High diagnostic performance of dynamic flat-panel computed tomography (Cone-Beam-CT) in detecting anatomical structures of temporal bones of healthy probands and patients with ear diseases

Introduction

Standard radiological imaging methods for examining temporal bone anatomy and pathologies include conventional computed tomography (CT), Cone-Beam-CT and angiography, as well as magnetic resonance imaging (MRI). The choice of imaging technique depends on the structure to be examined. Conventional CT provides better visualization of bone structures and spatial resolution than MRI. In contrast, soft tissues such as the vestibulocochlear nerve, the cochlea or closely spaced structures of similar density are better visualized using MRI.

In recent years, conventional CT imaging has increasingly been replaced by Cone-Beam-CT, which has proven to be a promising imaging method for the temporal bone (Reimann et al., 2020).

Compared to standard CT, Cone-Beam-CT achieves a very high spatial resolution while allowing radiation exposure to be reduced in selected temporal bone imaging protocols (Bartling et al., 2006; Debeaupte et al., 2019; Bai et al., 2012; Hecker, 2025). Studies have shown that the effective dose at the petrous bone can be reduced by 40%−45% while at the same time achieving better image quality than conventional CT (Bevis et al., 2021; Conte et al., 2017; Knollmann et al., 2006). Ideally, 6%−16% of the dose required for conventional CT is needed to achieve consistently good image quality (Boubaker et al., 2025; Helal et al., 2021; Taguchi and Anno, 2000; Theunisse et al., 2015). Newer Cone-Beam-CT algorithms eliminate metallic artifacts more effectively than before, thereby optimizing the assessment of cochlea implant (CI) electrode positioning (Stidd et al., 2014; Prell et al., 2010; Selles et al., 2024). These advantages make Cone-Beam-CT a valuable tool for temporal bone diagnostics and for outcome evaluation after ear surgery.

A few studies exist comparing the image quality of Cone-Beam-CT to conventional CT, but there is limited and inconclusive data on the ability of Cone-Beam-CT to detect pathological ear disease (Reimann et al., 2020; Bartling et al., 2006; Debeaupte et al., 2019; Piergallini et al., 2018). For this purpose, this retrospective study performed at a tertiary university center aimed to investigate the image quality between healthy and diseased middle and inner ears, and to identify the differences in evaluability between various disease groups. In addition, it was analyzed the which extent Cone-Beam-CT can be utilized for position control after CI implantations.

Materials and methods

Ethical considerations

The study was approved by the Ethics Committee of Jena University Hospital on February 3, 2020 (IRB No. 2020-1652). The Ethics Committee waived the requirement for informed consent of the patients because the study of the non-interventional retrospective design. All data were analyzed anonymously.

Cone-Beam-CT inclusion and exclusion criteria

Primarily, all radiological temporal bone examinations with Cone-Beam-CT (CBCT) performed from January 2015 to December 2018 in the Department of Radiology, Jena University Hospital, Jena, Germany were included. All Cone-Beam-CT scans were performed using the same angiography unit (Artis Q, Siemens Healthineers AG, Munich, Germany). In total, 421 patients were initially registered. The listed Cone-Beam-CTs were assessed for usability. The examinations were considered usable if at least one axial, coronal or sagittal reconstruction of the Cone-Beam-CT was available.

Patients were divided into three cohorts:

• Cohort 1: normal temporal bone: this cohort consisted of all patients with bilateral normal temporal bone. These had received the Cone-Beam-CT because of another disease than ear pathologies. The normal temporal bones were used for the initial anatomical evaluation of the images and assess which normal anatomical structures could be visualized with Cone-Beam-CT (Figure 1).
• Cohort 2: temporal bones with ear pathologies: this cohort consisted of the patients with middle or inner ear diseases, except for cochlear implant (CI) patients.
• Cohort 3: temporal bones after CI surgery: This cohort was used to assess the quality of Cone-Beam-CT images with regard to their diagnostic value to evaluate the localization und the type of CI implantation (cochleostomy or through the round window).

Examinations in which the petrous bone was not at least partially depicted were excluded. Finally, 216 patients with 291 Cone-Beam-CTs were included. All relevant parameters (patient characteristics, diagnosis, primary site of investigation, type of surgery, Cone-Beam-CT specifics,) were recorded. Table 1 summarizes the cohort data and specific parameters.

Table 1. Probands' and patients' characteristics of the three cohorts.

SD = standard deviation.

Measurement of anatomical data in normal and pathological temporal bones

The temporal bones of cohort 1 and 2 were to be used to demonstrate how valuable Cone-Beam-CT is for the diagnostic imaging of anatomical landmarks. The middle and inner ear structures, that were evaluated, are summarized in Table 2. The anatomical structures were measured according to a three-stage scale for the depiction of temporal bone structures (Chuang et al., 2006): 0: anatomical structure not visible; 1: anatomical structure visible but not demarcated; 2: anatomical structure clearly defined.

Table 2. Evaluated anatomical structures.

Measurement of anatomical data from temporal bone with CI insertions

Cohort 3 was used to determine the postoperative quality of CI implantation, as well as the way of implantation in Cone-Beam-CT. The analyzed parameters are listed in Supplementary Table 1. The angular insertion depth (AID) was measured as follows:

• A line was drawn between the entry point of the CI electrode into the cochlea (either through the round window or via cochleostomy) and helicotrema.
• A second line was drawn, connecting the helicotrema to the last CI electrode contact.
• Both lines then formed the angle “α.”

Figure 2 illustrates this procedure. The AID was then calculated as follows (Pearl et al., 2014):

• less than one turn: AID = 360° – α
• more than one turn, less than 1.5 turns: AID = 360° + α
• more than one turn, less than 1.5 turns: AID = 720° – α
• more than two turns: AID = 720° + α.

Statistical analysis

Statistical analysis of the extracted and measured data was performed using SPSS (IBM, Armonk, USA), version 28. A total of 55 variables were examined and analyzed. The descriptive analysis of the metric and ordinal variables included absolute and relative frequencies, means, medians, standard deviations, ranges, and maximum and minimum values. Frequency tables were used for straightforward variables to display the distributions of absolute and relative values. Linear regressions were used to analyze the correlation between slice thickness, slice spacing, and sex and the dependent variables (anatomical structures). To identify central tendencies in the evaluation of anatomical structures between normal and pathological temporal bones, as well as within the temporal bones with ear pathologies, Kruskal-Wallis H and Mann-Whitney U-tests were applied. To assess the postoperative location of CI placement in the third cohort, sensitivity, specificity, and positive and negative predictive values were calculated. Statistical significance was defined as p ≤ 0.05, indicating a significant correlation.

Results

Cohort 1: normal temporal bones

This cohort included 106 evaluable Cone-Beam-CTs from 58 patients, of whom 37 were female (63.8%) and 21 were male (36.2%). The mean age at the time of examination was 59.9 ± 13.8 years. Temporal bone was not the primary site of investigation. The mean slice thickness was 0.91 ± 0.74 mm, with a median of 0.5 mm. Notably, 41 out of 58 Cone-Beam-CTs (70.7%) exhibited slice thickness of 0.5 mm or less. The mean image distance was 1.22 ± 0.53 mm, with a median of 1 mm. After evaluating all anatomical structures from this cohort, the depiction qualities of the anatomical structures, as shown in Supplementary Table 2, were obtained.

Cohort 2: temporal bones with pathologies

This cohort comprised 135 evaluable Cone-Beam-CTs from 108 patients, including 50 female (46.3%) and 58 male (53.7%) participants. The mean age at the time of examination was 51.1 ± 24.1 years. The temporal bone was the primary site of investigation. The mean slice thickness of the Cone-Beam-CTs was 0.48 ± 0.08 mm, with a median of 0.5 mm. The mean image distance was 0.50 ± 0.01 mm, also with a median of 0.5 mm. For comparative purposes, all diagnoses were categorized into one of four subgroups, as detailed in Supplementary Table 3. Subgroup 1, comprising a single examination (1%), included malformations; subgroup 2, with two examinations (2%), encompassed temporal bones with neoplastic diseases; subgroup 3, consisting of 47 examinations (43.3%), included inflammatory diseases; and subgroup 4, with 63 examinations (53.7%), comprised all functional impairments without radiologically detectable correlates. After evaluating all anatomical structures from this cohort, the depiction qualities listed in Supplementary Table 4 were obtained.

Cohort 3: temporal bone after CI surgery

In this cohort, 50 evaluable Cone-Beam-CT scans were collected from 50 patients, comprising 23 females (46%) and 27 males (54%). The mean age at the time of examination was 62.9 ± 17.5 years. In all cases, the primary focus of the examination was the temporal bone with an implanted CI. The Cone-Beam-CTs had a mean slice thickness of 0.5 ± 0.05 mm, with a median of 0.5 mm, and the mean image distance was 0.51 ± 0.07 mm, also with a median of 0.5 mm. After evaluating the images in from this cohort, the CI was implanted via cochleostomy in 26 patients (52%) and via the round window in 23 patients (46%). In one case (2%), the classification based on Cone-Beam-CT could not be determined. Using this procedure, the ability to detect implantation via cochleostomy in the postoperative Cone-Beam-CT had a sensitivity of 68.2%, specificity of 60.7%, positive predictive value of 57.7%, and negative predictive value of 70.8%. The detection of implantation via the round window yielded a sensitivity of 73.9%, specificity of 44.4 %, positive predictive value of 53.1 %, and negative predictive values of 66.7%, respectively.

Measurement of the AIDs yielded a mean value of 392° ± 137.13°, corresponding to slightly more than one turn through the cochlea. The median was 399°. The smallest AID was 0°, which is equivalent to incorrect positioning, and the largest was 732°.

In four cases (8%), the CI electrode traversed both the scalae, resulting in a scala change. In 37 cases (78%), the electrode was correctly positioned in the scala tympani. However, in five cases (10%), it was incorrectly placed in the scala vestibuli. In addition, in two cases (4%), the electrode was not located in the cochlea. One instance involved compression of the electrode, while the other involved misimplantation into the eustachian tube.

The influence of slice thickness and image distance

The correlation between slice thickness and depiction quality of the anatomical structures in cohort 1 with normal and cohort 2 with diseased temporal bones was determined using linear regression, as shown in Tables 3, 4. The regression coefficients were significantly negative for most anatomical structures (p < 0.01), indicating a strong association between reduced slice thickness and increased depiction quality. Notable coefficients included those for the labyrinthine segment of the facial nerve (−0.97), the oval window (−0.96) and the round window (−0.96). The exceptions with non-significant results (p ≥ 0.05) were the cochlear aqueduct, the petrous part of the internal carotid artery, the sigmoid sinus, the aditus ad antrum, the mastoid cavity, and the endolymphatic sac. In cohort 2 significant negative associations were observed only for the sigmoid sinus and the endolymphatic sac. For all other structures the relationships were not statistically significant (p ≥ 0.05).

Table 3. Linear regression of slice thickness to depiction quality of cohort 1: healthy temporal bones.

Bold values are statistically significant at the p ≤ 0.05 level.

Table 4. Linear regression of slice thickness to depiction quality of cohort 2: temporal bones with ear pathologies.

Bold values are statistically significant at the p ≤ 0.05 level.

Table 5 shows the results of the linear regression analysis of image distance to display quality for cohort 1. Significant negative regression coefficients were found for most structures, including the semicircular canals (−0.91, p < 0.01) and the helicotrema (−0.77, p < 0.01), which had the lowest regression coefficient. The malleus (p = 0.02) and the stapedius muscle (p = 0.05) also exhibited statistical significance. No significant associations were observed for the cochlear aqueduct, aditus ad antrum, mastoid cavity, endolymphatic sac, petrous part of the internal carotid artery and sigmoid sinus (p ≥ 0.05).

Table 5. Linear regression of image distance to display quality of cohort 1: healthy temporal bones.

Bold values are statistically significant at the p ≤ 0.05 level.

Table 6 contains the corresponding linear regression analyses for cohort 2 regarding image distance to display quality. Only the endolymphatic sac and the sigmoid sinus showed significant positive regression coefficients (p < 0.01). For all other anatomical structures, the relationship was not statistically significant (p ≥ 0.05).

Table 6. Linear regression of image distance to display quality of cohort 2: temporal bones with pathologies.

Bold values are statistically significant at the p ≤ 0.05 level.

The influence of gender to depiction quality

Supplementary Table 5 shows the evaluation of the correlation between depiction quality and anatomical structures in relation to gender for cohort 1, and Supplementary Table 6 shows the evaluation for cohort 2. There was no significant difference (p ≤ 0.05) between the two genders in cohort 1 and cohort 2.

The influence of the diagnosis on depiction quality

Ultimately, Table 7 compares the mean depiction quality scores for different diagnostic groups (inflammatory, functional, malformation and neoplasia). A statistical comparison using revealed a significant difference between groups for the eardrum only (p ≤ 0.01). No significant differences between diagnostic groups were found for all other anatomical structures.

Table 7. Comparison of depiction quality depending on the diagnosis in cohort 2: temporal bones with pathologies.

SD = standard deviation.

The influence of temporal bone pathology on depiction quality

The comparison of depiction quality according to temporal bone pathology, contingent on cohort 1 and cohort 2, is shown in Table 8. Highly significant differences (p ≤ 0.01) were found in the geniculate ganglion, all three segments of the facial nerve, chorda tympani, tympanic membrane, oval and round windows, cochlear aqueduct, helicotrema, semicircular canals, vestibule, vestibular aqueduct, and sigmoid sinus. No significant difference (p ≥ 0.05) was found for the mastoid cavity, pyramidal eminence, malleus, stapes, stapedial muscle, endolymphatic sac, and petrous part of the internal carotid artery. Compared to the mean values, most anatomical structures were significantly more accurately identified in cohort 2 than in cohort 1. The only exception was the sigmoid sinus, which was identified more accurately in cohort 1.

Table 8. Comparison of depiction quality depending on the cohorts 1 and 2.

SD = standard deviation. Bold values are statistically significant at the p ≤ 0.05 level.

Discussion

Current literature on Cone-Beam-CT imaging lacks comprehensive comparisons between healthy and diseased middle and inner ears, especially regarding the impact of pathology on image quality. In literature, only one previous study has examined Cone-Beam-CTs of healthy and diseased temporal bones in a direct comparison, as was performed in our study (Güldner et al., 2017). Güldner et al. (2017) studied a group of diseased (N = 89) vs. healthy (N = 115) temporal bones, based on radiological opacity in the middle ear. Regarding the possibility of excellent visualization, significant differences were only found for small bony structures, such as the incudostapedial joint and the long process of the incus or stapes (Güldner et al., 2017). The comparison between healthy and diseased temporal bones was refined in our study by dividing the diseases into the four groups: “inflammatory,” “neoplastic,” “malformation,” and “functional.” This made it possible to better distinguish between the different qualities of the disease and their associated impact on image quality. This was considered necessary because not all pathologies necessarily have a radiological correlation.

Our data showed that the diagnostic quality of the examined anatomical structures differed between healthy and diseased temporal bones. Contrary to the assumption that healthy middle and inner ears can generally be imaged better, only the sigmoid sinus was imaged with a higher quality in the cohort of healthy patients in cohort 1. The geniculate ganglion, three segments of the facial nerve, chorda tympani, tympanic membrane, oval and round windows, cochlear aqueduct, helicotrema, semicircular canals, vestibule, and vestibular aqueduct were better visualized in the diseased cohort 2. The better depiction quality in diseased temporal bones may be mainly explained by the focused imaging protocol in this cohort, including thinner slices and smaller image spacing. In addition, pathology-related changes such as opacification, altered tissue density, or bony remodeling may have increased local contrast and facilitated the identification of small anatomical structures. Specifically, inflammatory opacification or soft-tissue filling in the middle ear can delineate the borders of ossicles, facial nerve canal, and round window more clearly than in a completely air-filled healthy ear, as also observed in previous studies (Güldner et al., 2017). The most important reason for better visualization in diseased temporal bones is the dedicated high–resolution protocol in this cohort, which used significantly thinner slices (mean 0.48 vs. 0.91 mm) and smaller image spacing (mean 0.50 vs. 1.22 mm) compared to healthy bones.

Reducing slice thickness in conventional CT can be associated with increased signal noise, lower contrast between neighboring voxels, and higher effective doses to the target organ (Monnin et al., 2017; Abdulkareem et al., 2023). Therefore, a smaller slice thickness does not automatically result in a better diagnostic image quality, and potential quality gains must always be weighed against the effective dose delivered to the examined tissue. Previous studies have demonstrated dose savings in the temporal bone using Cone-Beam-CT. However, it is unclear how a reduced slice thickness would affect the diagnostic quality of Cone-Beam-CT (Bevis et al., 2021; Conte et al., 2017; Theunisse et al., 2015; Diamandis et al., 2025).

It was possible to evaluate the correlation between the slice thickness and image quality in both healthy and diseased temporal bones. The distribution of slice thickness in cohort 1 with healthy temporal bones ranged from 0.25 to 3 mm, which was broader than that in cohort 2 with pathologies. In 17 of the 24 anatomical structures in cohort 1, highly significant negative regression coefficients were observed. This indicates that the quality of the anatomical structures increased as the slice thickness decreased. This is particularly important for small structures such as nerves, muscles, and membranes. It should be noted that 0.5 mm is a standard slice thickness for examination of the temporal bone. Theoretically, the minimum slice thickness of the Cone-Beam-CT device used in this study is 0.14 mm (Siemens Healthineers, 2026). Slice thicknesses of up to 0.3–0.6 mm are also possible in MRI for temporal bone diagnostics (Pyykkö et al., 2019). Thus, the current resolution of Cone-Beam-CT is superior to that of conventional CT and MRI. At the same time, the effective dose is reduced compared to conventional CT, and time is saved compared to MRI.

In combination with the results from this study, slice thickness was identified as a factor influencing the diagnostic accuracy of Cone-Beam-CT. This influence has not been previously investigated using Cone-Beam-CT. Whether a slice thickness of less than 0.5 mm leads to even better image quality must be evaluated in further series of measurements.

Similar to slice thickness, image quality was also correlated with image spacing. Cone-Beam-CTs from healthy temporal bones in cohort 1 had an average image spacing more than twice as large (1.22 mm). Here, 15 anatomical structures showed highly significant correlations with the slice thickness. Thus, reduced image spacing leads to better image quality. Overall, the results suggest that this cohort's outcome most accurately reflects real-life conditions. The Cone-Beam-CT is known for its high spatial resolution, allowing for thinner slices while maintaining excellent diagnostic image quality (Reimann et al., 2020; Bartling et al., 2006; Tunkel et al., 2019). This observation aligns with the measured values in our study. In our study sex had no a significant impact on depiction quality; therefore, Cone-Beam-CT can be used irrespective of this factor.

The influence of the disease groups showed a highly significant difference in recording quality only for the eardrum. The eardrum in the “functional” diagnosis group was the most clearly depicted. It is a positive finding that no other structure showed a difference, which means that all structures, regardless of diagnosis, were equally well-depicted and usable for diagnostics.

Cone-Beam-CT is quick to set up, mobile, can already be used for guidance during the operation, and since the patient is under anesthesia, motion artifacts in the head area can be neglected compared to an awake patient. The implantation pathway is vital for the quality of CI implantation. When a CI electrode is inserted through the round window, which is not always successful, it is associated with a better outcome in speech comprehension (Elafandi et al., 2020; O'Connell et al., 2016). Although CIs can differ in material, construction, and shape, they were all depicted equally well in the Cone-Beam-CT scans studied, with no significant differences between the manufacturers. However, analysis of specificity and sensitivity yielded values that did not recommend this method as a control for the access pathway.

Most studies comparing Cone-Beam-CT to conventional CT demonstrate an advantage in visual ability for at least some anatomical structures in Cone-Beam-CT, with better average image quality and a lower effective dose (Conte et al., 2017; Neubauer et al., 2014; Pein et al., 2014). Even with a short acquisition time, Cone-Beam-CT achieves significantly better results in depicting anatomical structures compared to conventional CT (Reimann et al., 2020). Digital volume tomography devices, based on the principle of Cone-Beam-CT, are also capable of depicting the smallest structures in the temporal bone but with diagnostic quality scarcely better than conventional CT (Bremke et al., 2015; Dalchow et al., 2006).

One way in which Cone-Beam-CT can be used for access control has been identified. This involves a preoperative analysis of retrofacial access to the stapedius muscle during CI implantation to test CI function. If the position within the cochlea is suitable, the stapedius reflex can be elicited by stimulating the cochlea with the CI (Marquez et al., 2021; Volk et al., 2020).

Measurement of the AID as a gauge of implantation success using Cone-Beam-CT has been examined in previous studies. These studies have consistently shown that Cone-Beam-CT is suitable for distinguishing individual CI electrode contacts, thus allowing reliable measurement of AID (Cohen et al., 1996; Heutink et al., 2019; Trieger et al., 2011). The foundation for measuring AID was laid by Cohen et al. (1996), using conventional CT. Trieger et al. (2011) utilized this method in combination with the much more advanced Cone-Beam-CT and showed that this approach is particularly suitable for the postoperative position assessment of CI.

Misplacement of the CI is associated with significantly poorer hearing outcomes (O'Connell et al., 2016). Studies have already investigated postoperative visualization of the CI position in Cone-Beam-CT on temporal bones in patients after CI implantation (Pearl et al., 2014; Guberina et al., 2017; Rotter et al., 2017; Stock et al., 2021; Yamamoto et al., 2019). In all cases examined in the studies, as well in our study, it was possible to evaluate the position using Cone-Beam-CT. Because this type of positional control can be applied directly postoperatively or even intraoperatively, immediate quality control and, if necessary, correction of position, as in the two cases of misplacement outside the cochlea studied here, is possible. Although Cone-Beam-CT excellently measured the CI electrode position and AID, postoperative cochlear access route determination should be avoided.

The evaluation of the CI position within the cochlea is particularly dependent on metal artifacts. In conventional CT, such artifacts are strongly pronounced when imaging the CI, and can therefore obscure the exact localization of the CI electrodes (Bartling et al., 2006; Prell et al., 2010). Consequently, it was possible to identify each CI electrode and its position within the cochlea. This result is in line with the existing data evaluating this method for determining the CI electrode position (Connor, 2018; Lane et al., 2007). The effective dose to the cochlea can also be reduced to 16%−6% compared to conventional CT (Theunisse et al., 2015).

A key limitation of this study was the evaluation by one radiology-trained person, whereas comparable studies used multiple investigators (Chuang et al., 2006; Güldner et al., 2017). Due to the retrospective approach, the diagnostic distribution of diseased temporal bones was predetermined, limiting conclusions regarding disease quality correlation. In healthy temporal bones, Cone-Beam-CTs only imaged the temporal bone as an additional region, not the primary focus, affecting the image quality. It should be noted that the neoplasia and malformation groups contained very few cases (N = 2 or N = 4 for various structures), which limits the statistical power and general ability of comparisons involving these groups. Considerable variability in depiction quality within each diagnostic category was indicated by substantial standard deviations observed for many structures.

Overall, Cone-Beam-CT appears to be a valuable and accurate imaging modality for temporal bone diagnostics, offering excellent spatial resolution and reduced radiation exposure compared to conventional CT. This combination of high-resolution osseous imaging with low radiation dose compared to conventional CT underscores its clinical utility for routine temporal bone evaluation. Further research with larger sample sizes and multiple reviewers is warranted to validate these findings and optimize Cone-Beam-CT protocols for specific clinical applications in otology.

Conclusions

In conclusion, this study demonstrates that Cone-Beam-CT provides high diagnostic performance for imaging the temporal bone, with several advantages over conventional CT. The image quality was generally better for diseased compared to healthy temporal bones, particularly for small anatomical structures. Reducing slice thickness and image spacing improved visualization of most anatomical landmarks. Cone-Beam-CT also showed utility for post-operative evaluation of cochlear implant positioning, though with some limitations in determining the exact implantation approach.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fauot.2026.1857060/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Bryan Kevin Ward, Johns Hopkins University, United States

Reviewed by: Pedro Marques, University of Porto, Portugal

Navendra Jha, Viraat Ramayan Institute of Medical Sciences, India