Stacking ML models achieve up to 96% accuracy for sex estimation from pelvic 3DCT in Black South Africans

A stacking ML model using 3DCT pelvic measurements achieved up to 96.1% accuracy for sex estimation in Black South Africans, outperforming traditional methods. Cranial measurements reached 94.3%. Retrospective, single-center, not externally validated.

• Retrospective study of 680 skeletal elements (400 crania, 280 pelvis) from 3DCT scans of contemporary Black South Africans.
• Stacking ensemble models achieved the highest accuracies: pelvic 86.1–96.1%, cranial 80.3–94.3% (exact figures from source).
• Single-center, retrospective design with no external validation; performance may not generalize to other populations or CT protocols.

RadPigeon summaries are original and for information only. They are not clinical advice.