Hi -
I'm very new at this - so I'm sure that this is a user error. I'm using
Chris Rorden's dcm2niiX version 1Jan2015 (64-bit)
on a Mac which I compiled from source.  I want to take 144 MR T2 Sagittal lumbar spine images (from a GE Signa HDxt, exported from Osirix) which I'd like to segment using sct. When I run
dcm2niix -v -z -y -o . ./xyz/*
I see the following:
Found 144 DICOM images
.... 
Then 144 entries like this
Patient Position 41.7495 -149.996 137.872
DICOM acq 1 img 1 ser 3 dim 512x512x1 mm 0.5078x0.5078x0.5 offset 15878 dyn 0 loc 72 valid 1 ph 0 mag 0 posReps 0 nDTI 1 3d 1 bits 16
then
Convert 144 DICOM as ./xyz (512x512x72x2)

When I load the *.nii in fsl it appears that only half of the data is present (the spinal column is basically sliced down the middle). The dimensions near the end 512x512x72x2 would suggest that only 72 of the 144 images were processed? If so - how do I convert the entire data set to *.nii?

Thanks for any help or suggestions.

Sean

Sean-

1.) Can you try a recent version of dcm2niix. Either get the code from Github to compile your own (7July2016) or get the pre-compiled version that comes with MRIcroGL (you can get 5May2016 on NITRC). The newer versions have been trained on a larger set of images and also tend to provide better messages regarding your images.

2.) Your command line options look confused, perhaps you want to start running dcm2niix from MRIcroGL's graphical interface (Import menu). The nice thing is when you define your options in the graphical interface it shows you what the command line will look like. I would think you meant "dcm2niix -v y -z y -o . ./xyz/"

3.) It looks like my software thinks that the spatial position is repeated so you have two copies of the same data. Hopefully this is because the 144 files represent 2 echoes, series, etc. On the other hand, it might be an error in the way the spatial position is reported in the DICOM file (DICOM is interpreted differently by different manufacturers, which makes decoding them much more complicated than simple explicit formats like NIfTI). In any case, if you still have problems with the latest version email me and provide me with a sample to explore.

Dr. Rorden -

Thanks. I'm still having issues with the latest version. I'll contact you via email with more details.

Sean

This is fixed in the 27July2016 release
  https://github.com/neurolabusc/dcm2niix
and seems specific to GE systems interpolating slices: Locations In Acquisition (0054, 0081) is inaccurate if slices are interpolated, the new version uses Images In Acquisition (0020,1002) if available. I would strongly recommend turning of interpolation - interpolation leads to larger files, slower processing, and specific interpolations algorithms have their own tradeoffs (e.g. blurring or ringing). However, my software should be able to cope with interpolated files.