Related Experiment Videos
Evaluation of selected two-dimensional segmentation techniques for computed tomography quantitation of lymph nodes
J Rogowska1, K Batchelder, G S Gazelle
1Department of Radiology, Massachusetts General Hospital/Harvard Medical School, Boston 02129, USA.
Investigative Radiology
|March 1, 1996
Summary
The Sobel/watershed and interactive deformable contour algorithms accurately segment lymph nodes in computed tomography images, showing less than 5% error. These methods offer objective and efficient tools for evaluating contrast agents.
Area of Science:
- Medical imaging
- Image analysis
- Radiology
Background:
- Accurate lymph node segmentation is crucial for evaluating novel contrast agents.
- Existing methods require precise determination of nodal boundaries for quantitative analysis.
Purpose of the Study:
- To evaluate the accuracy of different 2D computer segmentation methods for lymph nodes.
- To compare manual tracing, thresholding, Sobel/watershed, and deformable contour algorithms.
Main Methods:
- Five segmentation techniques were tested on acrylic phantoms mimicking lymph nodes in fat.
- The top three methods were then applied to enhanced lymph node images in rabbits.
Main Results:
- The Sobel/watershed and interactive deformable contour algorithms achieved minimal area measurement errors (<5%) and interoperator variation (<5%) in phantoms.
- These two methods significantly outperformed simple thresholding and local criteria thresholding.
Conclusions:
- Sobel edge detection-based methods provide objective segmentation tools for lymph node-like objects in CT images.
- Both Sobel/watershed and interactive deformable contour algorithms are fast and user-friendly.