I’m preparing a Siemens DICOM series for preprocessing with fMRIPrep, but I’ve noticed two variants of my functional runs (i.e. the same DICOM series but with one more additional tag):
[ 'ORIGINAL', 'PRIMARY', 'M', 'ND', 'MOSAIC' ]
[ 'ORIGINAL', 'PRIMARY', 'M', 'ND', 'NORM', 'MOSAIC' ]
- Series 1 has no “NORM” tag
- Series 2 includes the “NORM” tag (is that vendor‑applied intensity normalization? I can see visually that the time series with the NORM looks more uniform)
Which series should I feed into my BIDS-Converter so that fMRIPrep operates on the
correct input?
Importantly: for my T1w images, I only have the variant with ‘NORM’.
Thanks in advance for any insights!
Yes, the NORM (intensity normalization, you can indeed see it when you compare the images) is something that comes with the pulse sequence protocol. I always use the NORM data and either exclude the other image or save it as extra data. Fmriprep also does an intensity correction btw (called non-uniformity correction)
Okay thanks! I guess then I will take the NORM versions and set the others to ‘exclude’? Just out of curiosity: Will this cause any problems when fmriprep does the uniformity correction again although the image has already been corrected by the scanner software?
Mhhh, now that I think a bit more about it, I believe fmriprep only corrects the T1 image, not the BOLD data. It depends a bit on your subsequent analysis what is best, because the intensity correction may have some benefits for some algorithms, it also affects the noise distribution