Researchers at the Mepco Schlenk College of Engineering have introduced an algorithm designed to infer a person’s biological sex from dental X-ray images, achieving a reported accuracy of 94 percent. The findings appear in a peer-reviewed venue within the International Journal of Biomedical Engineering and Technology, reflecting a growing interest in applying machine learning to forensic radiography and dental imaging.
Forensic practice has traditionally relied on dental records to identify remains. Yet when the objective shifts from comparing teeth to existing records to determining sex based solely on dental features, the effectiveness can vary. The new study addresses this gap by focusing on sex estimation directly from dental radiographs, offering a complementary tool for scenarios where postmortem identification must move beyond record comparison and toward biological attribute assessment.
In the methodology, researchers integrated a gradient-boosted regression trees (GBRT) workflow for image segmentation, followed by feeding the segmented data into a ResNet-50 convolutional neural network. The ResNet-50 model had been pre-trained on a dataset of about 3,000 tooth images to leverage learned radiographic patterns. When evaluated on a held-out set, the system demonstrated the ability to predict biological sex from tooth X-rays with a 94 percent accuracy rate, signaling a promising direction for computer-assisted sex estimation in forensic contexts.
The next stage of the project involves more rigorous testing with new data to gauge the method’s robustness under real-world conditions, including variations in imaging equipment, patient demographics, and imaging protocols. Researchers also plan to explore extending the technique beyond sex estimation to derive other biologically relevant attributes, such as age, from dental X-rays, potentially broadening the forensic value of dental imaging in future investigations.
The broader implication is a move toward more data-driven, image-based assessment in dental forensics, where machine learning can augment expert judgment. This approach aligns with ongoing efforts to translate radiographic features into actionable biological information while ensuring that findings are validated across diverse populations and imaging settings. In parallel with these advances, scientists are investigating natural substances and protective strategies that strengthen tooth enamel against caries and bacterial films, aiming to improve oral health outcomes as a foundation for accurate radiographic interpretation.