Cannot obtain pixel spacing from dicom

[zhang-qiang-github]

Describe the bug

I can not obtain pixel spacing using pydicom.

Steps To Reproduce

The code is:

ds = pydicom.dcmread('67161575')
print(ds.PixelSpacing)

And it reports: 'FileDataset' object has no attribute 'PixelSpacing'

Your environment

module	version
platform	Windows-10-10.0.19042-SP0
Python	3.9.12 (main, Apr 4 2022, 05:22:27) [MSC v.1916 64 bit (AMD64)]
pydicom	2.3.0
gdcm	module not found
jpeg_ls	module not found
numpy	1.22.3
PIL	9.1.0
pylibjpeg	module not found
openjpeg	module not found
libjpeg	module not found

Other information

I am not sure whether this issue is relatived with my dicom file. If I open it with radiant, it looks like:

which is different with the normal dicom file.

[darcymason]

That radiant information makes it look like PixelSpacing is buried under a sequence item, which may also be nested, judging by the leading spaces. You would have to specify the overlying data elements for pydicom to get to that element.

You could perhaps try

ds.PixelMeasuresSequence[0].PixelSpacing

but it doesn't look from the radiant output that that would work either.

Is it possible to provide a de-identified version of the file? From that output it looks like it might be corrupted somehow.

[zhang-qiang-github]

Actually, I can not find the PixelMeasuresSequence

How to upload the file to github?

[zhang-qiang-github]

I have upload the file in: https://github.com/zhang-qiang-github/dicomfile

[mrbean-bremen]

The sequence is inside the SharedFunctionalGroupsSequence, as usual with EnhancedMR images. So you need something like:

ds.SharedFunctionalGroupsSequence[0].PixelMeasuresSequence[0].PixelSpacing

[zhang-qiang-github]

But, after read dicom, how can I know which method is used to obtain pixel spacing? ds.PixelSpacing or ds.SharedFunctionalGroupsSequence[0].PixelMeasuresSequence[0].PixelSpacing? Does I must try all methods to find pixel spacing?

[mrbean-bremen]

This depends on the SOP class of the image. For single-frame SOP classes (MR, CT, ...) it is always in the root, for multi-frame enhanced SOP classes (Enhanced CT, Enhanced MR, ...) it can be found in the Shared Functional Groups Sequence.
Checking for both is also always an option, e.g something like:

if "PixelSpacing" in ds:
    pixel_spacing = ds.PixelSpacing
elif "SharedFunctionalGroupsSequence" in ds:
    pixel_spacing = ds.SharedFunctionalGroupsSequence[0].PixelMeasuresSequence[0].PixelSpacing
else:
    # handle no pixel spacing

[zhang-qiang-github]

@mrbean-bremen Thank you for your kindly reply.

Is it possible the PixelSpacing appear in other tags?

[mrbean-bremen]

Is it possible the PixelSpacing appear in other tags?

You probably mean in other sequences - as far as I know, no.

[zhang-qiang-github]

I find a dicom, where ds.SharedFunctionalGroupsSequence[0] do not include PixelMeasuresSequence.

I have upload this dynamic dicom image in: https://github.com/zhang-qiang-github/dicomfile

In this image, the PixelSpacing appear in ds.PerFrameFunctionalGroupsSequence[0].PixelMeasuresSequence[0].PixelSpacing.

Is there any general method to obtian PixelSpacing?

[mrbean-bremen]

You are right, any tag may appear in PerFrameFunctionalGroupsSequence, if it is not the same for all slices. I was aware of that but thought that it is not a real case, as PixelSpacing does not change in an image. In your case, however, the same value of PixelSpacing has been written into each of the 80 items of PerFrameFunctionalGroupsSequence. This is not exactly DICOM-conform, but appearently it happens, so I stand corrected - this would be the third possible location.

In theory, you would have to check the value in each of the sequence items in this case, but I think in the case of PixelSpacing we can be reasonably sure that it is the same in all slices, so it would probably be sufficient to check only one of the items (for example the first, as you did).

I'm converting this to a discussion item, as it maybe of interest for someone else.