When you have half the magnitude image data because of downstream processing that dropped what was deemed to be duplicate slices when in reality they came from a different echo, is it legitimate to combine the available data into a single “complete” volume as the lone magnitude image to use for fieldmap correction ? dcm2niix -m 1 will produce an image which has every slice, albeit some come from one echo, others come from the other. In this case, the lost data is not recoverable from the scanner.
This can often be a hail mary - there will be some striping due to contrast differences, but it will likely be usable. The trick would be to isolate the problematic series into its own folder, and then use the merge function (-m y) to concatenate images even though they have different echo numbers.
So the solution would be to rename your DICOMs to isolate the problematic series:
dcm2niix -r y /path/to/DICOMs
and then convert only the corrupted series with dcm2niix:
Thanks @neurolabusc . This worked, and produced identical output to
dcm2niix -m 1 /path/to/unrenamed/mangled/series
that I had tried before posting the thread. I am curious why you suggested renaming first. I can’t tell from the names of the renamed individual dicom files which ones come from which echo. Thanks!