[ITK-users] SimpleITK - Subpixel shift detection in MRI

[Yaniv, Ziv Rafael (NIH/NLM/LHC) [C]]

Hello Sara,

As you did not give the specifics of the problem you want to address the  suggested approaches are rather generic. Assuming here that the anatomical structure of interest in the scans did not deform (not MRs of the heart in different phases of the cardiac cycle):

1. Prospective approach -

If you can accurately identify three or more corresponding points in each of the datasets you can use the LandmarkBasedTransformInitializer (https://itk.org/SimpleITKDoxygen/html/classitk_1_1simple_1_1LandmarkBasedTransformInitializerFilter.html) to estimate the transformation. Ideally these points can be localized with sub-voxel accuracy - possibly derived from markers which occupy a large number of voxels so that undersegmentation/oversegmentation has minimal impact on the point localization.

2. Retrospective approach -

Run a rigid registration between the images and see if the results have a visibly better alignment. After registration you can use a linked cursor to see if corresponding points appear closer to their expected locations.

An example of using a linked cursor with SimpleITK Python notebooks can be found here:
https://github.com/InsightSoftwareConsortium/SimpleITK-Notebooks/blob/master/67_Registration_Semiautomatic_Homework.ipynb

Another option is to resample the moving image and view the images using ITK-SNAP's linked cursor.

    Ziv


From: Sara Gh <sg.ele.eng at gmail.com<mailto:sg.ele.eng at gmail.com>>
Date: Monday, October 24, 2016 at 12:49 PM
To: "insight-users at itk.org<mailto:insight-users at itk.org>" <insight-users at itk.org<mailto:insight-users at itk.org>>
Subject: [ITK-users] SimpleITK - Subpixel shift detection in MRI

Hello,

I have three orthogonal intrasubject MRI volumes. Is there any way in SimpleITK to check if the volumes have subpixel (subvoxel) shifts or not?

Thanks,
Sara Gh

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/insight-users/attachments/20161024/bdafe420/attachment.html>