Related Experiment Video
Updated: Jul 18, 2026
In Vivo Quantification of Hip Arthrokinematics during Dynamic Weight-bearing Activities using Dual Fluoroscopy
Published on: July 2, 2021
Deep learning in sex estimation from knee radiographs - A proof-of-concept study utilizing the Terry Anatomical
Petteri Oura1, Juho-Antti Junno2, David Hunt3
1Department of Forensic Medicine, Faculty of Medicine, University of Helsinki, Helsinki, Finland; Forensic Medicine Unit, Finnish Institute for Health and Welfare, Helsinki, Finland; Medical Research Center, Oulu University Hospital and University of Oulu, Oulu, Finland; Research Unit of Health Sciences and Technology, Faculty of Medicine, University of Oulu, Oulu, Finland.
Artificial intelligence can now estimate sex from knee radiographs with 90.3% accuracy. This deep learning approach shows promise for forensic anthropology applications using knee joint analysis.
Area of Science:
- Forensic Anthropology
- Medical Imaging
- Artificial Intelligence
Background:
- Knee measurements are effective for metric sex estimation.
- Few studies have explored artificial intelligence (AI) for sex estimation using knee joints.
- The Terry Anatomical Collection provides a valuable dataset for research.
Purpose of the Study:
- To develop and evaluate deep learning algorithms for sex estimation from knee radiographs.
- To assess the accuracy of AI models in classifying sex based on knee joint images.
- To explore the potential of AI in forensic anthropology for sex determination.
Main Methods:
- Utilized 199 knee radiographs from 100 reconstructed cadaver knee joints.
- Employed AIDeveloper software for training, validation, and testing of neural network architectures.
- Focused on image classification for sex estimation from tibiofemoral joints.
Main Results:
- An MhNet-based deep learning model achieved the highest testing accuracy of 90.3%.
- The model demonstrated 100.0% accuracy in classifying female knee radiographs.
- The model correctly classified 78.6% of male knee radiographs.
Conclusions:
- Deep learning algorithms show significant potential for sex estimation from knee radiographs.
- AI-based methods offer a promising, non-invasive approach for forensic sex determination.
- Further research combining radiographic data and validated AI algorithms can enhance forensic anthropology tools.
