Measuring 3D Cochlear Duct Length on MRI: Is It Accurate and Reliable?

Accuracy and Reliability of 3D Cochlear Length Measurements

Intraobserver accuracy of an experienced observer’s CT measurements (observer 1) with MR imaging measurements was the following: There was moderate reliability for the basal turn length, good reliability for the 2-turn length, and excellent reliability for the full-turn cochlear length. There have been limited studies evaluating MR imaging for similar objectives.^14,15 One of these studies measured the intraobserver and interobserver reliability of the A-value, while the other study used the spiral ganglion as the measurement target, and only 1 observer performed the measurements. Multiple studies have been performed with CT,³ and most measured the A-value and rarely dealt with the reproducibility of the method. Schurzig et al¹⁶ revealed the superiority of 3D measurements over spiral formulas. Likewise, Eser et al¹² evaluated 3D CT measurements and reported good-to-excellent intraobserver reliability in their study (ICC = 0.87).

Our results show that the cochlear lateral wall length measurements are accurate when done by an experienced observer. The second observer had no experience in temporal bone imaging, except for 3 years of radiology experience and was given only 1 hour of training before measurements. A short training period was chosen as per the study hypothesis, which was that even with a short training period, there would be high accuracy and reliability. This hypothesis was proved for CT-based measurements: Good-to-excellent reliability was seen between the 2 observers. A probable reason is that the bony anatomic structures can be visualized easily with high-resolution CT (Fig 3). On the contrary, the section thickness is higher in MR imaging, making it more difficult to detect anatomic bone landmarks. However, while the interobserver reliability was moderate-to-good with MR imaging, our results demonstrate high (good-to-excellent) intraobserver reliability with MR imaging.

• Download figure
• Open in new tab
• Download powerpoint

FIG 3.

CT (A) and MR imaging (B) of the cochlea of the same patient. The red ring seen in these images shows the lateral wall of the cochlea. The black arrow seen on the CT (A) image and the red arrow seen on the MR imaging (B) image indicate the outermost point of the lateral cochlear wall. The lateral cochlear wall, the target point of measurement, and other anatomic structures are summarized in a schematic view (C). Created with BioRender.com.

Although Observer 1’s mean MR imaging measurements were similar to the CT measurements, the second observer’s MR imaging measurements were longer than the CT measurements, a probable reason being the basic difference between the 2 imaging techniques.^12,14 The structure visualized on CT was the hyperdense bony cortex surrounding the cochlea, while on MR imaging, it was the hyperintense fluid in the cochlea. These may cause a partial volume artifact for both imaging modalities.¹² Therefore, the cochlea can be observed slightly shorter than its actual size with CT and is slightly longer than its actual length with MR imaging. Second, the anatomic landmarks required for measurement are probably more clearly differentiated on CT.¹⁷ For instance, the round window niche is an anatomic landmark that has been used unanimously in previous studies as the starting point for CT measurement and can be easily detected in the cochlear view.¹⁸ In this study, both observers faced difficulty in finding the round window niche because it is made of a thin bone structure and was hardly differentiated on MR imaging due to the hyperintense fluid signal of the inner ear.

The final challenge that reduces interobserver reliability is the difficulty in deciding the cochlea apical end point. In 2013, Erixon and Rask-Andersen¹⁹ included the hook segment in the measurement and found an approximately 1-mm variation in the cochlear apical turn between the 2 observers. Moreover, the measurements were taken with a highly sensitive micrometer made from the plastic casts of cochleae obtained from cadavers. Despite an increase in the interobserver reliability with this choice, we decided to exclude the hook segment in the apical turn, contrary to previous studies.

Last, it is important to consider how measurements are affected by the observer’s experience. As hypothesized, the experienced observer’s intraobserver reliability reached an ICC value of 0.94, while the inexperienced observer had an ICC of 0.86, which was still highly acceptable. Furthermore, the interobserver reliability was better for CT-based measurements (ICC = 0.90), while it was moderate (ICC = 0.69) for MR imaging due to the aforementioned reasons. These results suggest that MR imaging measurements are affected more negatively by the observer’s experience than CT measurements.

Is the Difference in Interobserver Reliability Clinically Significant?

On evaluating paired t test results across the intraobserver and interobserver reliability, we observed that the highest difference achieved in the 2 observers’ MR imaging measurements was for the full-turn cochlear length (−0.90 [SD, 1.71] mm; 95% CI, −1.31 to −0.47 mm). This result shows that there was a negligible difference (on average, 1 mm) between the 2 observers. Considering that cochlear implant lengths are currently produced at 2-mm intervals, this difference is not clinically essential in cochlear implant selection.²⁰ In a recent meta-analysis, Atalay et al³ reported a similar difference (0.61 [SD, 0.54] mm) in the organ of Corti length between people from the general population and patients with acquired hearing loss and stated that this difference was not significant with the same reasoning. However, Schurzig et al¹⁶ compared the accuracy of spiral formulas. For the same A-value, they found the organ of Corti length varying from 29.2 (SD, 2.30) mm to 43.4 (SD, 3.40) mm. Considering that 2 mm is important in this implant selection, it is far from an acceptable margin of error.^3,9,16,20 Therefore, CT measurements should preferably be considered in the selection of cochlear implants, and MR imaging may be used as a second-line alternative.

How Do the Technical Differences between CT and MR Imaging Influence the 3D Cochlear Length Measurements?

Imaging technology is improving progressively. The most commonly used imaging method for pre-cochlear implant evaluation is multidetector CT, followed by conebeam CT.³ In a recent study, 0.25-mm isotropic-voxel resolution was achieved with ultra-high-resolution CT in clinical images.²¹ However, with MR imaging, the section thickness cannot go below 0.60-mm isotropic-voxel resolution, even with 3D sequences. Although this voxel size difference does not seem very large, 3D CT images have approximately 14 times higher spatial resolution than MRI²¹; this spatial resolution difference is probably responsible for the high interobserver reliability with CT. Although significant improvements have been made recently for bone imaging with MR imaging, such as zero TE, to reduce these technical differences, CT is more useful clinically.^22,23 Nevertheless, if MR imaging sequences developed in the future can show the bone structure in detail, pre-cochlear implant imaging of patients can be performed with MR imaging only, significantly avoiding unnecessary radiation exposure. Extensive and comprehensive studies are needed to obtain a higher spatial resolution with MR imaging that allows detailed imaging of the bone structure.

Last, it is necessary to discuss the MR imaging sequences and magnets exclusively. Besides the 3D FIESTA-C and CISS sequences used frequently for imaging the internal auditory canal, cochlea, and labyrinth, 3D FSE sequences are also used.^{24⇓⇓⇓-28} The 3D T2 FSE sequences (Cube, GE Healthcare; sampling perfection with application-optimized contrasts by using different flip angle evolutions [SPACE], Siemens; driven equilibrium radiofrequency reset pulse [DRIVE]) can eliminate banding artifacts and have fewer flow and susceptibility artifacts.²⁸ Also, the use of 3T magnets provides a high signal-to-noise ratio and better spatial resolution than 1.5T, with shorter acquisition times.^24,25 However, more artifacts are encountered along with these advantages, especially with gradient recalled-echo sequences (banding and susceptibility artifacts). These artifacts may obscure fine anatomic detail of the cochlea and labyrinth.

Limitations

Our study has some limitations. Although the CT and MR imaging devices used in the study are currently the most used device technologies, ultra-high-resolution CT and 3T-magnet MR imaging devices are also available. Other researchers may repeat this study using currently available higher resolution isotropic 3D T2 sequences (Cube, SPACE, DRIVE) acquired at 3T. Studies with these new devices are likely to provide a greater interobserver reliability for MR imaging. If MR imaging scans were obtained in the cochlear view plane, the distortion-caused artifacts could be reduced, increasing the interobserver reliability. Another limitation is the difference in the level of experience between the 2 observers. However, to avoid this, we tested the effect of experience on measurements; high reliability in CT measurements showed us that the observer experience has a minimal effect on CT-based evaluations. Nevertheless, comparing MR imaging measurements of observers with comparable professional experience will help appreciate the true potential of MR imaging measurements.