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

From Slicer Wiki
Jump to: navigation, search
(Created page with "<noinclude>{{documentation/versioncheck}}</noinclude> <!-- ---------------------------- --> {{documentation/{{documentation/version}}/module-header}} <!-- --------------------...")
 
(ENH: Improved the general information of BVeR module)
Line 8: Line 8:
 
{{documentation/{{documentation/version}}/module-introduction-start|{{documentation/modulename}}}}
 
{{documentation/{{documentation/version}}/module-introduction-start|{{documentation/modulename}}}}
 
{{documentation/{{documentation/version}}/module-introduction-row}}
 
{{documentation/{{documentation/version}}/module-introduction-row}}
−
Extension: [[Documentation/{{documentation/version}}/Extensions/AnomalousFilters|AnomalousFilters]]<br>
+
Extension: [[Documentation/{{documentation/version}}/Extensions/BabyBrain|Baby Brain Toolkit]]<br>
 
Webpage: http://dcm.ffclrp.usp.br/csim/<br>
 
Webpage: http://dcm.ffclrp.usp.br/csim/<br>
−
Author: Antonio Carlos da S. Senra Filho, CSIM Laboratory (University of Sao Paulo, Department of Computing and Mathematics)<br>
+
Author: Antonio Carlos da S. Senra Filho and Fabricio Henrique Simozo, CSIM Laboratory (University of Sao Paulo, Department of Computing and Mathematics)<br>
 
Contact: Antonio Carlos da S. Senra Filho, <email>acsenrafilho@usp.br</email><br>
 
Contact: Antonio Carlos da S. Senra Filho, <email>acsenrafilho@usp.br</email><br>
 
{{documentation/{{documentation/version}}/module-introduction-row}}
 
{{documentation/{{documentation/version}}/module-introduction-row}}
Line 16: Line 16:
 
|Image:CSIM-logo.png|CSIM Laboratory  
 
|Image:CSIM-logo.png|CSIM Laboratory  
 
|Image:USP-logo.png|University of Sao Paulo
 
|Image:USP-logo.png|University of Sao Paulo
−
|Image:CAPES-logo.png|CAPES Brazil
+
|Image:CNPq-logo.png|CNPq Brazil
 
}}
 
}}
 
{{documentation/{{documentation/version}}/module-introduction-end}}
 
{{documentation/{{documentation/version}}/module-introduction-end}}
Line 22: Line 22:
 
<!-- ---------------------------- -->
 
<!-- ---------------------------- -->
 
{{documentation/{{documentation/version}}/module-section|Module Description}}
 
{{documentation/{{documentation/version}}/module-section|Module Description}}
−
This module offer a simple application to the Anisotropic Anomalous Diffusion (AAD) filter, which is able to increase the image SNR and preserve fine object's details around the image space. This method was studied on MRI structural images (T1 and T2), which other imaging modalities could be properly investigated in the future.
+
This is a CLI module for Brain Volume Refinement (BVeR) tool, which is useful to correct brain segmentation errors that may appear in commonly brain extraction methods. The BVeR algorithm was extensively tested on structural MRI images (T1 and T2) of normal individuals. Further details are found at '''PAPER'''
  
 
<!-- ---------------------------- -->
 
<!-- ---------------------------- -->
 
{{documentation/{{documentation/version}}/module-section|Use Cases}}
 
{{documentation/{{documentation/version}}/module-section|Use Cases}}
−
* Use Case 1: Noise reduction as a preprocessing step for tissue segmentation
+
* Use Case 1: Segmentation outliers removal from previous brain extraction process
−
**When dealing with single voxel classification schemes running noise reduction as a preprocessing scheme will reduce the number of single misclassified voxels.
+
**It is frequent to appear segmentation errors (outliers) in commonly used brain extraction algorithms (BET, FreeSufer, BSE, AFNI, ROBEX, etc), thus the BVeR application can be helpful.
−
* Use Case 2: Preprocessing to volume rendering
+
* Use Case 2: Improving brain morphological measurement in large-scale studies
−
**Noise reduction will result in nicer looking volume renderings
+
**Due to decrease of segmentation errors, the outcome of large-scale morphological measurements can have better precision (e.g. cortical thickness and brain atrophy)
−
* Use Case 3: Noise reduction as part of image processing pipeline
 
−
**Could offer a better segmentation and classification on specific brain image analysis such as in Multiple Sclerosis lesion segmentation
 
  
 
<gallery widths="300px" perrow="3">
 
<gallery widths="300px" perrow="3">
−
Image:MRI_raw.png|Raw T1 weighted MRI Image
+
Image:T1_BET.png|A T1 weighted MRI image with the brain extracted using BET.
−
Image:MRI_AAD.png|T1 weighted MRI Image with AAD filter (q=1.2)
+
Image:T1_BET_BVeR.png|The BVeR application showing segmentation outlier corrections (brain external frontier)
 
</gallery>
 
</gallery>
  
Line 42: Line 40:
 
{{documentation/{{documentation/version}}/module-section|Panels and their use}}
 
{{documentation/{{documentation/version}}/module-section|Panels and their use}}
  
−
[[Image:aad_scalar_gui.png|thumb|380px|User Interface]]
+
[[Image:BVeR_gui.png|thumb|480px|User Interface]]
 
'''IO:'''
 
'''IO:'''
 
*'''Input Volume'''
 
*'''Input Volume'''
−
**Select the input image
+
**This is a previous brain extracted image which presents small segmentation outliers on the brain frontier (e.g. dura matter or bone marrow).
−
*'''Output Volume'''
+
*'''Updated Volume'''
−
**Set the output image file which the filters should place the final result
+
**Updated Volume with the outlier segmentation being corrected.
 +
*'''Updated Brain Mask'''
 +
**Updated brain mask obtained from the output image.
  
−
'''Diffusion Parameters:'''
+
'''Brain Volume Refinement Parameters:'''
−
*'''Conductance'''
+
*'''Neighbourhood Radius'''
−
**The conductance regulates the diffusion intensity in the neighborhood area. Choose a higher conductance if the input image has strong noise seem in the whole image space.
+
**A list of 3 values indicating the (x,y,z) size of the neighbourhood. This should large enough in order to get a consistent local statistics (values around 3 to 8). Example: a radius of (1,1,1) creates a neighbourhood of (3,3,3) in image space.
−
*'''Use Auto Conductance'''
+
*'''Convergence'''
−
**Choose if you want to use an automatic adjustment of conductance parameter. If this is checked, the inserted value is ignored and the optimization function below is used.
+
**A relative value that indicates how permissive is the algorithm to keep changing the brain borders. This counts how many voxels were changed in previous iterations and then estimate whether the total amount of changes reaches a limit (regarding the total number of voxels changed in the brain volume). High values will result in conservative outputs (fewer changes in the brain volume), on the other hand, low values will force a strong volume difference.
−
*'''Optimization Function'''
+
 
−
**A set of optimization function for automatic estimation of conductance parameter. This is helpful is you do not have an initial guess on what value is appropriate to the conductance setting. (Canny, MAD and Morphological). Please see the [http://www.insight-journal.org/browse/publication/983 Insight-Journal article] that explain each of these automatic conductance adjustment methods.
+
'''Advanced Parameters'''
−
*'''Number of Iteractions'''
+
*'''Number Of Iterations'''
−
**The number of iterations regulates the numerical simulation of the anomalous process over the image. This parameters is also related with the de-noising intensity, however it is more sensible to the noise intensity. Choose the higher number of iterations if the image presents high intensity noise which is not well treated by the conductance parameter.
+
**Maximum number of iterations. The brain mask is iteratively updated in order to vanish bigger error in the previous brain extraction result. This is an up limit threshold in order to avoid infinite loop in the brain volume correction. If the convergence level does not reach a stable result, then the number of iteration limit will stop the algorithm.
−
*'''Time Step'''
+
*'''Apply Binary Hole Filling'''
−
**The time step is a normalization parameters for the numerical simulation. The maximum value, given as default, is set to 3D images. Lower time step restrict the numerical simulation of the anomalous process.
+
**Choose if you want to fill holes in the first brain mask. This is important to not place a searching window inside the brain area. One can avoid this if a previous visual check was made in order to confirm that there are no zero values inside the brain area.
−
*'''Anomalous parameter'''
+
*'''Fill Holes Radius'''
−
**The anomalous parameter (or q value) is the generalization parameters responsible to give the anomalous process approach on the diffusion equation. See the reference paper<ref>Da S Senra Filho, A. C., Garrido Salmon, C. E., & Murta Junior, L. O. (2015). Anomalous diffusion process applied to magnetic resonance image enhancement. Physics in Medicine and Biology, 60(6), 2355–2373. doi:10.1088/0031-9155/60/6/2355</ref> to choose the appropriate q value (at moment, only tested in MRI T1 and T2 weighted images).
+
**A list of 3 values indicating the (x,y,z) size of the neighbourhood used in the binary filling hole procedure. This parameter is only used if "Apply Binary Hole Filling" option is checked. Example: a radius of 1 creates an isotropic neighbourhood of (3,3,3).
 +
*'''Selection Mode'''
 +
**The BVeR algorithm relies on the estimate of non-brain voxels that are still present in the input image. These outliers voxels are detected by a determined grey level threshold, which is iteratively updated. This option will define what is the type of the grey level threshold used in the entire process.  
 +
***Local = The threshold is calculated for each fixed neighbourhood
 +
***Global=The threshold is set by the global image mean value (zeros are not considered)  
 +
***Manual=The user defines a fixed threshold.
  
 
<!-- ---------------------------- -->
 
<!-- ---------------------------- -->
 
{{documentation/{{documentation/version}}/module-section|Similar Modules}}
 
{{documentation/{{documentation/version}}/module-section|Similar Modules}}
−
*[[Documentation/{{documentation/version}}/Modules/IADImageFilter|IAD Image Filter]]
+
N/A
−
*[[Documentation/{{documentation/version}}/Modules/GradientAnisotropicDiffusion|Gradient Anisotropic Image Filter]]
 
  
 
<!-- ---------------------------- -->
 
<!-- ---------------------------- -->
 
{{documentation/{{documentation/version}}/module-section|References}}
 
{{documentation/{{documentation/version}}/module-section|References}}
−
* Senra Filho, A.C. & Murta Junior, L. O., 2017. Automatic Conductance Estimation Methods for Anisotropic Diffusion ITK Filters. Insight-Journal. website: http://www.insight-journal.org/browse/publication/983
+
* PAPER
−
* da S Senra Filho, A.C., Garrido Salmon, C.E. & Murta Junior, L.O., 2015. Anomalous diffusion process applied to magnetic resonance image enhancement. Physics in Medicine and Biology, 60(6), pp.2355–2373. DOI: 10.1088/0031-9155/60/6/2355
 
−
* Filho, A.C. da S.S. et al., 2014. Anisotropic Anomalous Diffusion Filtering Applied to Relaxation Time Estimation in Magnetic Resonance Imaging. In Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE, pp. 3893–3896.
 
−
* Filho, A.C. da S.S., Barizon, G.C. & Junior, L.O.M., 2014. Myocardium Segmentation Improvement with Anisotropic Anomalous Diffusion Filter Applied to Cardiac Magnetic Resonance Imaging. In Annual Meeting of Computing in Cardiology.
 
  
 
<!-- ---------------------------- -->
 
<!-- ---------------------------- -->

Revision as of 14:15, 7 March 2018

Home < Documentation < Nightly < Modules < BrainVolumeRefinement


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


Introduction and Acknowledgements

Extension: Baby Brain Toolkit
Webpage: http://dcm.ffclrp.usp.br/csim/
Author: Antonio Carlos da S. Senra Filho and Fabricio Henrique Simozo, CSIM Laboratory (University of Sao Paulo, Department of Computing and Mathematics)
Contact: Antonio Carlos da S. Senra Filho, <email>acsenrafilho@usp.br</email>

CSIM Laboratory  
University of Sao Paulo  
CNPq Brazil  

Module Description

This is a CLI module for Brain Volume Refinement (BVeR) tool, which is useful to correct brain segmentation errors that may appear in commonly brain extraction methods. The BVeR algorithm was extensively tested on structural MRI images (T1 and T2) of normal individuals. Further details are found at PAPER

Use Cases

  • Use Case 1: Segmentation outliers removal from previous brain extraction process
    • It is frequent to appear segmentation errors (outliers) in commonly used brain extraction algorithms (BET, FreeSufer, BSE, AFNI, ROBEX, etc), thus the BVeR application can be helpful.
  • Use Case 2: Improving brain morphological measurement in large-scale studies
    • Due to decrease of segmentation errors, the outcome of large-scale morphological measurements can have better precision (e.g. cortical thickness and brain atrophy)

  • A T1 weighted MRI image with the brain extracted using BET.

  • The BVeR application showing segmentation outlier corrections (brain external frontier)


Panels and their use

User Interface

IO:

  • Input Volume
    • This is a previous brain extracted image which presents small segmentation outliers on the brain frontier (e.g. dura matter or bone marrow).
  • Updated Volume
    • Updated Volume with the outlier segmentation being corrected.
  • Updated Brain Mask
    • Updated brain mask obtained from the output image.

Brain Volume Refinement Parameters:

  • Neighbourhood Radius
    • A list of 3 values indicating the (x,y,z) size of the neighbourhood. This should large enough in order to get a consistent local statistics (values around 3 to 8). Example: a radius of (1,1,1) creates a neighbourhood of (3,3,3) in image space.
  • Convergence
    • A relative value that indicates how permissive is the algorithm to keep changing the brain borders. This counts how many voxels were changed in previous iterations and then estimate whether the total amount of changes reaches a limit (regarding the total number of voxels changed in the brain volume). High values will result in conservative outputs (fewer changes in the brain volume), on the other hand, low values will force a strong volume difference.

Advanced Parameters

  • Number Of Iterations
    • Maximum number of iterations. The brain mask is iteratively updated in order to vanish bigger error in the previous brain extraction result. This is an up limit threshold in order to avoid infinite loop in the brain volume correction. If the convergence level does not reach a stable result, then the number of iteration limit will stop the algorithm.
  • Apply Binary Hole Filling
    • Choose if you want to fill holes in the first brain mask. This is important to not place a searching window inside the brain area. One can avoid this if a previous visual check was made in order to confirm that there are no zero values inside the brain area.
  • Fill Holes Radius
    • A list of 3 values indicating the (x,y,z) size of the neighbourhood used in the binary filling hole procedure. This parameter is only used if "Apply Binary Hole Filling" option is checked. Example: a radius of 1 creates an isotropic neighbourhood of (3,3,3).
  • Selection Mode
    • The BVeR algorithm relies on the estimate of non-brain voxels that are still present in the input image. These outliers voxels are detected by a determined grey level threshold, which is iteratively updated. This option will define what is the type of the grey level threshold used in the entire process.
      • Local = The threshold is calculated for each fixed neighbourhood
      • Global=The threshold is set by the global image mean value (zeros are not considered)
      • Manual=The user defines a fixed threshold.

Similar Modules

N/A

References

  • PAPER

Information for Developers

Work in progress
 Section under construction.


Retrieved from "https://www.slicer.org/w/index.php?title=Documentation/Nightly/Modules/BrainVolumeRefinement&oldid=58731"