-
-
Notifications
You must be signed in to change notification settings - Fork 380
Inputs do not occupy the same physical space
This error will occur if you attempt to use images in different physical space in a context that requires a common physical space. For example, a brain mask passed to N4BiasFieldCorrection should occupy the same physical space as the image being corrected, or you will see an error like this
Exception caught:
itk::ExceptionObject (0x2c1a7c0)
Location: "unknown"
File: /data1/lib/build/ANTS/CentOS7/antsbin/ITKv5-install/include/ITK-5.0/itkImageToImageFilter.hxx
Line: 240
Description: itk::ERROR: N4BiasFieldCorrectionImageFilter(0x2c044e0): Inputs do not occupy the same physical space!
InputImage Origin: [-9.5968330e+01, 7.5178741e+01, -4.6519180e+01], InputImage_1 Origin: [-9.5723050e+01, 7.0265378e+01, -4.5940466e+01]
Tolerance: 8.0000000e-06
InputImage Direction: 1.0000000e+00 0.0000000e+00 0.0000000e+00
0.0000000e+00 1.0000000e+00 0.0000000e+00
0.0000000e+00 0.0000000e+00 1.0000000e+00
, InputImage_1 Direction: 9.9634532e-01 8.5416688e-02 -1.9561758e-06
8.2807286e-02 -9.6590231e-01 2.4530730e-01
-2.0951447e-02 2.4441094e-01 9.6944537e-01
Other programs can produce similar errors if the headers do not agree.
The obvious cause of this error is when the images are truly in different physical spaces. For example, if one image is in a subject's native T1 space, and another is in a group template space. If the T1 has been registered to the template, you can use antsApplyTransforms to resample the template space images into the native space, and vice versa.
Dimension mismatch between images can cause direction matrices to be wrong.
This happens when a multi-component image like a time series or vector image is used when a 3D scalar image is expected. Usually you will want to use a 3D image (for example, the average BOLD image or the average b=0 image from DWI) to do things like motion correction to the T1 image space. To warp a time series or multicomponent image, use antsApplyTransforms -d 3 and the -e option, with a 3D reference image.
In general, supplementary images such as masks or tissue probabilities must be in the same voxel space as the main input image. If the images are in the same physical space but sampled on different grids, the image can be resampled with antsApplyTransforms. For example:
antsApplyTransforms \
-d 3 \
-i mask.nii.gz \
-r t1.nii.gz \
-o maskResampled.nii.gz \
-n NearestNeighbor
will resample mask.nii.gz into the voxel space of t1.nii.gz.
ITK reads and writes a variety of file formats. Images are read into a generic ITK image object, and after processing the results are written out in the desired format. As a result, there can be precision errors in the output NIFTI header.
ThresholdImage 3 brain.nii.gz brainMask.nii.gz 1 Inf
will write brainMask.nii.gz, but its header is created from the ITK image in memory. One could (though probably shouldn't) replace ".nii.gz" in the output with ".mha", ".nrrd", or one of the other supported formats. A diff of the headers may show small differences in the qform between brain.nii.gz and brainMask.nii.gz. This can cause errors in downstream processing, for example:
Description: itk::ERROR: ImageToImageFilter(0x352ff90): Inputs do not occupy the same physical space!
InputImage Origin: [9.5322963e+01, -1.1251340e+02, -1.3474442e+02], InputImage_1 Origin: [9.5322960e+01, -1.1251340e+02, -1.3474442e+02]
Tolerance: 1.8750000e-06
You can use CopyImageHeaderInformation after processing to set headers consistently. You can also minimize problems by using a consistent reference space when warping images.
But all ANTs programs use ITK for I/O, so there is inherent loss of precision when reading and writing images. If CopyImageHeaderInformation does not work, you can use SetOrigin, and if necessary SetDirectionByMatrix, to set both image headers to a constant reference value. For example:
CopyImageHeaderInformation t1w.nii.gz mask.nii.gz mask.nii.gz 0 1 0
will normally make the headers consistent within the tolerance. But if it does not, then
SetOrigin t1w.nii.gz t1w.nii.gz -9.5968330e+01, 7.5178741e+01, -4.6519180e+01
SetOrigin mask.nii.gz mask.nii.gz -9.5968330e+01, 7.5178741e+01, -4.6519180e+01
should make the origins exactly consistent, though they are not guaranteed to be precisely (-9.5968330e+01, 7.5178741e+01, -4.6519180e+01).