Validation of surface-based registration for genioplasty osteotomies: comparison with regional voxel-based registration.

Despite lower reported accuracy of genioplasty osteotomies, significant variation exists in the accuracy analysis techniques, without full validation. The purpose of this study was to compare the accuracy of surface-based registration (SBR) and regional voxel-based registration (R-VBR) for genioplasty osteotomies, to allow greater rigor in scrutinizing accuracy data. In a retrospective cross-sectional study of consecutive genioplasty patients at a single-center over a 5-year period, postoperative cone-beam computed tomography (CBCT) scans were collected. The CBCTs were oriented to a known pitch/roll/yaw and set as the reference, and compared with the unoriented raw volumes. Both volumes were segmented to surface models, and SBR was performed using the genioplasty segment as the mutual region of interest. Angular (°) and linear (mm) errors at the pogonion and menton were measured, and compared with those of R-VBR, with accuracy thresholds of 0.5° and 0.3 mm. In 34 eligible subjects, angular errors ranged from 0.32° to 0.64°, and linear errors ranged from 0.11 to 0.44 mm for SBR (P < 0.001). R-VBR showed greater angular errors, but smaller linear errors of 0.10 to 0.20 mm (P < 0.001). SBR and R-VBR have variable accuracy for genioplasty, with angular and linear accuracy favoring SBR and R-VBR, respectively.