Related Experiment Video
Updated: Jun 7, 2025
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
1.7K
Human identification via digital palatal scans: a machine learning validation pilot study.
Ákos Mikolicz1, Botond Simon2, Aida Roudgari1,3
1Department of Restorative Dentistry and Endodontics, Semmelweis University, Budapest, Hungary.
BMC Oral Health
|November 15, 2024
Summary
Machine learning algorithms using palatal intraoral scans show high accuracy for identification, proving reliable across different populations. Sex determination using this method, however, has moderate reliability.
Area of Science:
- Forensic Science
- Biometrics
- Machine Learning
Background:
- Validation of a previously developed machine learning algorithm on a new population.
- Evaluation of discrimination potential using palatal intraoral scan-based geometric and superimposition methods.
- Study registered on ClinicalTrials.gov (NCT05349942).
Purpose of the Study:
- To validate a machine learning algorithm for palatal identification using geometric and superimposition methods.
- To assess the reliability and population independence of these methods.
- To investigate the influence of geographical factors and sex on palatal measurements.
Main Methods:
- Palatal scans obtained from 23 participants from diverse countries using the Emerald intraoral scanner.
- Geometric-based method: Measurement of palatal vault dimensions (height, width, depth) and input into Fisher's linear discriminant equations.
- Superimposition method: Comparison of scan repeatability with between-subjects differences using mean absolute differences (MAD); multiple linear regression analysis for influencing factors.
Main Results:
- Geometric-based method achieved 91.2% sensitivity and 97.1% specificity, consistent with training set results.
- Latitude and longitude did not significantly affect geometric-based matches; superimposition MAD ranges did not overlap with repeatability ranges.
- Sex determination function showed 69.0% sensitivity but decreased specificity (62.5%).
Conclusions:
- Palatal scan-based geometric and superimposition methods demonstrate robust reliability, independent of population.
- Sex distinction using this method has only moderate reliability.
- Significant correlation between geographical coordinates and palatal height suggests potential for large-scale origin determination studies.
