Related Experiment Video
Updated: Dec 26, 2025
09:02
Cortical Bone Assessment Using Ultrasonic Guided Waves: A Reproducibility Study in a Healthy Population
Published on: January 31, 2025
1.4K
Fully Automated Bone Age Assessment on Large-Scale Hand X-Ray Dataset
Xiaoying Pan1, Yizhe Zhao1, Hao Chen1
1School of Computer Science and Technology, Xi'an University of Posts and Telecommunications, Xi'an, Shaanxi 710121, China.
International Journal of Biomedical Imaging
|March 20, 2020
Summary
A new automated method for bone age assessment (BAA) uses deep active learning and convolutional neural networks to accurately predict bone age from X-rays, reducing observer variability in pediatric evaluations.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence
Background:
- Manual bone age assessment (BAA) is crucial for evaluating pediatric biological maturity but is time-consuming and observer-dependent.
- Automated BAA methods are needed to improve efficiency and consistency in clinical practice.
Purpose of the Study:
- To develop a fully automated bone age assessment (BAA) method using deep learning.
- To reduce the need for extensive manual data annotation in training AI models for BAA.
Main Methods:
- Utilized U-Net with deep active learning (AL) to efficiently segment hand images from X-rays, minimizing annotation burden.
- Employed transfer learning with six pre-trained Convolutional Neural Networks (CNNs) for feature extraction from hand images.
- Applied ensemble regression algorithms for the final bone age prediction.
Main Results:
- Achieved a Dice score of 0.95 for hand segmentation using the AL-enhanced U-Net model.
- The automated BAA method demonstrated comparable accuracy to state-of-the-art performance, with discrepancies of approximately 6.96 months (males) and 7.35 months (females) on the RSNA dataset.
Conclusions:
- The proposed fully automated BAA method effectively segments hand images and predicts bone age with high accuracy.
- Deep active learning significantly reduces annotation requirements for developing robust medical imaging AI models.
- This approach offers a promising alternative to manual BAA, enhancing efficiency and consistency in clinical settings.
