[ENH] added interface for dcm2niix #1435

mgxd · 2016-04-15T17:43:27Z

support for @neurolabusc's dcm2niix converter - compatible upto 4.12.16 version

satra · 2016-04-19T21:04:16Z

nipype/interfaces/dcm2nii.py

@@ -180,3 +180,127 @@ def _gen_filename(self, name):
            f.close()
            return config_file
        return None
+
+## dcm2niix update


remove this line

satra · 2016-04-19T21:06:48Z

@mgxd - could you again merge with current upstream master? and also run make check-before-commit

DimitriPapadopoulos · 2016-04-20T07:56:53Z

Some initial comments:

Under "Examples" the generated cmdline contains options -b and -m which are not available in versions of dcm2niix distributed by NeuroDebian on Ubuntu 14.04 for example. This doesn't seem to bother dcm2niix which seems to be skipping unknown options. No BIDS file is created though so in that case it's not part of the output. Not sure how to handle this.
Also I would change Uses MRICRON's dcm2niix to Uses Chris Rorden's dcm2niix or at least Uses MRIcroGL's dcm2niix.

mgxd · 2016-04-20T14:27:29Z

@DimitriPapadopoulos Thanks for the comments.

If bids option is empty (or doesn't exist), it shouldn't be returned (there's a check for an empty list). Since we'll be transitioning to bids format across all our datasets, it's a useful feature to include.
Changed MRIcron's dcm2niix to Chris Rorden's dcm2niix

Let me know if you catch anything else!

DimitriPapadopoulos · 2016-04-20T14:41:35Z

I'm not sure it can work in case DICOM file names are not passed as absolute paths.

I'm in a directory with hundreds of DICOM files:

$ cd /foo/bar
$ ls
[...]
20140614163106701.MR.dic
[...]
20140614163106761.MR.dic
[...]
$

Then:

>>> from nipype.interfaces.dcm2nii import Dcm2niix
>>> converter = Dcm2niix()
>>> converter.inputs.source_names = ['20140614163106701.MR.dic', '20140614163106761.MR.dic']
>>> converter.inputs.compress = 'i'
>>> converter.inputs.output_dir = '.'
>>> converter.cmdline
'dcm2niix -b y -z i -m n -f %q -o . -s n -v n 20140614163106701.MR.dic'
>>>

However dcm2niix doesn't seem to be processing paths relative to the current directory:

$dcm2niix -b y -z i -m n -f %q -o . -s n -v n 20140614163106701.MR.dic
Chris Rorden's dcm2niiX version 9Sept2015 (64-bit)
Version 9Sept2015 (64-bit)
Error: unable to find any DICOM images in 
Conversion required 0.000067 seconds.
$

unlike dcm2nii:

$ dcm2nii -a y -c y -b config.ini -v y -d y -e y -g y -i n -n y -o . -p y -x n -f n 20140614163106701.MR.dic
Chris Rorden's dcm2nii :: 4AUGUST2014 (Debian) 64bit BSD License
reading preferences file config.ini
0 files in working directory /volatile/070000160913/TEST
Data will be exported to ./
Validating 552 potential DICOM images.
Found 551 DICOM images.
Converting 202/551  volumes: 202
20140614161725601.MR.dic->20140614_154657EPIfacess004a001.nii
[...]
$

What about os.path.join(os.getcwd(), ...) at least for file names?

DimitriPapadopoulos · 2016-04-20T15:02:41Z

From my test directory with 2 DICOM sequences, Dcm2nii and Dcm2niix create 2 NIfTI files.

With Dcm2nii default output file names used to be:

20140614_154657EPIfacess004a001.nii
20140614_154657EPIstopsignals005a001.nii

With Dcm2niix default output file names are:

EP.nii.gz
EPA.nii.gz

I understand the default has changed between dcm2nii and dcm2niix but would it be worth changing the default output format of Dcm2niix from %q to something more like Dcm2nii? Or is Nipype supposed to be as transparent as possible?

coveralls · 2016-04-20T15:19:26Z

Coverage decreased (-0.05%) to 72.489% when pulling 62ca473 on mgxd:master into 60f7630 on nipy:master.

mgxd · 2016-04-20T16:15:09Z

@DimitriPapadopoulos

When converting a single image, the -s flag should change to 'y'. I was able to successfully convert one of my dicoms:

dcm2niix -b y -z i -m n -f %q -o . -s y -v n 996000-3-1.dcm
Compression will be faster with /usr/local/bin/pigz
Chris Rorden's dcm2niiX version 12Apr2016 (64-bit)
Version 12Apr2016 (64-bit)
Convert 1 DICOM as ./EP (108x108x65x1)
Conversion required 0.240000 seconds.

I've updated the doctest to include the single file flag.
EDIT: It seems the doctest fails with this flag.

As for default naming - I have no problem making the output look more like dcm2nii's. I've changed the default from %q to %t%p (time and protocol).

satra · 2016-04-21T13:39:40Z

nipype/interfaces/dcm2nii.py

+    def _run_interface(self, runtime):
+        new_runtime = super(Dcm2niix, self)._run_interface(runtime)
+        (self.output_files,
+         self.bvecs, self.bvals, self.bids) = self._parse_stdout(new_runtime.stdout)


have you tested this with and without the bids option?

DimitriPapadopoulos · 2016-04-21T23:30:09Z

@mgxd
I'm not certain about changeset 3603523:
Unless I'm mistaken it merely modifies the example and switches from the initial use case (converting a directory) that (still) doesn't work to a different use case (converting a single image) that works.
I would have expected the documentation to stay as is and the code to be modified so that the initial example does work (perhaps by adding os.path.join(os.getcwd(), ...) somewhere in the code).

Indeed in single file mode dcm2niix is able to handle relative pathnames:

$ dcm2niix -b y -z i -m n -f %q -o . -s y -v n seminoma.dic
Chris Rorden's dcm2niiX version 9Sept2015 (64-bit)
Version 9Sept2015 (64-bit)
Convert 1 DICOM as ./EP (64x64x40x1)
Conversion required 0.027342 seconds.
$

but it cannot convert directories identified by a relative pathname:

$ dcm2niix -b y -z i -m n -f %q -o . -s n -v n seminoma.dic
Chris Rorden's dcm2niiX version 9Sept2015 (64-bit)
Version 9Sept2015 (64-bit)
Error: unable to find any DICOM images in 
Conversion required 0.000075 seconds.
$

unlike dcm2nii:

$ dcm2nii -a y -c y -v y -d y -e y -g y -i n -n y -o . -p y -x n -f n seminoma.dic
Chris Rorden's dcm2nii :: 4AUGUST2014 (Debian) 64bit BSD License
0 files in working directory /volatile/000045835265/TEST
Data will be exported to ./
Validating 397 potential DICOM images.
Found 393 DICOM images.
Converting 191/393  volumes: 191
[...]
$

I believe Dcm2niix should accept relative pathames outside of single file mode, perhaps by adding os.path.join(os.getcwd(), ...) to the Dcm2niix code (not to the example code). Then the initial example would work.

neurolabusc · 2016-04-22T11:24:35Z

@DimitriPapadopoulos
I fixed the relative paths in the 22Apr2016 version of dcm2niix. The fix works on OSX and Linux, not sure about Windows (not sure if '.' is current working directory in Windows).
https://github.com/neurolabusc/dcm2niix

DimitriPapadopoulos · 2016-04-22T13:26:58Z

@mgxd

Since Chris confirms the relative path issue has been fixed in recent versions of dcm2niix, perhaps it's not worth working around it in nypipe. By the time a new version of nipype is officially released, I guess the recent version of dcm2niix might have made it to NeuroDebian.
Also I'm not totally convinced about changing the default NIfTI output file name after all. I was just asking a question about consistency between Dcm2nii and Dcm2niix, not firmly suggesting Dcm2niix should not follow the default behaviour of dcm2niix. Perhaps this should be best discussed and handled at the dcm2nii-dcm2niix level.

Perhaps 3603523 should be reverted then. What do you think? Sorry about that.

mgxd · 2016-04-22T14:16:34Z

@DimitriPapadopoulos

I agree - since it has been patched in the latest dcm2niix, I'd rather not add a change to this interface and then have to undo it. @satra what do you think?
I actually like the new naming better, more information as a default isn't a bad idea. I arbitrarily set the default to sequence anyways.

Thanks for the constant updates @neurolabusc!

satra · 2016-04-22T15:09:28Z

@mgxd - the one thing i would suggest is going back to the single file example in the doctest

satra · 2016-04-23T02:22:41Z

nipype/interfaces/dcm2nii.py

+
+class Dcm2niixInputSpec(CommandLineInputSpec):
+    source_names = InputMultiPath(File(exists=True), argstr="%s", position=-1,
+                                  copyfile=False, mandatory=True, xor=['source_dir'])


desc missing for all inputs.

i know dcm2nii didn't have any either. but that should be fixed as well.

coveralls · 2016-04-23T17:24:09Z

Coverage decreased (-0.05%) to 72.477% when pulling 252a6d6 on mgxd:master into 9121042 on nipy:master.

satra · 2016-04-25T15:10:39Z

nipype/interfaces/dcm2nii.py

+    output_dir = Directory(exists=True, argstr='-o %s', genfile=True,
+                           desc="Output directory")
+    bids_format = traits.Bool(True, argstr='-b', usedefault=True,
+                              desc="BIDS sidecar")


perhaps change to: Create a BIDS sidecar file

coveralls · 2016-04-25T15:53:14Z

Coverage decreased (-0.05%) to 72.477% when pulling 269ca84 on mgxd:master into 9bc6c8b on nipy:master.

mgxd and others added 8 commits April 15, 2016 11:56

added interface for dcm2niix

01e5792

dcm2niix add

231e813

added new dcm2niix flag

778dccf

Update CHANGES

28e1e1a

fixed test

8c21803

test fix

a473e8b

Merge https://github.com/mgxd/nipype

26818ba

Update dcm2nii.py

13e131f

satra reviewed Apr 19, 2016
View reviewed changes

mgxd and others added 5 commits April 19, 2016 17:11

Update dcm2nii.py

fff34ae

removed white lines

24863c9

upstream merge

7fa0ce8

CHANGES updated

68a9e8f

removed white line

2f4f46d

mgxd added 3 commits April 20, 2016 10:19

doc test fix

d0ccfde

Merge https://github.com/mgxd/nipype

a2c8764

Merge https://github.com/nipy/nipype

62ca473

default outname and single file doc test

3603523

mgxd added 2 commits April 20, 2016 12:23

Merge https://github.com/nipy/nipype

4def5b7

reverted doc test

d3d30f4

satra reviewed Apr 21, 2016
View reviewed changes

mgxd added 2 commits April 21, 2016 10:22

Merge https://github.com/nipy/nipype

d16e693

fix: compatible without bids option

c3c7cfb

fix: single file doctest

a67b26f

satra reviewed Apr 23, 2016
View reviewed changes

enh: desc for all inputs

252a6d6

satra reviewed Apr 25, 2016
View reviewed changes

mgxd added 2 commits April 25, 2016 11:34

Merge https://github.com/nipy/nipype

8619c30

fix: additional desc info

269ca84

satra merged commit aeecd2f into nipy:master Apr 27, 2016

[ENH] added interface for dcm2niix #1435

[ENH] added interface for dcm2niix #1435

Uh oh!

Conversation

mgxd commented Apr 15, 2016

Uh oh!

satra Apr 19, 2016

Choose a reason for hiding this comment

Uh oh!

satra commented Apr 19, 2016

Uh oh!

DimitriPapadopoulos commented Apr 20, 2016

Uh oh!

mgxd commented Apr 20, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

DimitriPapadopoulos commented Apr 20, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

DimitriPapadopoulos commented Apr 20, 2016

Uh oh!

coveralls commented Apr 20, 2016

Uh oh!

mgxd commented Apr 20, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

satra Apr 21, 2016

Choose a reason for hiding this comment

Uh oh!

DimitriPapadopoulos commented Apr 21, 2016

Uh oh!

neurolabusc commented Apr 22, 2016

Uh oh!

DimitriPapadopoulos commented Apr 22, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mgxd commented Apr 22, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

satra commented Apr 22, 2016

Uh oh!

satra Apr 23, 2016

Choose a reason for hiding this comment

Uh oh!

satra Apr 23, 2016

Choose a reason for hiding this comment

Uh oh!

coveralls commented Apr 23, 2016

Uh oh!

satra Apr 25, 2016

Choose a reason for hiding this comment

Uh oh!

coveralls commented Apr 25, 2016

Uh oh!

Uh oh!

mgxd commented Apr 20, 2016 •

edited

Loading

DimitriPapadopoulos commented Apr 20, 2016 •

edited

Loading

mgxd commented Apr 20, 2016 •

edited

Loading

DimitriPapadopoulos commented Apr 22, 2016 •

edited

Loading

mgxd commented Apr 22, 2016 •

edited

Loading