Related Experiment Video
Updated: May 1, 2026
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
996
Multi-atlas segmentation with robust label transfer and label fusion
Information Processing in Medical Imaging : Proceedings of the ... Conference
|April 2, 2014
Summary
This study introduces improved multi-atlas segmentation for medical imaging by enhancing label transfer and fusion techniques. The methods reduce errors from image registration, improving segmentation accuracy in 3D transesophageal echocardiography (TEE).
Area of Science:
- Medical Image Analysis
- Computational Anatomy
Background:
- Multi-atlas segmentation is crucial in medical imaging, using registration to transfer labels from atlases to target images.
- Registration errors can degrade segmentation quality, necessitating robust label fusion methods.
Purpose of the Study:
- To enhance registration-based label transfer by generating multiple warped atlases via composed registration paths.
- To improve label fusion performance against registration errors by integrating probabilistic models with joint label fusion.
Main Methods:
- A novel label transfer scheme using composed registration paths and atlas selection guided by segmentations.
- Integration of a probabilistic correspondence model with joint label fusion for enhanced error reduction.
Main Results:
- The proposed label transfer scheme effectively addresses cumulative registration errors through atlas selection.
- The integrated label fusion technique significantly improves performance against registration inaccuracies.
Conclusions:
- The developed techniques enhance the accuracy and robustness of multi-atlas segmentation in medical image analysis.
- The methods demonstrate effectiveness, particularly for mitral-valve segmentation in 3D transesophageal echocardiography (TEE).
