Error in position field for multi-stack dataset - Siemens raw data

[spichardo]

Hi,

We noticed an error in the calculation of the "position" field for a 2-stack dynamic (cine) scan acquisition. Each stack is orthogonal to each other. We never had this issue for other tests before including single slice slice dynamic scan, or a multi-slice (but one single stack) acquisitions.

I used the latest master branch version of the converter. The following link points to a google drive directory with the original Siemens raw data, the converted noise and data ISMRMRD files, a few of the dicom files exported from the Siemens console and a very simple Python Jupyter Notebook to show the error.

https://drive.google.com/drive/folders/1Pp7WRDF9NDT8qid53T5bHVWn-tkRvW-D?usp=sharing

It is the Z coordinate the one that is wrong; X, and Y are ok. Below we have print of the `position' field for the first ten entries.

F=h5py.File('DATA_meas_MID00038_FID04558_probe_test.h5','r')
d=F['dataset']['data']
for n in range(10):
    print n, d[n][0][16]
F.close()
0 [-23.68401909  70.12107086 -81.45278168]
1 [-23.68401909  70.12107086 -81.45278168]
2 [-28.18401909  68.66828156 -79.51573944]
3 [-23.68401909  70.12107086 -81.45278168]
4 [-28.18401909  68.66828156 -79.51573944]
5 [-23.68401909  70.12107086 -81.45278168]
6 [-28.18401909  68.66828156 -79.51573944]
7 [-23.68401909  70.12107086 -81.45278168]
8 [-28.18401909  68.66828156 -79.51573944]
9 [-23.68401909  70.12107086 -81.45278168]

When opening the exported DICOMs we can see the expected values

A=pydicom.read_file('Siemens DICOM/mag_1.dcm')
print A.ImagePositionPatient,CenterImage(A)
A=pydicom.read_file('Siemens DICOM/mag_2.dcm')
print A.ImagePositionPatient, CenterImage(A)
['-113.68401908875', '-19.878929138184', '-111.45278167725'] [ -23.68401909   70.12107086 -111.45278168]
['-28.184019088745', '-21.331718444824', '-19.515739440918'] [ -28.18401909   68.66828156 -109.51573944]

You can especially see the problem in the [ -23.68401909 70.12107086 -111.45278168] vs [-23.68401909 70.12107086 -81.45278168], which is for a transverse image, so Z remains constant in all image.

Data was converted with siemens_to_ismrmrd -f meas_MID00038_FID04558_probe_test.dat -z 2 -o DATA_meas_MID00038_FID04558_probe_test.h5

Any help to point us how we can fix this would be greatly appreciated, as I mentioned, we only have seen this in this specific multi-stack scan, any other datasets do not have this issue.

Thanks

Sam

[spichardo]

Apologies, digging a little more on the code (and adding some printing) I noticed this issue comes from the very source of the Siemens raw data , so it is not a problem of the converter. So I'm closing this non-issue for the converter.
S