Difference between revisions of "Projects:RegistrationLibrary:RegLib C29"

From NAMIC Wiki
Jump to: navigation, search
(Created page with 'Back to ARRA main page <br> Back to Registration main page <br> [[Projects:RegistrationDocumentation:UseCaseInv…')
 
Line 6: Line 6:
 
=== Input ===
 
=== Input ===
 
{| style="color:#bbbbbb; " cellpadding="10" cellspacing="0" border="0"
 
{| style="color:#bbbbbb; " cellpadding="10" cellspacing="0" border="0"
|[[Image:RLib02_SPGR.png|150px|lleft|this is the fixed reference image. All images are aligned into this space]]  
+
|[[Image:RegLib_C29_thumb1.png|150px|lleft|this is the fixed reference image. All images are aligned into this space]]  
 
|[[Image:RegArrow_Affine.png|100px|lleft]]  
 
|[[Image:RegArrow_Affine.png|100px|lleft]]  
|[[Image:RLib02_FLAIR_150.png|150px|lleft|this is the moving image. The transform is calculated by matching this to the reference image]]
+
|[[Image:RegLib_C29_thumb2.png|150px|lleft|this is the T2 reference image, serves as target to the DTI baseline, but is itself aligned to the SPGR]]
 +
|[[Image:RegArrow_NonRigid.png|100px|lleft]]
 +
|[[Image:RegLib_C29_thumb3.png|150px|lleft|this is the DTI Baseline scan, to be registered with the T2]]
 +
|[[Image:RegLib_C29_thumb4.png|150px|lleft|this is the DTI tensor image, in the same orientation as the DTI Baseline]]
 
|-
 
|-
 
|fixed image/target
 
|fixed image/target
 
|
 
|
|moving image
+
|moving image 1
|-
+
|
|resampled labelmap<br>in reference space
+
|moving image 2a
|[[Image:ResampleArrow_Affine.png|100px|lleft]]
+
|moving image 2b
|[[Image:RLib02_FLAIR+LesionSeg_150.png|150px|lleft|this is a passive image to which the calculated transform is applied. It is a label-map in the same space as the moving FLAIR image]]
 
|-
 
|result
 
|resample
 
|segmentation labelmap
 
 
|}
 
|}
  
Line 27: Line 25:
  
 
===Objective / Background ===
 
===Objective / Background ===
This scenario occurs in many forms whenever we wish to align all the series from a single MRI exam/session into a common space. Alignment is necessary because the subject likely has moved in between series. As additional files we have a labelmap for the moving image we need to move along .
+
This is a classic case of a multi-sequence MRI exam we wish to spatially align to the anatomical reference scan (T1-SPGR). The scan of interest is the DTI image to be aligned for surgical planning/reference.  
  
 
=== Keywords ===
 
=== Keywords ===
MRI, brain, head, intra-subject, FLAIR, T1, defacing, masking, labelmap, segmentation
+
MRI, brain, head, intra-subject, DTI, T1, T2, non-rigid, tumor, surgical planning
  
 
===Input Data===
 
===Input Data===
*reference/fixed : T1 SPGR , 1x1x1 mm voxel size, 256 x 256 x 146, sagittal,
+
*reference/fixed : T1 SPGR , 0.5x0.5x1 mm voxel size, 512 x 512 x 176
*moving: T2 FLAIR 1.2x1.2x1.2 mm voxel size, sagittal
+
*moving 1: T2 0.5x0.5x1.5 mm voxel size, 512 x 512 x 92
*tag: segmentation labelmap obtained from above FLAIR, to be resampled with result transform
+
*moving 2: DTI  baseline: 1 x 1 x 3 mm, 256 x 256 x 41
 +
*moving 2b: 1 x 1 x 3 mm, 256 x 256 x 41 x 9 (tensor), original: DWI 256 x 256 x 41 x 32 directions
  
 
===Registration Challenges===
 
===Registration Challenges===
*the amount of misalignment is small. Subject did not leave the scanner in between the two acquisitions, but we have some head movement.
+
*The DTI sequence (EPI) contains string distortions we seek to correct via non-rigid alignment
*we know the underlying structure/anatomy did not change, but the two distinct acquisition types may contain different amounts of distortion
+
*the DTI baseline is similar in contrast to a T2, albeit at much lower resolution
*the T1 high-resolution had a "defacing" applied, i.e. part of the image containing facial features was removed to ensure anonymity. The FLAIR is lower resolution and contrast and did not need this. The sharp edges and missing information in part of the image may cause problems.
+
*we have different amounts of voxel-anisotropy
*we have one or more label-maps attached to the moving image that we also want to align.
 
*the different series have different dimensions, voxel size and field of view. Hence the choice of which image to choose as the reference becomes important. The additional image data present in one image but not the other may distract the algorithm and require masking.
 
*hi-resolution datasets may have defacing applied to one or both sets, and the defacing-masks may not be available
 
*the different series have different contrast. The T1 contains good contrast between white (WM) and gray matter (GM) , and pathology appears as hypointense. The FLAIR on the other hand shows barely any WM/GM contrast and the pathology appears very dominantly as hyperintense.
 
  
 
===Key Strategies===
 
===Key Strategies===
*'''Slicer 3.6 recommended modules:  [http://www.slicer.org/slicerWiki/index.php/Modules:BRAINSFit BrainsFit], [http://www.slicer.org/slicerWiki/index.php/Modules:RegisterImagesMultiRes-Documentation-3.6 Robust Multiresolution Affine]'''  
+
*'''Slicer 3.6.1 recommended modules:  [http://www.slicer.org/slicerWiki/index.php/Modules:BRAINSFit BrainsFit], [http://www.slicer.org/slicerWiki/index.php/Modules:RegisterImagesMultiRes-Documentation-3.6 Robust Multiresolution Affine]'''  
*we use an affine transform with 12 DOF (rather than a rigid one) to address distortion differences between the two protocols
+
*we use an 2-step approach: 1) we first co-register the T2 with the SPGR T1.  2) we register the DTI baseline to the T2 (the T1 provides insufficient similarity), 3) combine this with the Affine1 transform obtained from registering the T2 to the T1. affine transform with 12 DOF (rather than a rigid one) to address distortion differences between the two protocols; 4) resample the DTI volume with the new transform
*we choose the SPGR as the anatomical reference. Unless there are overriding reasons, always use the highest resolution image as your fixed/reference, to avoid loosing data through the registration.
+
 
*the defacing of the SPGR image introduces sharp edges that can distract the registration algorithm; we use a multi-resolution approach (initialization + 6 DOF for BrainsFit or the Robust Multires module) to avoid instability
 
*because of the contrast differences and the defacing we use '''Mutual Information''' as the cost function (default for both recommended modules)
 
  
 
=== Procedures ===
 
=== Procedures ===
Line 83: Line 76:
 
=== Registration Results===
 
=== Registration Results===
 
[[Image:RegLib_C02_Unreg_AnimGif.gif|500px|Unregistered Data + segmentation labelmap]] unregistered <br>
 
[[Image:RegLib_C02_Unreg_AnimGif.gif|500px|Unregistered Data + segmentation labelmap]] unregistered <br>
[[Image:RegLib_C02_Result_AnimGif.gif|500px|Robust Multiresolution Registration Result: FLAIR + segmentation aligned with SPGR]] : registered w. Multiresolution affine <br>
 
[[Image:RegLib_C02_AGif_BrainsFit.gif‎|500px|BrainsFit Result: FLAIR aligned with SPGR]]: registered w. BrainsFit <br>
 
 
<br>
 
<br>
'''Conclusion:''' Both Multiresolution and BrainsFit modules produce good registrations without the need for additional initialization or masking. The BrainsFit solution is the faster of the two (3 minutes vs. 15 minutes).
+
'''Conclusion:''' BrainsFit provides good registration but requires intermediate step of registering to T2.  
  
 
=== Download ===
 
=== Download ===
*DATA:[[Media:RegLib_C02_DATA.zip‎|'''Registration Library Case 02: MSBrain intra-subject multi-contrast''' <small> (Data & Solution Xforms, zip file 18 MB) </small>]]
+
*DATA:[[Media:RegLib_C29_DATA.zip‎|'''Registration Library Case 29 (MRI) '' <small> (Data & Solution Xforms, zip file 18 MB) </small>]]
*PRESETS: both recommended modules run with default settings. No presets required.
+
*PRESETS:  
 
*TUTORIALS:  
 
*TUTORIALS:  
 
**[[Media:RegLib_C02_ScreenCast.mov| Screencast movie (Quicktime, 16MB)]]
 
**[[Media:RegLib_C02_ScreenCast.mov| Screencast movie (Quicktime, 16MB)]]

Revision as of 16:14, 30 August 2010

Home < Projects:RegistrationLibrary:RegLib C29

Back to ARRA main page
Back to Registration main page
Back to Registration Use-case Inventory

v3.6.1 Slicer3-6Announcement-v1.png Slicer Registration Library Case #29: Intra-subject Brain DTI

Input

this is the fixed reference image. All images are aligned into this space lleft this is the T2 reference image, serves as target to the DTI baseline, but is itself aligned to the SPGR lleft this is the DTI Baseline scan, to be registered with the T2 this is the DTI tensor image, in the same orientation as the DTI Baseline
fixed image/target moving image 1 moving image 2a moving image 2b

Modules

Objective / Background

This is a classic case of a multi-sequence MRI exam we wish to spatially align to the anatomical reference scan (T1-SPGR). The scan of interest is the DTI image to be aligned for surgical planning/reference.

Keywords

MRI, brain, head, intra-subject, DTI, T1, T2, non-rigid, tumor, surgical planning

Input Data

  • reference/fixed : T1 SPGR , 0.5x0.5x1 mm voxel size, 512 x 512 x 176
  • moving 1: T2 0.5x0.5x1.5 mm voxel size, 512 x 512 x 92
  • moving 2: DTI baseline: 1 x 1 x 3 mm, 256 x 256 x 41
  • moving 2b: 1 x 1 x 3 mm, 256 x 256 x 41 x 9 (tensor), original: DWI 256 x 256 x 41 x 32 directions

Registration Challenges

  • The DTI sequence (EPI) contains string distortions we seek to correct via non-rigid alignment
  • the DTI baseline is similar in contrast to a T2, albeit at much lower resolution
  • we have different amounts of voxel-anisotropy

Key Strategies

  • Slicer 3.6.1 recommended modules: BrainsFit, Robust Multiresolution Affine
  • we use an 2-step approach: 1) we first co-register the T2 with the SPGR T1. 2) we register the DTI baseline to the T2 (the T1 provides insufficient similarity), 3) combine this with the Affine1 transform obtained from registering the T2 to the T1. affine transform with 12 DOF (rather than a rigid one) to address distortion differences between the two protocols; 4) resample the DTI volume with the new transform


Procedures

with BrainsFit (ca. 3 min):

  1. download example dataset
  2. load into 3DSlicer 3.6
  3. open Registration : BrainsFit module
  4. Input Parameters: set SPGR as fixed and FLAIR as moving image
  5. Registration Phases:
    1. select Initialize with CenterofHeadAlign
    2. select Include Rigid registration phase
    3. select "Include Affine registration phase"
  6. Output Settings: select "New Linear Transform" under Output Transform
  7. accept all other defaults & click apply
  8. program will automatically move FLAIR image under the result transform; also manually move the labelmap
  9. right click on either image and select Harden Transform to apply & resample
  10. save result images/scene


with MultiresolutionAffine (ca. 15 min):

  1. download example dataset
  2. load into 3DSlicer 3.6
  3. open Registration : RobustAffineMultiresolution module
  4. set SPGR as fixed and FLAIR as moving image
  5. accept all defaults & click apply
  6. go to Data module and move FLAIR and labelmap under the result transform
  7. right click on either image and select Harden Transform to apply & resample
  8. save result images/scene

for more details see the tutorial under Downloads

Registration Results

Unregistered Data + segmentation labelmap unregistered

Conclusion: BrainsFit provides good registration but requires intermediate step of registering to T2.

Download