AI Tool to Estimate Biological Sex from Skull Scans
Artificial intelligence is changing the way forensic experts estimate biological sex from skeletal remains. Traditionally, anthropologists rely on visual assessment of key skull traits such as the glabella, mastoid process, supraorbital margin, nuchal crest and mental eminence. These methods depend on human expertise and often vary in accuracy across different populations.
A recent study used a deep-learning model trained on three-dimensional cranial CT scans from an Indonesian population. Instead of focusing on just a few traits, the model analyses the entire skull structure, detecting patterns too subtle for the human eye.
The AI system achieved an accuracy of about 97%, compared with 82% for a trained human assessor. It also showed very low bias between male and female classifications. Visualisation tools revealed that the model examines both traditional sex-related traits and broader skull shape features, suggesting it identifies additional cues beyond standard human scoring.
This approach could offer faster and more consistent forensic analysis, especially when remains are incomplete or when human assessment is challenging. Training on clinical CT scans also means the model could be more practical for real-world forensic casework.
While further research is needed to validate the system across diverse populations and ensure interpretability, this study demonstrates the potential of AI to enhance accuracy, consistency and objectivity in forensic anthropology.
This content is republished from Scientific Reports (original article can be found here).
Note: Content has been adapted and edited for clarity.