GE Multi-echo sequences

[neurolabusc]

Current GE software (DV26.0_R03_1831.b) running research multi-echo sequences create invalid DICOM images. The required public EchoTime (0018,0081) attribute lists the shortest echo time for the series, rather than the actual echo time for the given DICOM image. The public tag EchoNumber (0018,0086) reports 1 for all echoes. In addition, these sequences do not provide unique InstanceNumbers (0020,0013) for different slices in the sequence. While this latter behavior is not strictly prohibited by the DICOM standard, some tools and naming conventions may not distinguish different slices.

These sequences can be detected by the private tag Number of Echoes (0019,10a9) of VR type DS. A value greater than 1 indicates a multi-echo sequence. In practice, the echo number can then be inferred by sorting the data based on the private tag Raw Data Run Number (0019,10A2) or the the public SOPInstanceUID (0008,0018). Note the DICOM standard does not require sequential UIDs, but assuming that these are sequential works with observed datasets.

The developmental release of dcm2niix detects these sequences and directs the user to this issue. This issue is a limitation of the research GE sequence, not dcm2niix. It is hoped that GE will properly report EchoTime and EchoNumber prior to making this a product sequence (and unique InstanceNumbers would be appreciated). In the meantime, users encountering these images should consider:

• Vinai Roopchansingh's sortme script
• The developmental release provides the %y option to include the private Raw Data Run Number in the output filename. This provides a convenient way to avoid name clashes with these images and conduct subsequent sorting. A minimal DICOM renaming would be /dcm2niix -r y -f %s_%p_%3r_%5y.dcm -o ~/out ~/tmp. This kludge may or may not be included in future stable releases.

[neurolabusc]

Vinai Roopchansingh notes this was specific to custom sequences developed by the NIH and shared with other teams. This should not impact product sequences:

I think at NIH, we may have to take a bit of the blame for the multi-echo EPI data that was missing echo proper echo time labeling. AFAIK – that was our implementation that was shared with other sites we collaborate with. The new ATSMs from GE (I think we have access to all of them) now provide proper labeling of multi-echo EPI data.