This example demonstrates how to segment image based computed fields. It also shows how to use the results of the segmentation to generate data points to describe the boundary of an anatomical feature.
Segmentation refers to the process of partitioning an image into various regions. These regions are identified by sets of pixels with the same intensity values. Segmentation of medical images is typically used to identify the boundaries between different types of tissue and pinpoint the position of objects such as bones and organs.
The human brain is very good at segmenting images but it is quite a difficult task to automate. Many different segmentation algorithms have been developed and for a given image one approach may produce better results than another.
Cmgui implements segmentation via computed fields, using the Insight Tookit (ITK) library code. For background information on segmentation see chapter 9 of the ITK software guide. The latest version of the guide can be found at www.itk.org/ItkSoftwareGuide.pdf
Currently cmgui supports the following segmentation algorithms:
This example demonstrates the use of both methods to segment out different areas from a sample brain image
![]() |
| ![]() |
| ![]() | ||||||||||
Input image | Smoothed | Segmented image |
The images show the stages in segmenting out the left and right ventricle. Two seed points were used for the connected threshold filter, one within the left ventricle and one within the right.
![]() |
| ![]() |
| ![]() |
| ![]() |
| ![]() |
| ![]() | ||||||||||||||||||||||||||||||
Input image | Smoothed | Gradient | Speed term | Time-crossing map | Segmented image |
The images show the stages in segmenting out the right ventricle and the gray matter. Two seed points were used for the fast marching filter, one within the right ventricle and one within the left half of the gray matter.
Here the speed term is calculated using a sigmoid function, but it could also be calculated using the negative exponential exp(-x) or the reciprocal 1/(1+x).
Created by Peter Bier, July 2007.