Difference between revisions of "Documentation/Nightly/Modules/PETTumorSegmentationEffect"

From Slicer Wiki
Jump to: navigation, search
Line 43: Line 43:
  
 
<!-- ---------------------------- -->
 
<!-- ---------------------------- -->
{{documentation/{{documentation/version}}/module-section|Tutorial - Segmentation Refinement}}
+
{{documentation/{{documentation/version}}/module-section|Segmentation Refinement}}
  
 +
In cases where the algorithm was not able to produce the desired segmentation with one click, the segmentation can be refined using these three options:
  
{{documentation/{{documentation/version}}/module-developerinfo}}
+
* Refinement style "none": Create a new segmentation with the same label as the previous segmentation. After selection, click approximately at the center of the new region to segment.
 +
* Refinement style "global": Choose this option in cases where the whole segmentation appears to be too large or too small. After selection, click on a point the new boundary should go through. This changes the whole boundary of the segmentation.
 +
* Refinement style "local": Choose this option in cases where most of the segmentation appears correct, but there are local segmentation errors. After selection, click on a point the new boundary should go through. This changes the boundary of the segmentation in the area of the segmentation error.
  
 +
All refinement options can be applied several times until the desired segmentation result is obtained.
  
 
<!-- ---------------------------- -->
 
<!-- ---------------------------- -->
Line 69: Line 73:
 
<!-- ---------------------------- -->
 
<!-- ---------------------------- -->
 
{{documentation/{{documentation/version}}/module-section|References}}
 
{{documentation/{{documentation/version}}/module-section|References}}
* Markus L. van Tol (2014): A Graph-Based Method for Segmentation of Tumors and Lymph Nodes in Volumetric PET Images, The University of Iowa, 2014
+
* Markus L. van Tol: A Graph-Based Method for Segmentation of Tumors and Lymph Nodes in Volumetric PET Images, The University of Iowa, 2014
 
 
  
 
<!-- ---------------------------- -->
 
<!-- ---------------------------- -->

Revision as of 01:21, 8 January 2015

Home < Documentation < Nightly < Modules < PETTumorSegmentationEffect


For the latest Slicer documentation, visit the read-the-docs.


Introduction and Acknowledgements

Acknowledgments: This work is funded in part by Quantitative Imaging to Assess Response in Cancer Therapy Trials NIH grant U01-CA140206 and Quantitative Image Informatics for Cancer Research (QIICR) NIH grant U24 CA180918.
Authors: Christian Bauer (University of Iowa), Markus van Tol (University of Iowa)
Contact: Christian Bauer, <email>christian-bauer@uiowa.edu</email>

License: Slicer License

The University of Iowa (UIowa)  
Quantitative Image Informatics for Cancer Research  
National Alliance for Medical Image Computing (NA-MIC)  

Module Description

Editor Effect for segmentation of tumors and hot lymph nodes in PET scans. Click on a lesion (tumor or hot lymph node) in a PET scan to segment it with one click. For cases where the initial segmentation does not provide the desired result, the tool provides functionality to refine the segmentation with few clicks. Options may help deal with complicated cases such as segmenting individual lymph nodes in a lymph node chain.

Screenshot

Tutorial - Basic Tumor and Lymph Node Segmentation

Step 1) Load PET scan.

Step 2) Go to the “Editor” module, select the volume loaded in Step 1 as the “Master Volume” in the “Create and Select Label Maps” drop-down menu.

Step 3) Press the “PETTumorSegmentation” button.

Step 4) Click on approximately the center of lesion (tumor or hot lymph node) to segment it. The algorithm automatically produces a segmentation, which can be inspected for quality and modified by the user if necessary (see tutorials further below).

Step 5) After segmentation of a lesion, change the label and segment the next lesion.

Segmentation Refinement

In cases where the algorithm was not able to produce the desired segmentation with one click, the segmentation can be refined using these three options:

  • Refinement style "none": Create a new segmentation with the same label as the previous segmentation. After selection, click approximately at the center of the new region to segment.
  • Refinement style "global": Choose this option in cases where the whole segmentation appears to be too large or too small. After selection, click on a point the new boundary should go through. This changes the whole boundary of the segmentation.
  • Refinement style "local": Choose this option in cases where most of the segmentation appears correct, but there are local segmentation errors. After selection, click on a point the new boundary should go through. This changes the boundary of the segmentation in the area of the segmentation error.

All refinement options can be applied several times until the desired segmentation result is obtained.

Tool Options

After changing segmentation options, click "Apply" to update the segmentation result.

PET Tumor Segmentation Editor Effect Options

Options:

  • Splitting: Cut off adjacent objects to the target via watershed or local minimum. Useful for lymph node chains.
  • Assist Centering: Move the center to the highest voxel within 7 physical units, without being on or next to other object labels. Improves consistency.
  • Sealing: Close single-voxel gaps in the object or between the object and other objects, if above the threshold. Useful for lymph node chains.
  • Allow Overwriting: Ignore other object labels.

Advanced Options:

  • Necrotic Region: Prevents cutoff from low uptake. Use if placing a center inside a necrotic region.
  • Denoise Threshold: Calculates threshold based on median-filtered image. Use only if scan is very noisy.
  • Linear Cost: Cost function below threshold is linear rather than based on region. Use only if little/no transition region in uptake.

References

  • Markus L. van Tol: A Graph-Based Method for Segmentation of Tumors and Lymph Nodes in Volumetric PET Images, The University of Iowa, 2014

Information for Developers