ObjectiveSince the introduction of cone-beam computed tomography (CBCT) in dentistry, this technology has enabled distortion-free three-dimensional cephalometric analysis for orthodontic and orthognathic surgery diagnosis. However, CBCT is associated with significantly higher radiation exposure than traditional routine bidimensional examinations for orthodontic diagnosis, although low-dose protocols have markedly reduced radiation exposure over time.The objective of this preliminary feasibility study is to compare the accuracy and diagnostic capabilities of an already-validated three-dimensional cephalometric analysis on CBCT to those of an analysis on 3-T magnetic resonance imaging (3T-MRI) to assess whether the latter can deliver a comparable quality of information while avoiding radiation exposure.Materials and methodsIn order to test the feasibility of three-dimensional cephalometry on 3T-MRI, 18 subjects (4 male; 14 female) with mean age 37.8 ± SD 10.2, who had undergone both maxillofacial CBCT and maxillofacial 3T-MRI for various purposes within 1 month, were selected from the archive of the Department of Dentistry and Maxillofacial Surgery of Fondazione Ospedale Policlinico Maggiore, IRCCS, Milano, Italy.A three-dimensional cephalometric analysis composed of ten midsagittal and four bilateral landmarks and 24 measurements (11 angular, 13 linear) was performed on both scans using Mimics Research® v. 17.0 (NV, Technologielaan 15, 3001 Leuven, Belgium). Cephalometric analysis was performed twice by two independent orthodontists for each scan, and each orthodontist repeated the measurements 3 weeks later. Statistical analysis was performed with SPSS® 20.00 for Windows (IBM® Corporation, Sommers, NY, USA). A Bland-Altman test for each cephalometric value was performed to assess the agreement between the procedures. The intraclass correlation coefficient (ICC) was used to assess interobserver and intraobserver reliability. The coefficient of variation was used to evaluate precision.ResultsBoth procedures showed good reliability, with mean intraobserver ICCs of 0.977/0.971 for CBCT and 0.881/0.912 for MRI. The average interobserver ICCs were 0.965 for CBCT and 0.833 for MRI. A Bland-Altman analysis for the cephalometric tracing revealed a similar range of agreement between the two modalities; the bias range (mean ± SD) was − 0.25–0.66 mm (0.174 ± 0.31) for distances and − 0.41–0.54° (0.12 ± 0.33) for angles.ConclusionsWithin the main limitation of this pilot study, that is, the small sample, it is possible to state that cephalometric measurements on 3T-MRI seem to possess adequate reliability and repeatability and that they show satisfying agreement with values measured on CBCTs. An MRI examination does not expose patients to ionizing radiation and could provide an alternative to CBCT for three-dimensional cephalometrics in the future.