Definition of a coordinate system for multi-modal images of the temporal bone and inner ear.

Purpose: The position and orientation of the head is maintained to be relatively similar during the CT / MR imaging process. However, the position / orientation dissimilarities present in the resulting images between patients, or between different scans of the same patient, do not allow for direct comparison of the images themselves or features / metrics extracted from them. This paper introduces a method of defining a coordinate system which is consistent between patients and modalities (CT and MR) for images of the temporal bone, using easily identifiable landmarks within the semicircular canals.

Methods: Cone Beam CT and high resolution MRI (T2) images of the temporal bone from 20 patients with no cochlear or temporal bone pathology in either modality were obtained. Four landmarks within the semicircular canals were defined that can be identified in both modalities. A coordinate system was defined using these landmarks. Reproducibility of landmark selection was assessed using intra- and inter-rater reliability (for three expert raters and two repeats of the landmark selection). Accuracy of the coordinate system was determined by comparing the coordinates of two additional landmarks in CT and MR images after their conversion to the proposed coordinate system.

Results: Intraclass Correlation Coefficients at a 95% level of confidence showed significant agreement within and between raters as well as between modalities. The differences between selections, raters, and modalities (as measured using mean, standard deviation, and maximum) were low and acceptable for clinical applications.

Conclusion: The proposed coordinate system is suited for use in images of the temporal bone and inner ear. Its multi-modal nature enables the coordinate system to be used in tasks such as image co-registration.