Fails on bvecs/bval copy part of DWI conversion

[mih]

I have a Philips DWI scan (nothing fancy, standard clinical protocol). heudiconv fails like this:

  File "/usr/lib/python2.7/dist-packages/nipype/interfaces/base.py", line 1128, in aggregate_outputs
    raise FileNotFoundError(msg)
nipype.utils.filemanip.FileNotFoundError: File/Directory '['/tmp/heudiconvtmpm4a6EF/convert/20130717_141500DTIhigh2isoSENSEs701a1007.bvec']' not found for Dcm2nii output 'bvecs'.
Interface Dcm2nii failed to run.

That is, because the file it is looking for carries an x prefix -- which is strange, as dcm2nii was executed with this command:

Running: dcm2nii -a y -c y -b config.ini -v y -d y -e y -g y -i n -n y -o /tmp/heudiconvtmpm4a6EF/convert -p y -x n -f n

This could be considered a bug in nipype, dcm2nii, or heudiconv. It would be relatively simple to work around in heudiconv: Check if res.outputs.bvecs is present and a corresponding file exists. If it is present, but the file doesn't exist, look for the same file with a x prefix.

What approach would you suggest?

[chrisgorgo]

I think it's nipype...

On Thu, Nov 5, 2015 at 2:00 AM, Michael Hanke notifications@github.com
wrote:

I have a Philips DWI scan (nothing fancy, standard clinical protocol).
heudiconv fails like this:

File "/usr/lib/python2.7/dist-packages/nipype/interfaces/base.py", line 1128, in aggregate_outputs
raise FileNotFoundError(msg)
nipype.utils.filemanip.FileNotFoundError: File/Directory '['/tmp/heudiconvtmpm4a6EF/convert/20130717_141500DTIhigh2isoSENSEs701a1007.bvec']' not found for Dcm2nii output 'bvecs'.
Interface Dcm2nii failed to run.

That is, because the file it is looking for carries an x prefix -- which
is strange, as dcm2nii was executed with this command:

Running: dcm2nii -a y -c y -b config.ini -v y -d y -e y -g y -i n -n y -o /tmp/heudiconvtmpm4a6EF/convert -p y -x n -f n

This could be considered a bug in nipype, dcm2nii, or heudiconv. It would
be relatively simple to work around in heudiconv: Check if
res.outputs.bvecs is present and a corresponding file exists. If it is
present, but the file doesn't exist, look for the same file with a x
prefix.

What approach would you suggest?

—
Reply to this email directly or view it on GitHub
#13.

[satra]

it's nipype crashing (this should be a simple fix), but it's dcm2nii generating that prefix (i don't know why). it might be good to know from chris why that filename has an x prefix.

[mih]

I think the x prefix appears whenever dcm2nii performs some kind of correction or normalization. I went to the sources:

$ git grep ChangeFilePrefix.*\'x\'
niftiutil.pas:     lClipName := ChangeFilePrefix (lOutImgName,'x');
nii_3dto4d.pas:       lHdrName := ChangeFilePrefixExt (lHdrName,'x');
nii_3dto4d.pas:       lImgName := ChangeFilePrefixExt (lImgName,'x');
nii_3dto4d.pas:       lHdrName := ChangeFilePrefixExt (lHdrName,'x');
nii_3dto4d.pas:       lImgName := ChangeFilePrefixExt (lImgName,'x');
nii_4dto3d.pas:        //lTempName := lFilename;//ChangeFilePrefixExt (lFileName,'x');
nii_4dto3d.pas:       lOutImgName := ChangeFilePrefix (lHdrName,'x');
nii_4dto3d.pas:       lHdrName := ChangeFilePrefixExt (lHdrName,'x');
nii_4dto3d.pas:       lImgName := ChangeFilePrefixExt (lImgName,'x');
nii_4dto3d.pas:       lHdrName := ChangeFilePrefixExt (lHdrName,'x');
nii_4dto3d.pas:       lImgName := ChangeFilePrefixExt (lImgName,'x');
nii_crop.pas:        //lTempName := lFilename;//ChangeFilePrefixExt (lFileName,'x');
nii_math.pas:     lOName := ChangeFilePrefix(lAName,'x');
nii_math.pas:     lOName := ChangeFilePrefix(lAName,'x');

[mih]

Here is how it looks if I replicate the nipype call:

% dcm2nii -a y -c y -b config.ini -v y -d y -e y -g y -i n -n y -o /tmp/heudiconvtmpvIys62/convert -p y -x n -f n /tmp/heudiconvtmpvIys62/convert/MR.1.3.46.670589.11.38317.5.0.8636.2013071715092953613
Chris Rorden's dcm2nii :: 4AUGUST2014 (Debian) 64bit BSD License
0 ERROR: unable to find config.ini
Data will be exported to /tmp/heudiconvtmpvIys62/convert/
Validating 2388 potential DICOM images.
Found 2380 DICOM images.
Converting 2380/2380  volumes: 34
MR.1.3.46.670589.11.38317.5.0.8636.2013071715073853001->20130717_141500RegDTIhighisos702a1007A.nii
 brightest voxel was 2731: data will be saved as 16-bit signed integer.
GZip...20130717_141500RegDTIhighisos702a1007A.nii.gz
Number of diffusion directions = 34
GZip...x20130717_141500RegDTIhighisos702a1007A.nii.gz
Removed DWI from DTI scan - saving volumes 1..33

The last bit is the culprit. The corresponding code in dcm2nii says:

              end else if lDicomData.ManufacturerID = kPhilipsID then begin
                   //-->PhilipsCorrectBvecs(lDicomData,lDTIra,lDTIdir);
                 siemensPhilipsCorrectBvecs(lDicomData,lDTIra,lDTIdir,false);
                   //next: philips scans can include DWI images with bval>0 and v1=0,v2=0,v3=0 - we want to exclude these

This is followed by a call to SaveNIfTICoreCrop, which add the prefix.

There you go ;-)

[mih]

I have a patch for nipype. Will push shortly.

[mih]

Closing this one, as a fix went into nipype.