Accuracy of seven three-dimensional facial scanning systems compared with cone-beam computed tomography: a pilot...

Accurate three-dimensional (3D) reproduction of facial soft-tissue geometry is a prerequisite for planning and fabricating maxillofacial epitheses. This single-subject pilot study compared the surface agreement of seven 3D facial scanning configurations available on the Ukrainian market against a cone-beam computed tomography (CBCT)-derived reference surface, reporting trueness-related surface-deviation metrics and acquisition time. Because only one participant was examined and no repeated acquisitions were obtained, the work is exploratory and does not establish device precision (repeatability) or population-level accuracy. A prospective comparative pilot acquisition was performed on one healthy adult volunteer at rest on a single day. Three radiopaque fiducial markers were placed over low-mobility sites (glabella, both temporal regions). A CBCT volume (Planmeca ProMax 3D Mid) was segmented in 3D Slicer (v5.10.0) to yield the reference soft-tissue surface. Seven configurations were assessed: Shining 3D MetiSmile, Creality CR-Scan Raptor Pro, Revopoint Miraco Pro, RAYFace Facial Scanner, Planmeca ProMax 3D ProFace, and Qlone Dental App (iPad and iPhone). In CloudCompare (v2.14.alpha), fiducial-based Point-Pairs registration was followed by Iterative Closest Point (ICP) alignment, and cloud-to-mesh (C2M) surface-deviation distances were computed. Only descriptive statistics were used. Across the evaluated low-mobility region, the structured-light devices showed sub-millimetre surface deviation, with Shining 3D MetiSmile and Creality CR-Scan Raptor Pro showing the smallest dispersion (RMSE 0.288 and 0.328 mm). RAYFace recorded the highest proportion of points within ≤ 2 mm (98.74%) and the shortest acquisition time (5 s). Revopoint Miraco Pro and Planmeca ProMax 3D ProFace showed larger but still sub-millimetre deviations. Qlone deviated markedly (iPad and iPhone), with its upper error range approaching the fixed computation limit, so its tail is under-represented. In this single volunteer and over rigid, low-mobility regions only, structured-light scanners showed closer agreement with the CBCT-derived surface than the smartphone photogrammetry application. Differences among the leading structured-light devices were within the expected uncertainty of the reference surface and should not be over-interpreted as a device ranking. These preliminary observations require confirmation in multi-subject, repeated-measures studies before any clinical recommendation.