What Do We Know Beyond Reliability in Voxel-Based Registration? Validation of the Accuracy of Regional Voxel-Based Registration (R-VBR) Techniques for Orthognathic Surgery Analysis.

PURPOSE: Despite having excellent reproducibility, the accuracy of regional voxel-based registration (R-VBR) techniques used for postoperative orthognathic surgical analysis has not been validated. The purpose of this study was to validate the accuracy of R-VBR. METHODS: Preoperative (T0) and postoperative (T1) cone beam computed tomography (CBCT) of consecutive patients treated at a single center with nonsegmental LeFort I and bilateral sagittal split osteotomy were included. T1 CBCTs were oriented to match that of the standardized T0, and thus were assigned a known rotational transformation matrix in pitch/roll/yaw (P/R/Y), to create T1-prime (T1'). A copy of T1 (cT1) was made and was superimposed to T1' using R-VBR for 4 regions of interest (ROI): maxilla, distal mandible, right proximal mandible, and left proximal mandible, to create cT1'. The transformation matrix for each of the ROI was compared to those of T1' using paired t test and Bland-Altman analysis. RESULTS: Twenty-eight eligible subjects' CBCTs were analyzed. Mean difference between T1' and cT1' ranged from -0.08 to 0.14° (maximum 0.73°), with no statistically significant differences (P = 0.216 to 1). Mean absolute difference ranged from 0.13 to 0.31° (maximum 0.73°). Bland-Altman analysis showed good agreement between T1' and cT1', indicating excellent accuracy. CONCLUSIONS: R-VBR using the maxilla, distal mandible, and the bilateral proximal mandibular segments as ROI has excellent accuracy in terms of rotational measurements.