DOC: Dicom doc. improvement

[jond01]

Hi, your package and docs are great!
I found some possible improvements for the dicom documentation, as suggested in this PR:

• (-1 + N) in the affine formulas -> (N - 1) for better readability
• Further explanation about slice spacing (\Delta{s})
• Small typo

[codecov]

Codecov Report

Merging #910 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #910   +/-   ##
=======================================
  Coverage   91.73%   91.73%           
=======================================
  Files          97       97           
  Lines       12359    12359           
  Branches     2177     2177           
=======================================
  Hits        11338    11338           
  Misses        684      684           
  Partials      337      337           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update bfde527...5dcdccd. Read the comment docs.

[matthew-brett]

Oh dear - I see now that it is more complicated than I thought when I wrote the original - https://stackoverflow.com/questions/14930222/how-to-calculate-space-between-dicom-slices-for-mpr - what do you think we should do?

[jond01]

I think that the method with Image Position (Patient) is more general, as some dicoms may miss the Spacing Between Slices attribute.
If both attributes appear, then filling the third column in the affine is possible with either:

1. Image Position (Patient) of multiple sorted slices
2. Spacing Between Slices + the normal n from Image Orientation (Patient).

And they are equivalent (at least are expected to be).

Method 1 is described thoroughly in the doc:
https://nipy.org/nibabel/dicom/dicom_orientation.html#d-affine-formulae
Method 2 may be helpful in some special case: you have just one slice (N=1) but still want the third column of the 4x4 affine matrix. It can be achieved with:

• The cross product of Image Orientation (Patient) = n
• Spacing Between Slices = Δs

In summary, it does not contradict anything but worth mentioning.

[matthew-brett]

Yes - I think the question is whether to add the explicit mention of the DICOM Spacing Between Slices attribute here - because it looks like it is a pretty unreliable method of getting the thing we want, which is the distance between the center of one slice and the center of the next: https://groups.google.com/forum/#!msg/comp.protocols.dicom/gyvsgAj4y6o/KOjJywTWFlAJ

[jond01]

I agree that the bold part here can be moved to some footnote:

• $\Delta{s}$ - Spacing in mm between slices, given by the Spacing ...

However, from my experience when this tag is there - it is correct, i.e. it agrees with the current method.
David A. Clunie is a reliable source, but it was 20 years ago. I think that today most of the manufacturers comply with the definition in the standard, which is exactly what we want:

Spacing Between Slices | (0018,0088) | 3 | Spacing between adjacent slices, in mm. The spacing is measured from the center-to-center of each slice, and if present shall not be negative.

From:
http://dicom.nema.org/MEDICAL/dicom/2015b/output/chtml/part03/sect_C.7.6.16.2.html#table_C.7.6.16-2

[matthew-brett]

That's reasonable - maybe with some discussion like here so people know to be a little careful.

[matthew-brett]

Totally agree about the rest of the changes though.

[matthew-brett]

Thanks for persisting. Looks good.