Difference between revisions of "Modules:MRIBiasFieldCorrection-Documentation-3.6"

From Slicer Wiki
Jump to: navigation, search
Line 27: Line 27:
 
* Ron Kikinis: Harvard Medical School SPL
 
* Ron Kikinis: Harvard Medical School SPL
 
* Contact: sylvain at csail.mit.edu
 
* Contact: sylvain at csail.mit.edu
|[[Image:MIT_NAMIC_Logo.png|thumb|280px|]]
+
|[[Image:MIT_NAMIC_Logo.png|thumb|280px|This work is part of NAMIC.]]
 
|}
 
|}
 +
 +
===Module Description===
 +
MRIBiasFieldCorrection corrects for the intensity inhomogeneity due to MRI gain field distortion.
 +
It first estimates the bias field and then 'subtract' it from the image.
 +
This module implements the so called methods 'N3' and 'N4' (see references at the bottom of this page).
  
 
{|
 
{|

Revision as of 04:23, 2 May 2010

Home < Modules:MRIBiasFieldCorrection-Documentation-3.6

Return to Slicer 3.6 Documentation

Gallery of New Features


Module Name

MRIBiasFieldCorrection

User Interface
Output


General Information

Module Type & Category

Type: CLI

Category: Filtering

Authors, Collaborators & Contact

  • Sylvain Jaume: MIT CSAIL
  • Nicolas Rannou: ISEN IRISA
  • Ron Kikinis: Harvard Medical School SPL
  • Contact: sylvain at csail.mit.edu
This work is part of NAMIC.

Module Description

MRIBiasFieldCorrection corrects for the intensity inhomogeneity due to MRI gain field distortion. It first estimates the bias field and then 'subtract' it from the image. This module implements the so called methods 'N3' and 'N4' (see references at the bottom of this page).

  • Step 1/24. Select input image:
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 1
  • Step 2/24. Select mask image:
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 2
  • Step 3/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 3
  • Step 4/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 4
  • Step 5/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 5
  • Step 6/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 6
  • Step 7/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 7
  • Step 8/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 8
  • Step 9/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 9
  • Step 10/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 10
  • Step 11/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 11
  • Step 12/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 12
  • Step 13/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 13
  • Step 14/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 14
  • Step 15/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 15
  • Step 16/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 16
  • Step 17/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 17
  • Step 18/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 18
  • Step 19/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 19
  • Step 20/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 20
  • Step 21/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 21
  • Step 22/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 22
  • Step 23/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 23
  • Step 24/24. :
    • Input Image: input image
    • Mask Image: label map providing an approximative segmentation of the ROI
    • Output Volume: corrected image
Step 24
Screenshot of the MRIBiasFieldCorrection module in Slicer3 during development. The screenshot illustrates the application of the MRI Bias Field Correction on the SPL Atlas. Note that the interface was simplified for the production version. The top left window shows the mask used to defined the ROI where the correction will be applied. The top right window shows the rendering of the 3D model created using the mask. The bottom left window shows one slice in the input image. Note the intensity inhomogeneity from the bottom left corner (dark) to the top right corner (bright). The bottom right window shows the output image after the application of the MRI Bias Field Correction module. The parameters used for the correction are shown on the left panel of the Slicer3 interface.

Usage

  • Load the input dataset (Main menu: Add Volume)
  • Create a Mask Volume using the Editor (Modules > Editor) (the Threshold tool should give a good result)
  • Select the MRIBiasFieldCorrection module (Modules > Filtering > MRIBiasFieldCorrection)
  • In the left panel, select the Input Volume
  • Select the Mask Volume
  • In the Preview Volume menu , select 'Create New Volume'
  • Do the same for the Output Volume menu
  • Modify the parameter values if desired (default values gave good results during our experiments)
  • Click on Apply.
Close-up of the above screenshot showing the image before and after correction. Note that the intensity inhomogeneity visible in the input image (left) has been corrected in the output image (right).

It took 32 min to process a 512x512x30 MRI volume on a MacBook laptop. To visualize the result:

  • Deselect the Overlay volume,
  • Select the input as Background volume and the output as Foreground volume.
  • Go to Modules > Volumes
  • Select the input volume, and write down the values for Window and Level
  • Select the output volume, and apply the same values
  • Move the lower left cursor in 'Manipulate Slice Views' between B (background) and F (foreground)

Tests

On the Dashboard, these tests verify that the module is working on various platforms:

Known bugs

Links to known bugs in the Slicer3 bug tracker


Usability issues

Follow this link to the Slicer3 bug tracker. Please select the usability issue category when browsing or contributing.

Source code & documentation

Links to the module's source code:

Source code:

Doxygen documentation:

More Information

Acknowledgment

This work is part of the National Alliance for Medical Image Computing (NAMIC), funded by the National Institutes of Health through the NIH Roadmap for Medical Research, Grant U54 EB005149 (PI: Ron Kikinis).

References for algorithms implemented in this module

References for related algorithms

  • Parametric Estimate of Intensity Inhomogeneities Applied to MRI, M. Styner, C. Brechbhuler, G. Szekely, and G. Gerig, IEEE Transactions on Medical Imaging, 19(3):153–165, Mar 2000.