Related Experiment Video
Updated: Jun 28, 2026
07:53
Three-Dimensional Reconstruction for the Whole Lung with Early Multiple Pulmonary Nodules
Published on: October 13, 2023
Hierarchical shape statistical model for segmentation of lung fields in chest radiographs
1Digital Medical Research Center, Fudan University, Shanghai, 200032, China. Yonghong.Shi@fudan.edu.cn
Medical Image Computing and Computer-Assisted Intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention
|November 5, 2008
Summary
This study introduces hierarchical shape statistical models for improved image segmentation. By utilizing sub-population models, the method refines segmentation accuracy beyond standard Active Shape Models (ASM).
Area of Science:
- Medical image analysis
- Computer vision
- Statistical modeling
Background:
- Standard Active Shape Models (ASM) use a single population model for segmentation, which can be suboptimal for diverse sample variations.
- Testing samples may resemble specific sub-populations within the training data, necessitating more tailored statistical guidance.
Purpose of the Study:
- To develop a hierarchical shape statistical model for enhanced image segmentation.
- To improve segmentation performance by leveraging sub-population specific shape statistics.
Main Methods:
- Generated a whole-population shape model from linearly aligned samples.
- Employed affinity propagation to cluster samples into sub-populations.
- Constructed sub-population specific shape models using aligned samples within each cluster.
Main Results:
- The proposed hierarchical model guides initial segmentation with the whole-population model.
- Sub-population models are selected based on shape similarity for refinement.
- Experimental results demonstrate significant improvements in segmentation performance compared to conventional ASM.
Conclusions:
- Hierarchical shape statistical models offer a more effective approach to image segmentation.
- Tailoring models to sub-populations significantly enhances segmentation accuracy.
- The method provides a robust framework for diverse medical image segmentation tasks.
