BUG: Set NIFTI header xyzt_units field correctly in output #4595

cookpa · 2024-04-18T03:05:48Z

Previously, the correct xyzt code was assigned to the 'xyz_units' variable, but time gets masked out because it's only for space units, resulting in xyzt_units=2.

Fix by setting 'time_units' to NIFTI_UNITS_SEC.

This produces images with xyzt_units == 10, which is correct for ITK output (NIFTI_UNITS_MM | NIFTI_UNITS_SEC)

PR Checklist

[*] No API changes were made (or the changes have been approved)
[*] No major design changes were made (or the changes have been approved)
[*] Added test (or behavior not changed)
[*] Updated API documentation (or API not changed)
[*] Added license to new files (if any)
Added Python wrapping to new files (if any) as described in ITK Software Guide Section 9.5
Added ITK examples for all new major features (if any)

Refer to the ITK Software Guide for
further development details if necessary.

Modules/IO/NIFTI/test/itkNiftiImageIOTest14.cxx

N-Dekker · 2024-04-18T15:26:18Z

Modules/IO/NIFTI/test/itkNiftiImageIOTest14.cxx

+  typename OutputImageType::IndexType  start;
+  start[0] = 0;
+  start[1] = 0;
+  start[2] = 0;


Obviously, if the start index is all zero, it does not need to be set, explicitly. By default, a region starts at index zero.

Thanks, I took this from another test. Is it OK if I fix the original also in this PR?

ITK/Modules/IO/NIFTI/test/itkNiftiImageIOTest5.cxx

Lines 146 to 148 in 7f1c584

start[0] = 0;

start[1] = 0;

start[2] = 0;

Is it OK if I fix the original also in this PR?

I think so, yes 😃 My preference would be to fix the original in a separate commit. (Rather than squashing everything into one commit.) But it could be part of the same PR, sure. 👍

Thanks, I updated with your other feedback, will add a separate commit for the other files

Sorry if I was unclear! I meant to suggest to remove the start variable and the region.SetIndex(start) call! A default-constructed image region starts at index zero.

Right you are, I removed start also. Just testing now

I ended up not using the code to create a blank image, because I added test images to capture more failures. Now I test this by writing one of them back out and checking that the units are correct.

Modules/IO/NIFTI/test/itkNiftiImageIOTest14.cxx

Modules/IO/NIFTI/src/itkNiftiImageIO.cxx

cookpa · 2024-04-20T13:54:22Z

Quick update, I found some further problems with nonstandard units. Here are two headers with the same physical space, except one is in units of micron | milliseconds (xyzt_units = 19) and the other is in the usual mm | s (xyzt_units = 10).

 Spacing [0.4, 0.4, 0.4, 1.2]				      |	 Spacing [0.4, 0.4, 0.4, 1200]
 Origin [-0.5, -0.5, 0.5, 0]				      |	 Origin [-500, -500, 500, 0]
 Direction 							 Direction 
1 0 0 0								1 0 0 0
0 -1 0 0							0 -1 0 0
0 0 1 0								0 0 1 0
0 0 0 1								0 0 0 1

 Size : [8, 8, 4, 5]						 Size : [8, 8, 4, 5]

  Image Dimensions   : [8, 8, 4, 5]				  Image Dimensions   : [8, 8, 4, 5]
  Bounding Box       : {[-0.5 -0.5 0.5 0], [2.7 2.7 2.1 6]}   |	  Bounding Box       : {[-500 -500 500 0], [-496.8 -496.8 501
  Voxel Spacing      : [0.4, 0.4, 0.4, 1.2]		      |	  Voxel Spacing      : [0.4, 0.4, 0.4, 1200]
  Intensity Range    : [0, 0]					  Intensity Range    : [0, 0]
  Mean Intensity     : 0					  Mean Intensity     : 0
  Direction Cos Mtx. : 						  Direction Cos Mtx. : 
1 0 0 0								1 0 0 0
0 -1 0 0							0 -1 0 0
0 0 1 0								0 0 1 0
0 0 0 1								0 0 0 1

  Voxel->RAS x-form  : 						  Voxel->RAS x-form  : 
  Image Metadata: 						  Image Metadata: 
    ITK_original_direction of unsupported type N3itk6MatrixId	    ITK_original_direction of unsupported type N3itk6MatrixId
    ITK_original_spacing of unsupported type NSt3__16vectorId	    ITK_original_spacing of unsupported type NSt3__16vectorId
    ITK_sform_corrected = NO					    ITK_sform_corrected = NO
    bitpix = 64							    bitpix = 64
    cal_max = 0							    cal_max = 0
    cal_min = 0							    cal_min = 0
    datatype = 64						    datatype = 64
    dim[0] = 4							    dim[0] = 4
    dim[1] = 8							    dim[1] = 8
    dim[2] = 8							    dim[2] = 8
    dim[3] = 4							    dim[3] = 4
    dim[4] = 5							    dim[4] = 5
    dim[5] = 1							    dim[5] = 1
    dim[6] = 1							    dim[6] = 1
    dim[7] = 1							    dim[7] = 1
    dim_info = 0						    dim_info = 0
    intent_code = 0						    intent_code = 0
    intent_p1 = 0						    intent_p1 = 0
    intent_p2 = 0						    intent_p2 = 0
    intent_p3 = 0						    intent_p3 = 0
    nifti_type = 1						    nifti_type = 1
    pixdim[0] = 0						    pixdim[0] = 0
    pixdim[1] = 0.4					      |	    pixdim[1] = 400
    pixdim[2] = 0.4					      |	    pixdim[2] = 400
    pixdim[3] = 0.4					      |	    pixdim[3] = 400
    pixdim[4] = 1.2					      |	    pixdim[4] = 1200
    pixdim[5] = 0						    pixdim[5] = 0
    pixdim[6] = 0						    pixdim[6] = 0
    pixdim[7] = 0						    pixdim[7] = 0
    qfac = -1							    qfac = -1
    qform_code = 1						    qform_code = 1
    qform_code_name = NIFTI_XFORM_SCANNER_ANAT			    qform_code_name = NIFTI_XFORM_SCANNER_ANAT
    qoffset_x = 0.5					      |	    qoffset_x = 500
    qoffset_y = 0.5					      |	    qoffset_y = 500
    qoffset_z = 0.5					      |	    qoffset_z = 500
    qto_xyz of unsupported type N3itk6MatrixIfLj4ELj4EEE	    qto_xyz of unsupported type N3itk6MatrixIfLj4ELj4EEE
    quatern_b = 0						    quatern_b = 0
    quatern_c = 1						    quatern_c = 1
    quatern_d = 0						    quatern_d = 0
    scl_inter = 0						    scl_inter = 0
    scl_slope = 1						    scl_slope = 1
    sform_code = 1						    sform_code = 1
    sform_code_name = NIFTI_XFORM_SCANNER_ANAT			    sform_code_name = NIFTI_XFORM_SCANNER_ANAT
    slice_code = 0						    slice_code = 0
    slice_duration = 0						    slice_duration = 0
    slice_end = 0						    slice_end = 0
    slice_start = 0						    slice_start = 0
    srow_x = -0.4 0 0 0.5				      |	    srow_x = -400 0 -0 500
    srow_y = 0 0.4 0 0.5				      |	    srow_y = 0 400 -0 500
    srow_z = 0 0 0.4 0.5				      |	    srow_z = 0 0 400 500
    toffset = 0							    toffset = 0
    vox_offset = 352						    vox_offset = 352
    xyzt_units = 10					      |	    xyzt_units = 19

Correctly scaled:

Spacing [x,y,z]

Incorrectly scaled:

Spacing [t]
Bounding Box
Origin

To do for me:

Fix scaling where necessary
Test images in memory have equivalent spacing in 4D, bounding box, origin
Verify that the image on the right above is correctly written to disk with quantities in mm | s units (xyzt_units=10).

BUG: xyzt_units set incorrectly in output. Previously, the correct xyzt code was assigned to the 'xyz_units' variable, but time gets masked out because it's only for space units, resulting in xyzt_units=2. Fix by setting 'time_units' to NIFTI_UNITS_SEC. This produces images with xyzt_units == 10, which is correct for ITK output (NIFTI_UNITS_MM | NIFTI_UNITS_SEC) ENH: Convert the origin to mm and s. Previously, the spacing of the first three dimensions were converted to mm, but the fourth dimension (time) was not converted to s. Also, the origins of space and time did not change units on read. ENH: Use the toffset field to encode the origin in the fourth dimension.

cookpa · 2024-04-22T05:10:52Z

I think this is good to go now. I made some more changes to ensure that both space and time get scaled into mm and seconds respectively, in both the pixel dimensions and the origin fields. The xyzt_units should now be read and written correctly.

dzenanz · 2024-04-22T13:05:09Z

/azp run ITK.macOS.Python

jhlegarreta · 2024-05-03T12:57:36Z

@cookpa itkNiftiSpatialTemporalUnitsTest2 is failing on a number of sites in the dashboard:
https://open.cdash.org/index.php?project=Insight

See the +1/+2 failing indicators.

cookpa · 2024-05-03T13:08:53Z

Apologies, will look at this today

thewtex requested a review from hjmjohnson April 18, 2024 03:21

hjmjohnson approved these changes Apr 18, 2024

View reviewed changes