3D‑3D registration of labial anterior dentition: A potential method for personal identification in living individuals

Recent advancements in 3D video surveillance and dental reconstruction enable pre-acquisition of 3D labial anterior dentition models, offering forensic odontologists critical data for identification when facial obstructions impede traditional methods. However, no prior studies have validated the reliability of 3D-3D registration for exclusive use of labial anterior dentition features in dental identification. This study proposes a protocol for 3D-3D registration of labial dentition and systematically evaluates its forensic efficacy across two precision scenarios: high- (intraoral scans) and low-precision (3D facial scans). Thirty participants underwent sequential acquisition of maxillary intraoral scans (IOS-1) and 3D facial models (Face-1), with follow-up acquisitions (IOS-2, Face-2) obtained after a six-month interval to assess temporal consistency. Standardized extraction of labial crown regions of interest (ROIs) was performed using Geomagic Control X software, with subsequent calculation of root mean square (RMS) distances between matched and mismatched pairs. Relative Technical Error of Measurement (rTEM) and Mann-Whitney U tests were used to assess observer errors and group differences. In the high-precision scenario, intra- and inter-observer rTEM values were < 4.85 %, with matches (RMS: 0.07-0.19 mm) significantly lower than mismatches (0.45-1.62 mm; P < 0.001), showing no overlap. Receiver Operating Characteristic (ROC) analysis achieved 100 % accuracy at a 0.32 mm cutoff. In the low-precision scenario, rTEM values were < 8.21 %, with matches (0.26-0.63 mm) significantly lower than mismatches (0.50-1.89 mm; P < 0.001) but partial overlap. ROC analysis yielded 96.3 % sensitivity, 100 % specificity (99.7 % accuracy) at 0.63 mm. High-precision models demonstrated superior discriminatory power. The study establishes technical validity for labial dentition 3D-3D registration in living individual identification, with high-precision models demonstrating forensic-grade discriminative capability. This approach addresses critical challenges in contemporary forensic practice where conventional facial recognition is impeded by data limitations.