Musculoskeletal (MSK)AI / InformaticsResearch
YOLOv8 and YOLOv11 Near-Perfectly Estimate Sex from Hand-Wrist Radiographs
International journal of legal medicinetoday
YOLOv8 and YOLOv11 achieved near-perfect sex classification from hand-wrist radiographs of 1,800 individuals aged 8-18, with F1-scores of 0.99 and 0.98 and AUCs of 0.99 and 0.98; YOLOv5 performed poorly (F1 0.69, AUC 0.61).
- The study used a retrospective cohort of 1,800 hand-wrist radiographs from individuals aged 8-18 years, split into training, validation, and test subsets.
- YOLOv8 and YOLOv11 provided fully automated sex estimation requiring no manual annotation or region-of-interest selection, supporting forensic applications when preferred anatomical regions are unavailable.
Automated summary
RadPigeon summaries are original and for information only. They are not clinical advice.