Related Experiment Video
Updated: Jul 5, 2026
05:05
Four-Dimensional CT Analysis Using Sequential 3D-3D Registration
Published on: November 23, 2019
8.6K
Improved AI-Assisted Image Recognition of Cervical Spine Vertebrae Enables Motion Pattern Analysis in Dynamic X-Ray
Esther van Santbrink1,2,3, Tijmen H W Hijzelaar4, Valérie N E Schuermans1,2,3
1Department of Neurosurgery, Maastricht University Medical Centre, 6229 HX Maastricht, The Netherlands.
Bioengineering (Basel, Switzerland)
|March 28, 2026
Summary
Deep learning models can automate cervical spine motion analysis, with specific configurations (Models A and B) showing accuracy. Optimal performance requires high image contrast and sufficient range of motion for reliable results.
Area of Science:
- Biomechanics
- Medical Imaging Analysis
- Deep Learning Applications
Background:
- Cervical spine motion follows a consistent pattern.
- Current manual data analysis is time-consuming.
- Previous deep learning models showed promise but required improved segmentation accuracy.
Purpose of the Study:
- To enhance deep learning model performance for automated cervical motion analysis.
- To improve segmentation accuracy for reliable qualitative motion analysis.
Main Methods:
- Tested four nnU-Net configurations (baseline, pre-trained, with histogram equalization, pre-trained with histogram equalization).
- Evaluated segmentation using Dice Similarity Coefficient (DSC), Intersection over Union (IoU), and Hausdorff Distance (HD95).
- Assessed reliability of vertebral rotation estimation using Intraclass Correlation Coefficient (ICC) after sensitivity analyses.
Main Results:
- Models achieved mean DSC up to 0.92 and mean IoU up to 0.85.
- Intraclass Correlation Coefficient (ICC) for vertebral rotation estimation varied by model and conditions, with Models A and B showing higher reliability (up to 0.837).
- Sensitivity analyses indicated that low segmental range of motion and low-quality recordings impacted model performance.
Conclusions:
- Models A and B demonstrate accurate cervical motion pattern analysis.
- High image contrast and adequate segmental range of motion are crucial for robust model performance.
- This research advances automated qualitative motion analysis, improving accessibility.
