@safir: Thanks for reporting the problem and suggesting a solution. Could you test the following patch:

https://github.com/jcfr/Slicer/tree/3776-CheckForLabelVolumeValidity-fix-double-compare

Fixed in r23488
See http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=23488

@Steve: Thanks for reviewing.

@Safir: Thanks for contributing. After testing it works out as expected, make sure to close the issue.

Thanks

Thanks @jscfr for looking into this issue. The issue remains the same for me, with these two files: https://db.tt/YNaGQblq and https://db.tt/goyDta8k I get this "Warning: IJKToRAS mismatch at [1, 3] (45.929 != 45.929)" and "IJKToRAS mismatch at [2, 3] (176.197 != 176.197)".

/Arash

I think this problem still exist.

I've changed the format of a label map from nrrd to mha.

After opening the scene and the editor with the label map saved in mha format I receive the same warning.

What version of Slicer are you using? (is it something newer that 4.4?)

Can you share a small example dataset that demonstrates the problem?

Here is an example: http://www13.zippyshare.com/v/bW3crAV2/file.html

The same label map stored as nrrd and mha.

If I switch from nrrd to mha in the Editor, i get the error:
http://imgur.com/a/d1MdR

I use version 4.6.2.

Thanks for reporting - I'd like to try with the sample data you provided, but the zippyshare.com link has so many ads with download buttons that I cannot tell which one actually links to the data. Can you try again with dropbox or google drive? Or even just upload to the bug tracker if the file is small enough.

^^^^ done

Attached Files:
Slicer_Bug3776_Example.zip [^] (3,667,585 bytes) 2017-05-30 14:38

Thanks for the data - I was able to replicate the issue.

There is a truncation issue where the original data from the NRRD file:

space directions: (1.26953125,0,0) (0,1.26953125,0) (0,0,0.5)

Contains more digits than the Meta file:

ElementSpacing = 1.26953 1.26953 0.5

So the detection was correct, but the error message was incorrect because the error message only displayed 6 significant digits, but this included 3 before the decimal point while the code checks only the fractional part.

If I fix the error message I get this:

IJKToRAS mismatch at [0, 0] (-1.269531 != -1.269530)
IJKToRAS mismatch at [0, 3] (-420.214844 != -420.215000)
IJKToRAS mismatch at [1, 1] (-1.269531 != -1.269530)
IJKToRAS mismatch at [1, 3] (-264.062500 != -264.063000)

which is technically correct but still probably over sensitive for the nature of the problem.

Fix is committed so the dialog box accurately reports the numbers that are different:

http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=26061

It may still be the case that the test is overly sensitive, but I think that should be filed as an independent issue.

But if I understand correctly, the reason for the difference is the conversion from nrrd to mha, right?

Right - the mha file doesn't preserve all the significant digits. Probably this could be addressed by changing the print format in the Meta image code.