Abstract
Background/objectives Sex estimation is a fundamental step in constructing the biological profile of unidentified skeletal remains, and the sacrum, although less studied than the os coxae, offers a potentially useful complementary source of evidence. This study aimed to estimate sex from the sacrum and to evaluate whether the resulting classification accuracy reflects a genuine morphological signal or largely depends on the flexibility of the classification method. Materials and methods Seven sacral osteometric parameters were measured in 300 dry sacra from Northeastern Thailand, comprising 150 males and 150 females. Discriminant function analysis, support vector machine, and random forest were applied to the six parameters showing significant sex differences and evaluated using leave-one-out cross-validation. K-means and TwoStep clustering, performed without reference to sex, were compared with documented sex. Results Six of seven parameters differed significantly between sexes. The three supervised classifiers achieved comparable accuracy, ranging from 87.3 to 90.0 percent, with no statistically significant differences among them. K-means and TwoStep clustering achieved 81.3 and 73.7 percent concordance with documented sex respectively, both above chance but below supervised accuracy. The antero-posterior diameter of sacral segment 1 and the vertical length of the auricular surface were the strongest contributors to classification. Conclusion Sacral sexual dimorphism reflects a genuine but incomplete natural structure, with part of the supervised models' accuracy depending on training with sex labels. The sacrum may serve as a useful complementary element in multi-element forensic sex estimation.
| Original language | English |
|---|---|
| Article number | 100530 |
| Journal | Translational Research in Anatomy |
| Volume | 45 |
| DOIs | |
| Publication status | Published - Nov 2026 |
Keywords
- Forensic anthropology
- Machine learning
- Sacrum
- Sex estimation
- Unsupervised learning
Fingerprint
Dive into the research topics of 'Sex estimation from sacrum osteometrics: Distinguishing intrinsic morphological signal from algorithm-driven accuracy using machine learning'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver