SimpleITK, ResampleImageFilter for orthogonal views of MRI data

TanjaIvanovska · May 14, 2019, 1:44pm

Dear All,
i have the following question:
i have an MRI dataset and would like to compute the other two orthogonal projections of that dataset. If i understand correctly, ResampleImageFilter should do what i need. But i have a problem with parameter settings.

For instance, my dataset was taken in a sagittal view and its parameters are:

origin: (35.82217025756836, -281.53106689453125, -45.62310028076172)
size: (448, 448, 15)
spacing: (1.1160714626312, 1.1160714626312, 4.4000020027160645)
direction: (4.5691251730000336e-05, -0.0034908494453999993, -0.9999939059226608, 0.9999999989561548, 1.5929806359999996e-07, 4.569097404107829e-05, -2.0321856580351833e-10, -0.9999939069664997, 0.0034908494490346205)
pixel type: 16-bit unsigned integer
number of pixel components: 1

Now i want to get the coronal and axial views for this dataset to process them separately.
It is clear how to define the new size and new spacing for coronal and axial views, but i
do not understand how to get the correct origin and direction vectors. If i copy them from the original image (which is wrong, of course), the result is wrong as expected. But how to compute them?
Thank you!

Thank you.

dzenanz · May 14, 2019, 2:04pm

Welcome to the community Tanja!

To get orthogonal views, it is best to use OrientImageFilter. It does not introduce lossy resampling step, and you don’t have to worry about spacing.

blowekamp · May 14, 2019, 2:18pm

Hello TanjaIvanovska,

While SimpleITK does not have the OrientImageFilter. This ITK filter simply uses the PermuteAxesImageFilter and the FlipImageFilter to perform the desire re-orientation:

github.com

InsightSoftwareConsortium/ITK/blob/6938c77a5524ec8c94396610ee9263677643092f/Modules/Filtering/ImageGrid/include/itkOrientImageFilter.hxx#L158-L179


using PermuteFilterType = PermuteAxesImageFilter< InputImageType >;
using FlipFilterType = FlipImageFilter< InputImageType >;
using CastToOutputFilterType = CastImageFilter< InputImageType, OutputImageType >;


typename PermuteFilterType::Pointer permute = PermuteFilterType::New();
typename FlipFilterType::Pointer flip = FlipFilterType::New();
typename CastToOutputFilterType::Pointer cast = CastToOutputFilterType::New();
permute->SetInput(inputPtr);
permute->SetOrder(m_PermuteOrder);


flip->SetInput( permute->GetOutput() );
flip->SetFlipAxes(m_FlipAxes);
flip->FlipAboutOriginOff();


cast->SetInput( flip->GetOutput() );
cast->GetOutput()->SetRequestedRegion( outputPtr->GetRequestedRegion() );


// The input to the minipipeline is the input to this filter
// minipipeline. Therefore, the requested region of the minipipeline
// is the one needed by this filter.

This file has been truncated. show original

TanjaIvanovska · May 15, 2019, 10:40am

Thank you! The combination of PermuteFilter and FlipFilter worked out as expected!!

sag · December 13, 2019, 1:47pm

Hi tanja,

I need a MRI dataset for a program similar to yours, could you direct me where I can get couple of sets.

Regards,
sag

dzenanz · December 13, 2019, 2:24pm

TCIA has many datasets with MR modality.