Back to all papers

Predicting biological sex in pediatric skeleton X-rays using artificial intelligence.

December 12, 2025pubmed logopapers

Authors

Janisch M,Scherkl M,Stranger N,Elsayed H,Singer G,Till H,Zellner M,Hržić F,Tschauner S

Affiliations (7)

  • Division of Neuroradiology, Vascular and Interventional Radiology, Department of Radiology, Medical University of Graz, Auenbruggerplatz 9, 8036, Graz, Styria, Austria.
  • Division of Pediatric Radiology, Department of Radiology, Medical University of Graz, Auenbruggerplatz 34, 8036, Graz, Styria, Austria. [email protected].
  • Division of Pediatric Radiology, Department of Radiology, Medical University of Graz, Auenbruggerplatz 34, 8036, Graz, Styria, Austria.
  • Department of Pediatric and Adolescent Surgery, Medical University of Graz, Auenbruggerplatz 34, 8036, Graz, Styria, Austria.
  • Department of Imaging, University Children's Hospital Zurich, Lenggstrasse 30, 8008, Zurich, Switzerland.
  • Center for Artificial Intelligence and Cybersecurity, University of Rijeka, Radmile Matejčić 2, 51000, Rijeka, Primorsko-Goranska, Croatia.
  • Department of Orthopaedic Surgery and Sports Medicine, Boston Children's Hospital, Harvard Medical School, 300 Longwood Ave, Boston, MA, 02445, USA.

Abstract

Artificial intelligence (AI) is increasingly applied in medical imaging, yet its ability to predict biological sex from pediatric radiographs remains unclear. This study investigates the performance of convolutional neural network (CNN) models in sex classification using a large dataset of pediatric trauma imaging and compares results with human raters. Radiographs from computed and digital radiography systems were processed to normalize grayscale and enhance contrast. The EfficientNet family of CNN models (B0-B7) was trained on this dataset, with attention to balancing the test set by age, sex, and fracture visibility. A subset of 1,000 images was independently assessed by human raters for comparison. AI models achieved a mean precision of 0.731 ± 0.035, recall of 0.718 ± 0.110, accuracy of 0.722 ± 0.032, and F1-score of 0.724 ± 0.050 across all network variants. Performance improved with age, peaking in the 13-18 group. Pelvic X-rays achieved the highest classification metrics. Human raters showed significantly lower agreement. AI can classify biological sex from pediatric radiographs with high accuracy, surpassing human performance. Results vary by age and body region, supporting the potential for AI-assisted imaging in pediatric clinical practice.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 7,100+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.