Related Experiment Video
Updated: Nov 8, 2025
10:23
Author Spotlight: Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
3.3K
Three-Dimensional Facial Anthropometric Analysis With and Without Landmark Labelling: Is There a Real Difference?
Daniele Gibelli1, Annalisa Cappella2,3, Filippo Bertozzi1
1LAFAS, Laboratory of Functional Anatomy of the Stomatognathic System, Department of Biomedical Sciences for Health, University of Milan.
The Journal of Craniofacial Surgery
|April 19, 2021
Summary
Landmark labeling in 3D facial models improves measurement accuracy. Unlabeled or untextured models show decreased precision, especially for small or palpable facial landmarks, impacting 3D facial analysis.
Area of Science:
- 3D facial imaging
- Anthropometry
- Biometrics
Background:
- The necessity of landmark labeling for accurate 3D facial acquisition remains a subject of debate.
- Variations in 3D facial model types (textured labeled, unlabeled, untextured) may influence measurement reliability.
Purpose of the Study:
- To compare measurement accuracy and precision across different 3D facial model types.
- To evaluate the impact of landmark labeling on the reliability of anthropometric measurements from 3D facial scans.
Main Methods:
- Acquisition of 3D facial data from 50 subjects using stereophotogrammetry.
- Extraction of landmark coordinates and calculation of 33 linear/angular measurements, surface area, and volume from textured labeled (TL), unlabeled (NL), and untextured (NTL) models.
- Assessment of measurement accuracy using relative technical error of measurement (rTEM) and evaluation of intra- and inter-observer errors.
Main Results:
- Measurement errors (intra- and inter-observer) increased from TL to NTL and NL 3D models.
- Average rTEM values were 4.5±2.6% for TL and 4.7±2.8% for NTL/NL models, with most measurements rated 'good' or 'very good'.
- Specific measurements like orbital height, mandibular ramus length, and nasal convexity showed 'moderate' or 'poor' rTEM, indicating lower reliability.
Conclusions:
- Landmark labeling is crucial for enhancing the accuracy and precision of measurements in 3D facial models.
- Reduced accuracy is observed in unlabeled and untextured 3D facial models, particularly for measurements of small magnitude or those involving landmarks requiring palpation.
