Anatomy-Based Fitting in Cochlear Implants: Potential for Optimizing Postoperative Outcomes: A Pilot Study

INTRODUCTION

Up until recently, the frequencies of cochlear implant (CI) electrode contacts have been set uniformly for most CI recipients (1). In the default setting, there is a discrepancy between the frequencies that are allocated to the individual electrode channels of the CI and the tonotopic frequency representation on the basilar membrane along the cochlea. This discrepancy can potentially limit postoperative performance, leading to reduced sound quality and a challenging adaptation phase. In particular, individuals with postlingual hearing loss are known to have poor results due to the frequency shifts, as they have had previous experience with normal acoustic perception and frequency ranges (2). This “one size fits all” approach fails to provide optimal results for every individual and is a limiting factor in achieving favorable postoperative hearing results (1). In contrast, optimizing the electrode cochlear coverage by taking into account the individual size of the cochlea seems to have a positive effect on speech perception outcomes (3). In addition, reducing the mismatch between electrode frequency allocation and cochlear tonotopy is believed to contribute to better postoperative outcomes (4,5).

The OTOPLAN planning software (CASCINATION, Bern, Switzerland, and MED-EL, Innsbruck, Austria) allows measurements to be taken in individual cochleae using computed tomography (CT) images. After defining anatomical landmarks, such as the modiolus, round window, and cochlear dimensions, the software enables the visualization and 3D reconstruction of the inner ear, automatic measurements of the diameter, width, height, cochlear duct length (CDL), and choice of the correct electrode array. Studies validating the software showed good to excellent reliability in the calculation of the CDL (5–7). Using postoperative CT images, OTOPLAN enables the identification of electrode contacts and automatic calculation of the angular insertion depth. It also facilitates individualized frequency mapping based on the Greenwood Function (8). This aims to move away from the “one size fits all” approach, taking into account physiological tonotopy. In a comprehensive review of the application of the otological OTOPLAN surgical planning software, only five studies about visualizing the electrode position using postoperative images, the evaluation of postoperative imaging related to electrode position and insertion depth, and three studies about creating an individualized frequency map (the relocation of center frequencies based on a subject-specific frequency map) could be identified (9).

Several studies have indicated that personalized adjustments, including electrode array length, insertion depth, and anatomy-based fitting (ABF), may lead to improvements in postoperative hearing (6,8,10–12). In order to perform ABF, knowledge about the actual electrode contact locations is a prerequisite. As already mentioned, this information is typically extracted from postoperative imaging data, such as CT or digital volume tomography (DVT) scans. Obtaining these data is, however, not always feasible in everyday clinical routine. First, clinicians hesitate to perform nonessential radiologic procedures due to additional radiation exposure. Because of this, ABF is rarely performed in underage CI recipients. Second, clinics often struggle with capacity issues in their radiology departments, which makes obtaining the necessary postoperative data for ABF a logistic challenge. In theory, an updated version of the OTOPLAN software can now use postoperative x-ray imaging to determine the electrode contact locations (13) after implantation, but the applicability of this approach using real-world clinical data has yet to be investigated.

The aim of this retrospective work was to investigate whether using preoperative measurements of the cochlea parameters to predict the resulting electrode contact locations results in ABF maps that are different compared with clinically based fitting (CBF) using the standard allocation of frequency bands. In addition, the extent of potential differences was quantified to assess whether the approach to predict postoperative electrode locations from preoperative imaging could be feasible.

MATERIALS AND METHODS

Study Design, Inclusion Criteria, and Ethics Approval

The present monocentric retrospective registry study includes subjects who underwent cochlear implantation at the Department of Otorhinolaryngology, Kepler University Hospital GmbH, between 2004 and 2021. All adult patients who received a MED-EL CI, fully inserted through the round window, who had completed the fitting process in our department, and who had a preoperative CT with a maximum slice thickness of 1 mm available in the hospital information system were included. The exclusion criteria comprised incomplete insertion, incomplete fitting, or lack of a CT of appropriate quality. Children and adults who received an implant from a different brand were excluded from the chart review.

The study protocol was approved by the Ethics Committee in accordance with the Declaration of Helsinki (EK Nr. 1010/2023).

Clinical Data

The hospital information system and MAESTRO (MED-EL, Innsbruck, Austria) software were used to collect preoperative and postoperative data, including demographic information (age, sex, type of hearing loss), intraoperative data (side, type, and depth of implantation), CI characteristics (type of electrode array, CBF frequency band parameters of the electrode channels), and imaging data (preoperative CT of the temporal bone). Postoperative images, videos, and information from the surgical report were used to confirm the surgical approach, the inserted electrode type, and whether the electrode was fully inserted. Using Van de Heyning's audiological classification criteria, the preoperative audiological status was described (14).

Measurements

Because CT was not done routinely in our clinic after cochlear implantation, it was decided to conduct a simulation of the electrode array position based on the preoperative CT scans and the actually inserted electrode types. At first, the cochlear coordinate system was manually defined. Cochlea parameters were measured also manually: diameter of basal turn (A), defined as the distance between the round window center and the contralateral wall; cochlear width (B), the distance between cochlear walls perpendicular to the A value; and the height (H), the distance between the base of the cochlea and the apex on an orthogonal plane. The measurement procedure is depicted in Figure 1. Two CDL estimates were automatically calculated by the OTOPLAN software: whole linear CDL at the level of the organ of Corti (OC)/basilar membrane and CDL (OC from 0 degrees), taking into account the cutting angle effect, defined as the length of the cochlear duct from the center of the round window to the apical tip. Based on the actually inserted electrode type, the electrode array position was simulated with OTOPLAN Version 4 using the preoperative temporal bone CT scans. The following parameters describing the predicted electrode contact locations were recorded: linear insertion depth (LID) and angular insertion depth (AID). Estimated frequencies according to the Greenwood frequency map, corresponding to the ABF and measured cochlear parameters (length, width, height, and CDL), were recorded. Figure 1 illustrates the measurement of cochlear parameters within the OTOPLAN software.

FIG. 1.

Measuring the size of the cochlea in OTOPLAN. Diameter (A value): distance between green points; width (B value); distance between blue points; height (H value): distance between red points.

The measurements were manually performed by three different ENT doctors, and each researcher repeated the task three times. The accuracy of the measurements was verified by assessing intraclass correlation coefficients (ICCs).

In a next step, the resulting predicted electrode contact frequencies were compared with the actual filter band center frequencies of the subjects' clinical fitting maps in the MAESTRO fitting software.

Data Collection

All study data were collected retrospectively, pseudonymized, and stored in a registry in a spreadsheet file.

Statistical Analysis

Descriptive statistics were reported for all observations. The Shapiro-Wilk test was used to test for a normal distribution of data. A t test was performed for normally distributed data, and a nonparametric Wilcoxon test was performed for non-normally distributed data. All p values were purely descriptive and indicated trends. ICC for intra- and inter-rater reliabilities were estimates, and their 95% confident intervals were calculated, respectively, one-way random effects and on a two-way mixed effect model with agreement. Assessment of the ICC results followed previously published criteria (15,16). Missing data were replaced with mean values by assessing the ICC. In all other than ICC calculations, missing values were excluded. IBM SPSS Statistics software Version 3.0 (IBM; Armonk, NY) was used for statistical analyses.

RESULTS

Group Characteristics

Ninety-five people received 117 CIs between 2004 and 2021 at the Department of Otorhinolaryngology, Kepler University Hospital GmbH. Twenty-three subjects were ultimately included in our study (Supplemental Fig. 1, http://links.lww.com/MAO/C139). The median age at implantation was 46.2 years (range: 35.7–62.5 yr, IQR of 26.8 yr). Twelve subjects (52.2%) were male, and 11 (47.8%) were female. The indications for cochlear implantation were single-sided deafness in 5 subjects (21.7%), bilateral sensorineural hearing loss in 7 subjects (30.5%), and asymmetrical hearing loss in 11 subjects (47.8%).

The median time between implantation and study entry was 3.5 years (range 1.8–5.2 yr, IQR 3.4 yr). Twenty FLEX28 electrode arrays, two STANDARD electrode arrays, and one FLEXSOFT electrode array (both 31 mm long) were implanted. The preoperative CT scans had a slice thickness of 1 mm for one patient and 0.6 mm for all the others.

Cochlear Parameters

The averaged values of the cochlear parameters of all measurements of all three investigators, obtained using OTOPLAN, are shown in Table 1. The range of cochlea diameter, width, and height were 8.23 and 10.12, 5.81 and 7.53, and 3.20 and 4.17 mm, respectively. The CDL (full OC) ranged from 30.96 to 39.93 mm, and the CDL (OC from 0 degrees), was between 28.46 and 37.43 mm. Upon examining the CT scans of the cochleae, we observed no instances of malformations.

TABLE 1.

Cochlea parameter based on preoperative CT scans

	(n = 23)
Cochlea Parameter	Minimum	Maximum	Mean	SD	Q1	Median	Q3	IQR	p(SW)
Diameter A (mm)	8.23	10.12	9.14	0.48	8.79	9.13	9.49	0.70	0.982
Width B (mm)	5.81	7.53	6.66	0.43	6.44	6.66	6.92	0.48	0.807
Height H (mm)	3.20	4.17	3.81	0.23	3.67	3.86	4.01	0.34	0.363
CDL (full OC)	30.96	39.93	35.27	2.21	34.00	35.22	36.83	2.83	0.768
CDL (OC from 0°)	28.46	37.43	32.78	2.21	31.50	32.72	34.33	2.83	0.769

CDL indicates cochlear duct length; IQR, interquartile range; Q1, first quartile; Q3, third quartile; p(SW), p value for the Shapiro-Wilk test; SD, standard deviation.

Predicted Electrode Contact Parameters

Table 2 shows the estimated frequencies for the CI electrode positions based on preoperative temporal bone CT scans and the information about the inserted electrode type. LID, AID, and estimated frequency according to the Greenwood frequency map for the most apical electrode contact number 1 (C1) are presented. Minimal AID (in degrees) was 495.4, and maximum AID was 683.8.

TABLE 2.

Predicted parameters of electrode contact number 1

	(n = 23)
C1	Minimum	Maximum	Mean	SD	Q1	Median	Q3	IQR	p(SW)
AID (°)	495.4	683.8	583.7	51.1	549.9	567.6	602.4	52.5	0.117
LID (%)	57.6	78.4	67.3	5.7	64.0	65.7	69.3	5.3	0.136
LID (mm)	27.9	31.3	28.5	1.1	28.0	28.1	28.1	0.1	<0.001
f (Hz)	121.2	452.1	273.4	86.0	230.5	291.3	327.4	96.9	0.629

AID indicates angular insertion depth; C1, electrode contact number 1; LID, linear insertion depth; f, frequency at the most apical electrode contact; IQR, interquartile range; Q1, first quartile; Q3, third quartile; p(SW), p value for the Shapiro-Wilk test; SD, standard deviation.

Measurement Accuracy

Intraobserver agreement and interobserver reliability were good and excellent, as shown in Table 3.

TABLE 3.

Intra- and inter-rater reliability

Parameters	Intraclass Correlation	95% Confidence Interval		F Test With True Value 0
		Lower Bound	Upper Bound	Value	df1	df2	Sig
Single measures, intra-rater reliability
R1: CDL (OC from 0°)	0.886	0.789	0.945	24.203	22	44	0.000
R2: CDL (OC from 0°)	0.906	0.825	0.956	30.011	22	44	0.000
R3: CDL (OC from 0°)	0.819	0.679	0.911	14.597	22	44	<0.001
R1: C1, AID (°)	0.881	0.778	0.944	23.206	22	44	0.000
R2: C1, AID (°)	0.757	0.584	0.878	10.351	22	46	<0.001
R3: C1, AID (°)	0.841	0.715	0.923	16.897	22	46	<0.001
R1: C1, f (Hz)	0.874	0.767	0.941	21.902	21	44	<0.001
R2: C1, f (Hz)	0.740	0.559	0.869	9.545	22	46	<0.001
R3: C1, f (Hz)	0.812	0.667	0.907	13.928	22	46	<0.001
Average measures, inter-rater reliability
CDL (OC from 0°) (R1, R2, R3)	0.961	0.834	0.987	52.406	22	44	<0.001
C1, AID (°) (R1, R2, R3)	0.949	0.787	0.982	39.793	22	44	<0.001
C1, f (Hz) (R1, R2, R3)	0.940	0.777	0.978	31.197	22	44	<0.001

AID indicates angular insertion depth; C1, electrode contact number 1; CDL, cochlear duct length; R1, rater 1; R2, rater 2; R3, rater 3.

Comparison of Predicted Electrode Contact Positions and CBF

Three individual electrode contacts were deactivated by the audiologist because of poor sound quality and speech processing: contact 10, contact 11, and contact 12 (each 33.3%). Comparison of frequency values was possible only for electrodes that were activated. All electrode contact frequencies were significantly different between ABF and CBF (Supplemental Table 1, http://links.lww.com/MAO/C140). Figure 2A shows a comparison of the estimated electrode contact frequencies (ABF) and the center frequencies of the CBF map for each stimulation channel. Figure 2B shows the magnitude of the differences between the predicted frequencies (ABF) and CBF settings. The comparison of mean predicted frequencies and the average filter band settings (i.e., upper, lower, and center frequencies of each band) for the CBF maps is illustrated in Figure 3. When the dataset was manually analyzed, 84% (n = 233) of the predicted electrode contact frequencies appeared to be outside their allocated frequency band in the clinical fitting map.

FIG. 2.

A, Comparison between frequencies of electrode contacts in ABF estimated based on preoperative imaging data and CBF. B, Magnitude of the differences between ABF and CBF in Hz for each electrode contact.

FIG. 3.

Comparison between average frequency band parameters in CBF and the average simulated ABF center frequencies based on preoperative estimation of electrode contact locations.

DISCUSSION

The aim of this retrospective study was to investigate whether there are differences between estimated electrode contact frequencies based on preoperative CT scans and CBF. In addition, the goal was to quantify the extent of the potential differences. Based on the results, we can conclude that preoperative estimation of electrode contact locations indeed yields frequency values that are significantly different from the standard frequency band settings in MAESTRO. In addition, these differences show a large individual variance. Therefore, we may assume that contact frequency estimates derived from preoperative imaging data could also be used for creating ABF maps.

In our study, we compared electrode contact frequency estimates based on preoperative imaging—that could be potentially used for ABF—with CBF frequency parameters as applied in the fitting software in clinical routine. The data for analysis are from the period when the choice of electrode array was subjective, based on the surgeon's experience and preference. We consider that not only variation in cochlea size and shape but also surgery-related factors, such as type of electrode array, insertion length, and surgical approach, can affect frequency-to-place matching. It must be emphasized that only electrode insertions via the round window can yield valid electrode location estimates in OTOPLAN. Moreover, this approach is based on the assumption that the electrode array is fully inserted, and it has to be verified carefully that this is indeed the case. In the present study, measured cochlea parameters were similar to results previously described in the literature. The coronal basal turn diameter (A value), which predicts CDL, was estimated by OTOPLAN in the studies cited to be between 7.4 and 10.2 mm and the CDL between 29 and 41.4 mm (9). In the present study, the estimations confirm these observations with a mean A value of 9.14 mm (SD, 0.48; minimum, 8.23; maximum, 10.12) and CDL of 35.27 mm (SD, 2.21; minimum, 30.96; maximum, 39.93).

When analyzing the difference between predicted frequencies and actual clinical (CBF) frequency maps, we observed indeed a significant discrepancy and the estimated frequencies one would use for ABF deviated widely from the clinically set filter bank parameters for each electrode channel. No electrode contact was lying inside the frequency range of CBF. The frequency to place mismatch was statistically significant for all electrode contacts. In the study by Di Maro et al. (10), the subjectively set frequencies of electrode contacts are higher compared with ABF. Another study showed that the frequency set subjectively in comparison to ABF differs in one subject depending on the number of electrode contacts (12). In our study, contact frequencies predicted by ABF were higher than set clinically.

Even before the launch of OTOPLAN, Jiam at al. (17) used flat-panel CT from an angiography system and modified the Greenwoods function to create an individual frequency map for 17 CI users and 23 electrodes. He noticed that 83% predicted contact frequencies (216 out of 260) needed electrode contact adjustment. In our estimations, 84% of electrode contacts in the test group may benefit from an adjustment in pitch mapping. One aspect to consider when using estimated frequencies from preoperative data is the fact that OTOPLAN tends to underestimate the CDL from preoperative data compared with postoperative analysis.

Recent studies have demonstrated that OTOPLAN's preoperative electrode estimation tends to overestimate (7,18,19) or even underestimate the angular insertion depth of the most apical contact C1 compared with actual postoperative data (19), when analyzed using high-resolution imaging.

The mean error reaches 39 degrees, which is attributed not only to inaccuracies in the equations employed but also to additional factors such as imaging artifacts and variations in surgical outcomes (7,18,19). Nevertheless, the authors conclude that the tool is reasonably reliable for determining electrode insertion depth, leaving it ultimately up to the surgeon to decide on the criteria for electrode array selection based on these findings (18). The authors also note that it enables reliable predictions, despite minor inaccuracies that need to be considered (19).

To further reduce the error, it has been proposed to apply a correction factor (7). Specifically, a factor of 0.95 should be applied to the software's electrode visualization tool results to compensate for potential overestimations. Another approach to reduce this error is to use high-resolution imaging, as proposed in Ref. (19). This is, however, not always feasible in a clinical routine setting. Therefore, it is important to treat the preoperative frequency estimates with caution when applying them to actual ABF fitting maps.

Study Limitations

There are some technical restrictions, like imaging technique and measurement quality. In CT, the quality of the image plays a very important role, as do modality, slice thickness, and spatial resolution. It can influence the quality of measurement. Measurement accuracy substantially depends on radiological image modality, slice thickness, and correctly setting the oblique coronal plane. In this study, all subjects but one had CT images with a slice thickness of 0.6 mm available, one user presented with a slice thickness of 1 mm. Similar to the literature, we observed good to excellent intra- and inter-rater reliability results (18,20–22). Here it should be noted that starting from version 4.0, OTOPLAN is capable to detect and measure the cochlear anatomy and CDL automatically, given the appropriate resolution of the CT scans.

To our knowledge, this is the first study to use OTOPLAN to simulate a frequency map using preoperative CT scans to assess discrepancy between predicted and default frequencies of electrode contacts. It is necessary to conduct a study that compares the results of the simulation with the postoperative position of contacts.

CONCLUSIONS

The retrospective study examined the potential differences between the frequency band settings of the clinical fitting map and electrode contact locations estimated from preoperative imaging data. Setting the CI filter bank frequencies based on preoperative electrode location estimates could be a useful alternative when postoperative CT or DVT imaging is not feasible or unavailable. This can serve as a good starting point when trying to reduce frequency-to-place mismatch between frequency allocations of the fitting filter bank and the tonotopy of the cochlea. Further studies are necessary to verify if there is agreement between data from preoperative imaging and data from postoperative imaging.