Difference between revisions of "2010 Winter Project Week WMLS"

From NAMIC Wiki
Jump to: navigation, search
 
(One intermediate revision by the same user not shown)
Line 3: Line 3:
 
Image:itk_wmls_pipeline.png| Pipeline of WML segmentation
 
Image:itk_wmls_pipeline.png| Pipeline of WML segmentation
 
Image:UNCWMLSegmentation.png|One training dataset (T1, T2, PD, FLAIR images and wml segmentation)
 
Image:UNCWMLSegmentation.png|One training dataset (T1, T2, PD, FLAIR images and wml segmentation)
Image:WMLS_SlicerInterface.png| Slicer plug-in Interface
+
Image:itk_wmls.png| One testing dataset and segmentation result
Image:WMLS_ResultsOnFlair.png| Segmentation result
 
 
 
 
</gallery>
 
</gallery>
  
Line 17: Line 15:
  
 
<h3>Objective</h3>
 
<h3>Objective</h3>
We will continue developing and testing a Slicer module for the white matter lesion segmentation algorithm. Moreover, we will do more extensive testing by obtaining more data  within NA-MIC community.
+
We will continue developing and testing the white matter lesion segmentation algorithm implemented using ITK. The goal is to have an initial version ready by the end of the week that can be distributed within NA-MIC community for more extensive testing.
  
 
</div>
 
</div>
Line 24: Line 22:
  
 
<h3>Approach, Plan</h3>
 
<h3>Approach, Plan</h3>
We will improve the current version of WMLS Slicer module, especially the generalization performance of the training model and speed.  
+
We will develop a Slicer module for the white matter lesion segmentation algorithm. Base line results and test will be generated.  
  
 
</div>
 
</div>
Line 31: Line 29:
  
 
<h3>Progress</h3>
 
<h3>Progress</h3>
We have developed and tested a Slicer module for WML segmentation. Moreover, we have improved the speed of segmentation stage after training. We will provide the tutorial for this project meeting to present how to use our learning-based white matter lesion segmenation algorithm in Slicer 3. Subtasks implemented include: (1) a skull stripping algorithm working on T1 weighted images; (2) a fuzzy clustering algorithm for tissue segmentation; (3) a parametric model for gain field correction. All of these subtasks are implemented by using ITK. The training step uses AdaBoost and the segmenation step uses a support vector machine.  
+
Since summer project week in Boston, we have developed/implemented a WML segmentation algorithm using ITK classes. Also, we have integrated into the Slicer3. We have made the tutorial for this project meeting to present how to use our learning-based white matter lesion segmenation algorithm in Slicer 3. Subtasks implemented include: (1) a skull stripping algorithm working on T1 weighted images; (2) a fuzzy clustering algorithm for tissue segmentation; (3) a parametric model for gain field correction. All of these subtasks are implemented by using ITK. The training step uses AdaBoost and the segmenation step uses a support vector machine.  
  
 
</div>
 
</div>

Latest revision as of 19:17, 16 June 2010

Home < 2010 Winter Project Week WMLS


Key Investigators

  • UNC: Minjeong Kim, Dinggang Shen
  • GE : Xiaodong Tao, Jim Miller

Objective

We will continue developing and testing the white matter lesion segmentation algorithm implemented using ITK. The goal is to have an initial version ready by the end of the week that can be distributed within NA-MIC community for more extensive testing.

Approach, Plan

We will develop a Slicer module for the white matter lesion segmentation algorithm. Base line results and test will be generated.

Progress

Since summer project week in Boston, we have developed/implemented a WML segmentation algorithm using ITK classes. Also, we have integrated into the Slicer3. We have made the tutorial for this project meeting to present how to use our learning-based white matter lesion segmenation algorithm in Slicer 3. Subtasks implemented include: (1) a skull stripping algorithm working on T1 weighted images; (2) a fuzzy clustering algorithm for tissue segmentation; (3) a parametric model for gain field correction. All of these subtasks are implemented by using ITK. The training step uses AdaBoost and the segmenation step uses a support vector machine.

References

  • Zhiqiang Lao, Dinggang Shen, Dengfeng Liu, Abbas F Jawad, Elias R Melhem, Lenore J Launer, Nick R Bryan, Christos Davatzikos, Computer-Assisted Segmentation of White Matter Lesions in 3D MR images, Using Pattern Recognition, Academic Radiology, 15(3):300-313, March 2008.

[1]