Related Experiment Videos
An early vision-based snake model for ultrasound image segmentation
1Institute of Biomedical Engineering, National Taiwan Univ., Taipei, Taiwan. chung@lotus.mc.ntu.edu.tw
Ultrasound in Medicine & Biology
|March 21, 2000
Summary
This study introduces a novel algorithm for segmenting ultrasound (US) images, overcoming limitations of traditional methods. The new approach enhances image segmentation accuracy and noise immunity for clearer visualization.
Area of Science:
- Medical Imaging
- Image Processing
- Biomedical Engineering
Background:
- Classic image segmentation techniques struggle with ultrasound (US) images due to speckles and ill-defined edges.
- Effective segmentation of US images is crucial for accurate medical diagnosis and analysis.
Purpose of the Study:
- To develop a robust algorithm for segmenting general ultrasound images.
- To improve upon existing segmentation methods by addressing noise and boundary definition challenges.
Main Methods:
- A novel algorithm combining an early-vision model and a discrete-snake model was developed.
- The early-vision model suppresses speckle noise while capturing grayscale and textural edges.
- The discrete-snake model utilizes distance map peaks for improved noise immunity and convergence.
Main Results:
- The proposed algorithm demonstrated superior performance compared to the gradient vector flow (GVF) snake model.
- Segmentation accuracy was found to be comparable to manual delineation by experts.
- The new method significantly relaxes the constraint on initial contour placement.
Conclusions:
- The developed algorithm offers a significant advancement in ultrasound image segmentation.
- It provides a more accurate and noise-resilient solution for segmenting challenging US images.
- This technique has the potential to enhance diagnostic capabilities in medical imaging.