An evaluation of the use of augmented intelligence in 3D facial landmarking—an approach to facial analysis and craniofacial anomalies detection

Background: Accurate three-dimensional (3D) facial landmarking is essential for craniofacial analysis, but manual identification is time-consuming and subject to variability. Automated systems offer a solution but require robust validation and calibration. This study aims to validate a previously described automated augmented intelligence (AI) platform for 3D facial landmarking in a Caucasian population. Methods: Thirty-seven 3D facial scans from a single cohort of Caucasian adults aged 18–40 years were used. Twenty-one landmarks were manually identified on each 3D face with texture overlay. The same landmarks were automatically plotted by the AI platform (https://3dmedx.nl) using geometric data only (texture removed). Ten linear distances and angles were derived from the landmarks. Paired t-tests were used to compare manual and AI-derived measurements. Results: The mean error across all linear distances was 1.30±0.89 mm and for angles was 1.14±0.76 degrees, representing a mean percentage error of less than 2.5%. The percentage of landmarks within 2, 3, and 4 mm was 75.1%, 89.8%, and 96.7%, respectively. Statistically significant differences were found for landmarks, such as Nasion-Pronasale-Subnasale (P=0.02) and Nasion-Pronasale-Pogonion (P=0.043). Conclusions: The evaluated AI platform demonstrated good accuracy for 3D facial landmarking in a Caucasian cohort, with errors comparable to reported inter-observer variability. It shows promise as a time-saving tool for initial assessment and research, and as the technology evolves can be used in things like diagnostic facial imaging and precise surgical planning.