Related Experiment Video
Updated: Aug 6, 2026
06:45
Automated Joint Space Detection Improves Bone Segmentation Accuracy
Published on: November 28, 2025
Automated morphometric segmentation analysis of hand X-ray image using deep learning network
Yiyang Zhang1, Zhukai Zhuang1, Pengli Yu1
1Department of Thoracic and Cardiovascular Surgery, The Affiliated Drum Tower Hospital of Nanjing University Medical School, Kuang Yaming Honors School, Nanjing University, Nanjing, 210023, China.
BMC Medical Imaging
|July 20, 2026
Summary
This study introduces a deep learning approach for analyzing hand X-rays, accurately segmenting bones and extracting morphometric features like finger bone length and width. The automated method shows high consistency with expert measurements, paving the way for standardized clinical analysis.
Area of Science:
- Medical Image Analysis
- Deep Learning Applications
- Computer Vision in Healthcare
Background:
- Deep learning is increasingly used in medical image analysis, yet its application to hand X-ray morphometrics is underexplored.
- Existing research primarily focuses on CT and MR imaging, leaving a gap in hand X-ray analysis.
Purpose of the Study:
- To develop an efficient and accurate deep learning scheme for automatic segmentation and morphology feature extraction from hand X-ray images.
- To establish a clinically interpretable workflow for hand X-ray analysis.
Main Methods:
- Utilized a dataset of 668 hand X-ray images, split into training, validation, and testing sets.
- Employed a combination of nnU-net for region of interest segmentation and YOLO for bone-wise instance segmentation.
- Developed an in-house program to extract morphometric features such as Finger Bone Length (FBL) and Finger Bone Width (FBW).
Main Results:
- Achieved high IoU scores (0.9596 training, 0.9584 validation) with nnU-net and high accuracy (0.9816) with YOLOv8.
- Demonstrated strong consistency between automatically and manually extracted features using Pearson, Spearman, and ICC analyses (95% confidence level).
- The automated feature extraction method showed a mean difference close to zero compared to manual measurements.
Conclusions:
- Presented a two-stage workflow combining nnU-Net and YOLO for interpretable hand X-ray analysis.
- Quantitatively validated automated morphometric indicators (FBL/FBW) against expert measurements, showing consistent agreement.
- The proposed pipeline supports standardized clinical measurements and has potential utility in clinical settings.
