Hi all,

We found dcm2nii misses files when converting .IMA DICOM images.

We have 74 files in one folder on a GlusterFS filesystem(a distributed network storage).
We ran the following command:
dcm2nii -O outb -v y testb
But we only got 50 output instead of 74

Any ideas how to resolve this?

We also did a control run on local hard drive storage, all 74 files were successfully processed.
The dcm2nii program itself

Output when on GlusterFS
----------
Chris Rorden's dcm2nii :: Nov 11 2011
reading preferences file /home/user/.dcm2nii/dcm2nii.ini
Data will be exported to outb/
Validating 50 potential DICOM images.
Found 50 DICOM images.
Converting 50/50 50
......
----------

The GlusterFS is mounted with command like this:
mount -t glusterfs -o transport=rdma gfs01:gfsvolume /share

Output when on local drive
----------
Chris Rorden's dcm2nii :: Nov 11 2011
reading preferences file /home/user/.dcm2nii/dcm2nii.ini
Data will be exported to outb/
Validating 77 potential DICOM images.
Found 74 DICOM images.
Converting 74/74 74
----------


Thanks for any input!

----------
Bradley
bangyuzhou@gmail.com

This seems very odd behavior.
 1.) Can you try a recent version of dcm2nii. You are using a version that is almost 5 years old. Conversion tools need to be updated to keep up with changes implemented by the scanner manufacturers. 
 2.) Are any of the files hidden (e.g. the first character of the filename is '.'?
 3.) Is the full path to the folder "testb" or is it more like "/home/dir/testb"? Do you get the same effects when you run the command from the same path?
 4.) What does the terminal report when you try to list all the files using the "ls" function, for example "ls -l /home/dir/testb"
 5.) Have you tried dcm2niix (which comes with MRIcroGL). In general, I would only prefer dcm2nii for converting legacy datasets (e.g. data in historical formats like Elscint, Signa, etc), for modern datasets would suggest using dcm2niix (just as I would recommend MRIcron for old computers and MRIcroGL for modern computers).

My guess is that this is not a problem with my software, but reflects that your Gluster has entered a split brain state. If my theory is correct, you need to heal your volume

https://gluster.readthedocs.io/en/latest...
https://www.gluster.org/pipermail/gluste...
https://access.redhat.com/documentation/...

Thanks for the input

1 & 5)
We did try dcm2niix, and its working fine on the same dataset(same environment, GlusterFS configuration), all 74 files were recognized.
           We also tried the most recent version of dcm2nii included in MRIcron, it had the same problem as the original post.

2 & 3 & 4)
I substitute relative path 'testb' to full path '/home/user/testb', same problem as the original post.
The IMA files are all named "141219***.IMA" , all normal non-hidden files. The ls -l command outputs something like:
-rwxr-xr-x 1 user user  1001406 Aug 19 13:55 141219_WZ_AC88_ID012.MR.FBIP_FMRI.0021.0057.2014.12.19.15.09.33.156250.28921948.IMA


Originally posted by Chris Rorden:

All of this is consistent with your Gluster entering a split brain state. You need to heal your Gluster. This is a problem with your distributed file system becoming out of sync / corrupted and not my software.

Very helpful info, thanks!
We are looking into our Gluster volumes. We'll keep you updated about our findings.

Originally posted by Chris Rorden: