Repeatability, Reproducibility and Precision: Digital Dental Model Registration with Magnetic Resonance Imaging

Abstract

PURPOSE: Registration of digital dental models and cone beam computed tomography (CBCT) images provides multi-modal, 3D information for comprehensive orthodontic diagnosis, treatment planning and outcome evaluation. The purpose of our study was to compare the repeatability, reproducibility and precision of surface-based registration of a digital dental model to a magnetic resonance image (MRI). DESIGN: An observational retrospective study. SUBJECTS: CBCT, MRI, and digital dental model data were previously acquired from a singular patient for orthodontic treatment at the University of Minnesota. The patient had a full complement of teeth, bracket and metal artifact-free. METHODS: CBCT and MRI scans were segmented and saved in stereolithographic (STL) format. Manual digital dental model registration was completed with the CBCT and MRI to achieve best fit. Automatic surface-based registration was then initiated using incisal and facial surfaces of the maxillary dentition. Differences between the two registered surfaces were measured for the 1st molars and central incisors to evaluate the repeatability, reproducibility and precision of this process. Registrations were repeated 160 times, by three operators, for both the CBCT and MRI images. PRIMARY OUTCOME MEASURE: Root mean square (RMS) error in millimeters (mm) measured by 3dMD Vultus Software. Clinical acceptability was deemed as ≤ 0.5 mm RMS error. RESULTS: Repeatability of registrations between CBCT and MRI, as measured by RMS error, found no statistical significance between morning and evening (p= 0.92, p=0.86), or day of the week (p= 0.85, p= 0.50). Reproducibility of registrations between CBCT and MRI, as measured by RMS error, found a statistically significant difference between the three operators (p= 0.0001, p<.0001). The variables of dental landmark and image type had the most statistically significant influence on precision of registration (F= 472.59, F=37.78 with p-values <.0001). Overall, precision of registration for the CBCT and MRI with the digital dental model was 0.203 mm and 0.365 mm RMS error, with a statistically significant difference (p<.0001). Bland Altman-Analysis demonstrated excellent agreement when evaluating repeatability, reproducibility and precision. CONCLUSION: Using the proposed regional surface-based registration method, the results demonstrated repeatability, reproducibility, and precision of a digital dental model and an MRI image with clinically acceptable RMS errors. However, there was a statistically significant difference in overall precision based on image type; CBCT registrations exhibited higher precision. This is likely due to the MRI image quality, characterized by larger voxel size and fewer slices. Moreover, there was a statistically significant difference between selected dental landmarks, which is likely a result of differential soft tissue contrast and position of dentition. Increased quality image (decreased voxel size and increased number of slices), and optimization of the pulse sequence, image contrast, and dental landmark selection can improve precision of the surface-based registration method for MRI use in orthodontics.