Three-dimensional measurements of 600 first ribs from 300 documented northeastern Thai donors compared DFA, binary logistic regression, SVM and Random Forest. Accuracy ranged from 75.2% to 81.0%; Random Forest reached 81.0% and AUC .867, while logistic regression achieved 80.3% with 0% sex disparity and greater interpretability.
Key findings
- Inter-observer ICCs were .992–.999. All measures differed by sex at p
Why this matters globally
Forensic anthropology needs population-appropriate, courtroom-explainable methods. A marginal accuracy gain may matter less than group balance, transparency and reproducibility when pelvis or skull evidence is unavailable.
Thai researcher contribution
Teams from Naresuan, Thammasat and Khon Kaen universities developed a Thai-specific first-rib reference using KKU’s documented donor collection and 3D scanning.
Limitations to consider
Paired ribs are correlated; specimen-level rather than donor-level cross-validation could inflate performance and must be clarified. Northeastern documented donors with an older mean age do not represent all Thai populations, and casework remains may be more damaged than selected specimens.
Verify the original sources
HOMORead the original article↗DOI: 10.1127/homo/2015