I'm not sure what to tell you. You can try dcm2nii or MRIconvert (different package than mri_convert). You can also try an mri_convert from an older version of FS, which you can get from here
ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/
On 03/15/2017 03:18 PM, Mark Lescroart wrote:
Hello,
I'm trying to create a nifti file from some older (2006-2009) dicom files using mri_convert, and I'm having some difficulty. Part of this is uncertainty about the dicom files in question. The files were collected on Siemens scanner between 2006 and 2009, and I *believe* they should be in Siemens dicom format. They came off the magnet without a .dcm extension, if I recall correctly, but I used Brain Voyager to rename the dicom files (I believe all it does is re-title them to reflect the acquisition order, but I'm not 100% sure about that - that could be part of the problem). Running mri_convert to create a nifti file from the first file in each of the sequences I have yields the error I have copied below.
<redacted> bits are there for anonymity, I don't think those fields relate to the problem. The main problem with this call seems to be that some fields regarding the image position are not read correctly, so the affine transform has nans in it.
I also tried calling mri_convert -it dicom, which succeeded, but created an image that was of shape (384, 384, 1, 202). I don't know why that call to mri_convert doesn't correctly recognize that the images in the dicoms are, in fact, mosaic images (in the first call, there was an "IsMosaic" field that was set to 0, but I'm pretty sure it should be 1 or True or whatever); the resulting 4D array should be (64, 64, 32, 202).
However, the image that the second call created may have had a correct affine transform (at least there were no nans) - to check on it in python, I called:
nii = nibabel.load(file_name) nii.affine
... and got:
array([[ -3.5 , -0. , -0. , 672.67797852], [ -0. , -3.5 , -0. , 690.59564209], [ 0. , 0. , 3. , -52.07869339], [ 0. , 0. , 0. , 1. ]])
... So, now I'm a little stuck. -it seimens did not help, gave me the same error as below (same for -it siemens_dicom). I don't fully understand how to possibly override the bad info, or how to tell mri_convert where to find the correct info, or how to tell if I'm just out of luck with this whole endeavor.
mri_convert --version spits out "stable5", which might be a slightly older version.
Any help will be much appreciated.
Cheers, Mark
Error message:
Getting Series No Scanning Directory INFO: Found 993 files in old_data/ML_Localizers/Raw_Retinotopy_05_29_09 INFO: Scanning for Series Number 21 INFO: found 202 files in series INFO: loading series header info.
# <there were lots of these error lines, one for each file in the series; i think these are due to the space between the subject ID, which I've redacted, and -0021-0001-00001.dcm > # error: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/strings: invalid decimal number in flag: -0021-0001-00001.dcm Usage: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/strings [-] [-a] [-o] [-t format] [-number] [-n number] [[-arch <arch_flag>] ...] [--] [file ...]
WARNING: file <redacted> -0021-0001-00001.dcm does not contain a Siemens ASCII header has this file been anonymized? Proceeding as best as I can ...
INFO: (384 384 1), nframes = 202, ismosaic=0 PE Dir COL COL FileName <redacted>/<redacted> -0021-0001-00001.dcm Identification NumarisVer syngo MR B15 ScannerModel TrioTim PatientName <redacted> Date and time StudyDate <redacted> StudyTime 105825.156000 SeriesTime 122023.890000 AcqTime 122021.700000 Acquisition parameters PulseSeq *epfid2d1_64 Protocol <redacted> PhEncDir COL EchoNo 1 FlipAngle 90 EchoTime 30 InversionTime -1 RepetitionTime 2000 PhEncFOV 0 ReadoutFOV 0 Image information RunNo 20 SeriesNo 21 ImageNo 1 NImageRows 384 NImageCols 384 NFrames 202 SliceArraylSize 0 IsMosaic 0 ImgPos 672.6780 690.5956 -52.0787 VolRes 3.5000 3.5000 3.0000 VolDim 384 384 1 Vc -1.0000 -0.0000 0.0000 Vr -0.0000 -1.0000 0.0000 Vs nan nan nan VolCenter nan nan nan TransferSyntaxUID 1.2.840.10008.1.2.1 UseSliceScaleFactor 0 (slice 0: 1) INFO: no Siemens slice order reversal detected (good!). TR=2000.00, TE=30.00, TI=-1.00, flip angle=90.00 i_ras = (-1, -0, 0) j_ras = (-0, -1, 0) k_ras = (nan, nan, nan) Reslicing using trilinear interpolation MRIresample(): error inverting matrix; determinant is nan, matrix is: -3.500 -0.000 nan nan; -0.000 -3.500 nan nan; 0.000 0.000 nan nan; 0.000 0.000 0.000 1.000;
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer