HeuDiConv Criteria Spreadsheet (dicominfo.tsv)

HeuDiConv produces a spreadsheet, dicominfo.tsv, based on parsing the Dicom directory in Stage 1. This contains a lot of options for potential criteria (~27 criteria columns), but they are not all created equal.
The video below displays several columns in an example dicominfo.tsv file. I have color-coded these columns to facilitate exploration:
- Invariant columns are of no use (because they don't differentiate any of the DICOM series). They are unmarked in the example, e.g., study_description.
- Unstable columns marked in pink are not stable across subjects and scans. Avoid using these columns in your translator file if you can (though they can be useful for one-off conversion). Patient_id is a good example.
- Useful Columns distinguish series and should be stable across subjects. They are marked in green. If possible, use these for criteria. Protocol_name is a good choice, for example, the T1 Sagittal is uniquely identified by its protocol_name. However, protocol_name is not always sufficient to uniquely identify a sequence, for example, DTI_MB_3SHELL sequences cannot be differentiated by protocol name alone. Don't worry, HeuDiConv supports adding multiple criteria!
Derived Sequences: Look at the is_derived column. Some DICOM series (e.g., ASL, DWI) are partially processed by the scanner. For example, DWI images are optionally processed to create ADC, TRACEW, FA, and COLFA (color FA) images. These DERIVED images share many characteristics with the ORIGINAL images you want to convert, including the protocol_name, so you need criteria in your translator file that exclude these DERIVED images. All derived images are marked TRUE in the is_derived column, which can make is_derived an excellent additional criterion column.
Motion Corrected Sequences: Similar to the derived sequences, some datasets include motion-corrected fMRI files (marked MOCO); though there are none in the illustrated dataset, there are some in the tutorial dataset. You do not want to convert MOCO files (in fact, fMRIprep has been known to throw a fit if MOCO files are included in the BIDS directory). If any sequence is marked TRUE in the is_motion_corrected column, then you don't want to process it.
Keep in mind that you can ask the scanner tech not to create these derived and motion-corrected files. This might be a good idea since they only take up space and cannot be used for processing.

Derived and Motion Corrected images may be wrong

Ted Trouard remarks that some reconstructions created by Siemens are simply wrong. It is best not to rely on them.

Watch the 12.25-minute video below discussing these issues.

DICOM2BIDS 5: Heudiconv