Related Experiment Videos
Segmentation of carpal bones from CT images using skeletally coupled deformable models
Thomas B Sebastian1, Hüseyin Tek, Joseph J Crisco
1LEMS, Division of Engineering, Brown University, Providence, RI 02912, USA.
Medical Image Analysis
|December 7, 2002
Summary
Accurate segmentation of carpal bones from CT scans is crucial for wrist joint analysis. A novel unified framework combining region growing and region competition effectively segments these bones, overcoming common imaging challenges.
Area of Science:
- Medical Imaging
- Biomedical Engineering
- Computational Anatomy
Background:
- Accurate segmentation of carpal bones from 3D CT images is essential for in vivo wrist joint kinematics analysis.
- Challenges include bone tissue non-uniformity, irregular bone shapes, limited inter-bone space resolution, blood vessels, and CT image blurring.
Purpose of the Study:
- To review existing segmentation methods for carpal bones.
- To propose and validate a novel unified segmentation framework for carpal bones in CT images.
Main Methods:
- A review of statistical classification, deformable models, region growing, region competition, and morphological operations was performed.
- A novel method combining curve evolution region growing with skeletally-mediated inter-region competition was developed.
- The method utilizes subpixel representations of regions and inter-region skeletons for growth coupling and competition mediation.
Main Results:
- The proposed unified framework effectively overcomes common challenges in carpal bone segmentation.
- Demonstrated success on both synthetic and real CT imaging data.
- The method integrates advantages of active contour models, region growing, and region competition.
Conclusions:
- The developed segmentation method significantly improves carpal bone segmentation accuracy in CT images.
- The approach is robust and overcomes limitations of previous methods.
- Its domain-general nature suggests applicability to diverse medical imaging segmentation tasks.