# 3D palate scans for forensic ID tested on 81 twin pairs: accuracy drops across dentition stages

- Published: 2026-11-16 19:51 UTC
- Section: Life & Health › Medicine
- Tags: 3D palate scan, forensic identification, antemortem-postmortem matching, dental casts, twin cohort, RANSAC, ICP algorithm, dentition stages, University of Adelaide
- URL: https://realtech.media/news/rt-cd73c023f5006f289390d91e_3d_palate_scans/

**University of Adelaide researchers tested 3D palate scans for antemortem-postmortem matching. The method discriminated individuals, but accuracy fell as the developmental gap between records grew.**

![3D palate scans for forensic ID tested on 81 twin pairs: accuracy drops across dentition stages](https://realtech.media/media/2026/11/e4ecd2374dc19b98.jpg)
*Lee Haywood from Wollaton, Nottingham, England, Wikimedia Commons, CC BY-SA 2.0*

Researchers from the University of Adelaide evaluated three-dimensional surface scans of the human palate for antemortem-postmortem (AM-PM) matching, according to a study in the International Journal of Legal Medicine. The work used maxillary dental casts from a longitudinal twin cohort of 81 pairs, 162 individuals of Western European descent. [1] The authors report that palatal scans show discriminatory potential, but matching performance decreased across developmental transitions. [1]

Previous work focused mainly on discrete features such as palatal rugae or specific dimensions. This study compared the palate as an integrated 3D structure and examined how changes between primary, mixed and permanent dentition affect matching.

## Method and data

The participants came from a national longitudinal study conducted between 1995 and 2006 and included monozygotic and dizygotic twins. Using twins let the team test whether the method separates genetically related people. The casts were digitised with an extra-oral 3D scanner with an accuracy of 4 micrometers. [1]

To simulate postmortem data, the researchers removed dental crowns and alveolar structures and kept only the palatal vault. Full scans of permanent, mixed and primary dentitions served as AM references. Comparison used a two-stage registration framework: feature-based RANSAC for coarse alignment, then ICP for fine refinement. Performance was assessed with global alignment score, fitness (the final correspondence ratio after refinement), RMSE (mean surface distance after alignment) and rank-based identification. [1]

## Results and limitations

True matches showed significantly higher fitness than true mismatches in all comparisons, while RMSE was significantly lower only in permanent-to-permanent comparisons. Identification performance decreased as the developmental difference grew, with the lowest accuracy in permanent-to-primary comparisons. Fitness was more robust to cross-stage variation than RMSE. [1]

Within the permanent dentition database, monozygotic twins had significantly higher fitness than dizygotic twins among mismatched pairs. Perfect identification at rank 1 was achieved when both scans came from the permanent dentition, but this involved digitising the same casts twice, a controlled optimal condition rather than typical casework.

The authors note several limitations: permanent-dentition casts came from a single time point, the models were not scale-normalised, and the segmentation may not reflect irregular or fragmented palatal remains. They suggest that palatal comparison could serve as a supplementary tool for screening and candidate prioritisation. [1]

## Sources

1. [International Journal of Legal Medicine](https://doi.org/10.1007/s00414-026-04066-8)
2. [bioengineer.org](https://bioengineer.org/3d-palate-scans-aid-forensic-id-performance-drops-across-growth-stages)

---

Source: RealTech, realtech.media — https://realtech.media/news/rt-cd73c023f5006f289390d91e_3d_palate_scans/
