positive was recorded.

To find a reasonable compromise in minimizing the remaining uncertainty in the classification, we adapted the view of Hoppe et al. [18] by putting the patient’s interest first. As a result, sensitivities were adjusted so that correct placements were more likely to be overlooked or incorrectly rejected than true misplacements were to be missed. Thus, we were able to correctly identify all tip folds in our clinical sample. By contrast, a lower sensitivity was achieved in detecting genuine basal kinking, for which just 2 of 3 cases were identified correctly. This can be explained by the relative position of the basal electrode contacts involved, which did not overlap as distinctly as in the other cases of misplacement (see Fig. 5, case BK-2). In this particular case, only the two most basal electrodes were affected; that is very close to a correct placement, and can still result in monotonic and inconspicuous behavior in the side-diagonals. In all other cases of identified misplacements, at least four electrodes were affected.

In addition, it must be mentioned that, owing to the low occurrence of basal kinking in this study, only one single overlooked case impacts the sensitivity by 33.3% and thus worsens excessively the results shown. In general, it is to be noted that owing to the low number of misplacements the “correct position” specificity and “tipfold” and “basal kinking” PPV are strongly quantized and do not reflect the true general probability of these misplacement types. This thus calls for additional cases to further refine the algorithm, even though a considerable number of misalignments were investigated in this study. The incidence of tip folds identified in this study was 0.6%; that is within the range of earlier investigations concerning lateral-wall electrodes [10, 17], while sufficient valid numbers for the probability of basal kinking are not available.

Finally, the proven sensitivities and specificities support the assertion that the algorithm is well suited as a tool for the intraoperative detection of CI electrode placement. Owing to the low incidence of tip fold-over or basal kinking with lateral electrodes and the algorithm’s high sensitivity, it is most meaningful when correct placement is detected. In these cases, it could be considered to supplant intraoperative and/or postoperative radiological imaging in the future.

The authors thank Dr. Paul Woolley for providing language-editing service.

This study protocol was reviewed and approved by the Local Ethics Committee: Ethikkommission an der Technischen Universität Dresden (IRB00001473, IORG0001076), approval No. SR+BO-EK-47012022. Written informed consent from participants was not required in accordance with local guidelines.

The authors report no other potential or actual conflict of interest. The authors alone are responsible for the content and writing of this article.

This study was supported by MED-EL Elektromedizinische Geräte Deutschland GmbH.

A.F.-T.: conceptualization, funding acquisition, methodology, and writing, review, and editing S.L.: writing, review, and editing; S.S.: formal analysis, methodology, and software K.M.: data curation E.S.: writing, review, and editing M.N.: review T.Z.: review and editing, funding acquisition All authors have read and agreed to the published version of the manuscript.

The data that support the findings of this study are not publicly available due to information that could compromise the privacy of research participants but are available from the corresponding author upon reasonable request.

1.
Hoppe
U
,
Hast
A
,
Hocke
T
.
Validation of a predictive model for speech discrimination after cochlear impIant provision
.
HNO
.
2023
;
71
(
Suppl 1
):
53
9
.
2.
Hoppe
U
,
Hocke
T
,
Hast
A
,
Iro
H
.
Cochlear implantation in candidates with moderate-to-severe hearing loss and poor speech perception
.
Laryngoscope
.
2021
;
131
(
3
):
E940
5
.
3.
Perkins
EL
,
Labadie
RF
,
O’Malley
M
,
Bennett
M
,
Noble
JH
,
Haynes
DS
, et al
.
The relation of cochlear implant electrode array type and position on continued hearing preservation
.
Otol Neurotol
.
2022
;
43
(
6
):
e634
40
.
4.
Weller
T
,
Timm
ME
,
Lenarz
T
,
Büchner
A
.
Cochlear coverage with lateral wall cochlear implant electrode arrays affects post-operative speech recognition
.
PLoS One
.
2023
;
18
(
7
):
e0287450
.
5.
Cuda
D
,
Murri
A
.
Assessment of cochlear trauma and telemetry measures after cochlear implantation: a comparative study between Nucleus® CI512 and CI532 electrode arrays
.
Audiol Res
.
2019
;
9
(
1
):
223
.
6.
Ketterer
MC
,
Aschendorff
A
,
Arndt
S
,
Beck
R
.
Electrode array design determines scalar position, dislocation rate and angle and postoperative speech perception
.
Eur Arch Otorhinolaryngol
.
2022
;
279
(
9
):
4257
67
.
7.
Patro
A
,
Lindquist
NR
,
Schauwecker
N
,
Holder
JT
,
Perkins
EL
,
Haynes
DS
, et al
.
Comparison of speech recognition and hearing preservation outcomes between the mid-scala and lateral wall electrode arrays
.
Otol Neurotol
.
2024
;
45
(
1
):
52
7
.
8.
Zelener
F
,
Majdani
O
,
Roemer
A
,
Lexow
GJ
,
Giesemann
A
,
Lenarz
T
, et al
.
Relations between scalar shift and insertion depth in human cochlear implantation
.
Otol Neurotol
.
2020
;
41
(
2
):
178
85
.
9.
Cheung
LL
,
Kong
J
,
Chu
PY
,
Sanli
H
,
Walton
J
,
Birman
CS
.
Misplaced cochlear implant electrodes outside the cochlea: a literature review and presentation of radiological and electrophysiological findings
.
Otol Neurotol
.
2022
;
43
(
5
):
567
579
.
10.
Dhanasingh
A
,
Jolly
C
.
Review on cochlear implant electrode array tip fold-over and scalar deviation
.
J Otol
.
2019
;
14
(
3
):
94
100
.
11.
Hans
S
,
Arweiler-Harbeck
D
,
Kaster
F
,
Ludwig
J
,
Hagedorn
E
,
Lang
S
, et al
.
Transimpedance matrix measurements reliably detect electrode tip foldfold-over in cochlear implantation
.
Otol Neurotol
.
2021
;
42
(
10
):
e1494
502
.
12.
Kaul
VF
,
Brannan
Z
,
Keith
J
,
Hittle
B
,
Riggs
W
,
Hiss
M
, et al
.
Post-operative evaluation of computed tomography imaging following cochlear implantation
.
Am J Otolaryngol
.
2024
;
45
(
1
):
104081
.
13.
Trieger
A
,
Schulze
A
,
Schneider
M
,
Zahnert
T
,
Mürbe
D
.
In vivo measurements of the insertion depth of cochlear implant arrays using flat-panel volume computed tomography
.
Otol Neurotol
.
2011
;
32
(
1
):
152
7
.
14.
Vogl
TJ
,
Tawfik
A
,
Emam
A
,
Naguib
NNN
,
Nour-Eldin
A
,
Burck
I
, et al
.
Pre-intra- and post-operative imaging of cochlear implants
.
Rofo
.
2015
;
187
(
11
):
980
9
.
15.
Aschendorff
A
.
Imaging in cochlear implant patients
.
Laryngo-Rhino-Otol
.
2011
;
90
(
S 01
):
S16
21
.
16.
Gabrielpillai
J
,
Burck
I
,
Baumann
U
,
Stöver
T
,
Helbig
S
.
Incidence for tip foldover during cochlear implantation
.
Otol Neurotol
.
2018
;
39
(
9
):
1115
21
.
17.
Högerle
C
,
Englhard
A
,
Simon
F
,
Grüninger
I
,
Mlynski
R
,
Hempel
JM
, et al
.
Cochlear implant electrode tip foldfold-over: our experience with long and flexible electrode
.
Otol Neurotol
.
2022
;
43
(
1
):
64
71
.
18.
Hoppe
U
,
Brademann
G
,
Stöver
T
,
Ramos de Miguel
A
,
Cowan
R
,
Manrique
M
, et al
.
Evaluation of a transimpedance matrix algorithm to detect anomalous cochlear implant electrode position
.
Audiol Neurootol
.
2022
;
27
(
5
):
347
55
.
19.
Müller
A
,
Kropp
MH
,
Mir-Salim
P
,
Aristeidou
A
,
Dziemba
OC
.
Intraoperatives tip-foldover-screening mittels spread of excitation messungen
.
Z Med Phys
.
2021
;
31
(
3
):
276
88
.
20.
Dziemba
OC
,
Hocke
T
,
Müller
A
,
Kaftan
H
.
Excitation characteristic of a bipolar stimulus for broadband stimulation in measurements of electrically evoked auditory potentials
.
Z Med Phys
.
2018
;
28
(
1
):
73
7
.
21.
Vanpoucke
FJ
,
Boermans
P-PB
,
Frijns
JH
.
Assessing the placement of a cochlear electrode array by multidimensional scaling
.
IEEE Trans Biomed Eng
.
2012
;
59
(
2
):
307
10
.
22.
Joly
C-A
,
Reynard
P
,
Hermann
R
,
Seldran
F
,
Gallego
S
,
Idriss
S
, et al
.
Intra-cochlear current spread correlates with speech perception in experienced adult cochlear implant users
.
J Clin Med
.
2021
;
10
(
24
):
5819
.
23.
Klabbers
TM
,
Huinck
WJ
,
Mylanus
EAM
.
Comparison between transimpedance matrix (TIM) measurement and X-ray fluoroscopy for intraoperative electrode array tip foldfold-over detection
.
Otol Neurotol
.
2021
;
42
(
10
):
e1457
63
.
24.
Verbist
BM
,
Skinner
MW
,
Cohen
LT
,
Leake
PA
,
James
C
,
Boëx
C
, et al
.
Consensus panel on a cochlear coordinate system applicable in histologic, physiologic, and radiologic studies of the human cochlea
.
Otol Neurotol
.
2010
;
31
(
5
):
722
30
.
25.
Van Rossum
G
.
Python tutorial, Technical Report CS-R9526, Centrum voor Wiskunde en Informatica (CWI)
.
Amsterdam
;
1995
.
26.
Virtanen
P
,
Gommers
R
,
Oliphant
TE
,
Haberland
M
,
Reddy
T
,
Cournapeau
D
, et al
.
SciPy 1.0: fundamental algorithms for scientific computing in Python
.
Nat Methods
.
2020
;
17
(
3
):
261
72
.
27.
Sabban
D
,
Parodi
M
,
Blanchard
M
,
Ettienne
V
,
Rouillon
I
,
Loundon
N
.
Intra-cochlear electrode tip fold-over
.
Cochlear Implants Int
.
2018
;
19
(
4
):
225
9
.
28.
Trakimas
DR
,
Kozin
ED
,
Ghanad
I
,
Barber
SR
,
Curtin
H
,
Remenschneider
AK
.
Precurved cochlear implants and tip foldover: a cadaveric imaging study
.
Otolaryngol Head Neck Surg
.
2018
;
158
(
2
):
343
9
.
29.
Wanna
GB
,
Noble
JH
,
Carlson
ML
,
Gifford
RH
,
Dietrich
MS
,
Haynes
DS
, et al
.
Impact of electrode design and surgical approach on scalar location and cochlear implant outcomes
.
Laryngoscope
.
2014
;
124 Suppl 6
(
0 6
):
S1
7
.
30.
McJunkin
JL
,
Durakovic
N
,
Herzog
J
,
Buchman
CA
.
Early outcomes with a slim, modiolar cochlear implant electrode array
.
Otol Neurotol
.
2018
;
39
(
1
):
e28
33
.
31.
Dziemba
OC
,
Aristeidou
A
,
Brill
S
.
Slope of electrically evoked compound action potential amplitude growth function is site-dependent
.
Cochlear Implants Int
.
2021
;
22
(
3
):
136
47
.
32.
Aebischer
P
,
Meyer
S
,
Caversaccio
M
,
Wimmer
W
.
Intraoperative impedance-based estimation of cochlear implant electrode array insertion depth
.
IEEE Trans Biomed Eng
.
2021
;
68
(
2
):
545
55
.
33.
de Rijk
SR
,
Tam
YC
,
Carlyon
RP
,
Bance
ML
.
Detection of extracochlear electrodes in cochlear implants with electric field imaging/transimpedance measurements: a human cadaver study
.
Ear Hear
.
2020
;
41
(
5
):
1196
207
.
34.
Ayas
M
,
Muzaffar
J
,
Borsetto
D
,
Eitutis
S
,
Phillips
V
,
Tam
YC
, et al
.
A scoping review on the clinical effectiveness of Trans-Impedance Matrix (TIM) measurements in detecting extracochlear electrodes and tip fold overs in Cochlear Ltd devices
.
PLoS One
.
2024
;
19
(
3
):
e0299597
.