Use AcquisitionTime for acq_time column of scans tsv files #450
Description
Summary
It looks like heudiconv uses ContentTime in the scans tsv files instead of AcquisitionTime, although I think that AcquisitionTime would be more appropriate (i.e., closer to the actual start of the scan).
Quoting @pvelasco in physiopy/phys2bids#219 (comment):
In dicoms, SeriesTime, InstanceCreationTime, and ContentTime are very similar, but AcquisitionTime is generally ~20 seconds earlier than the rest. We need to determine which time is closest to the scanner's trigger pulse.
From some documentation I found from Siemens on image reconstruction:
AcquisitionTime: The Time the acquisition of data started.
From an answer given by Chris Roden (dcm2niix):
Instance Creation Time 0008,0013 initially reflects the time an image leaves a parallel processing reconstructor, so it has a variable delay with regards to the acquisition time. Acquisition Time is more meaningful for data.
So, if the reconstruction involves something more than a straight FFT (i.e., GRAPPA, SENSE, etc.), that will explain the ~20 sec delay you find between the AcquisitionTime and the other times.
If I remember correctly, for Siemens scanners, the "scanner trigger" is sent at the time of the excitation pulse of the first slice of each (collected) volume. So the very first trigger will be sent just a few milliseconds before the first chunk of data for the first slice for the first volume is acquired (AcquisitionTime).
This issue is related to #447.