I am trying to conducting distortion correction on my EPI (collected by Siemens Prisma 3T scanner) using spm toolbox FieldMap. The original .DCM files was converted to .nii files using MRIconvert.
Although the scanning sequence and the setting for MRIconvert were the same for all participants, MRIconvert generated two magnitude maps and one phase map for some participants, which seems to be the normal case, but only one magnitude map (together with one phase map) for several others.
I followed the instruction on this website Chris Rordens Neuropsychology Lab :: Fieldmaps (very clear and helpful btw). However, I can only do correction for participants with one magnitude map. When I tried those of two magnitude maps, I got an error “images don’t all have same dimensions”.
I was wondering if there is any clue on:
why the same setting would lead to two types of magnitude maps for different participants? and
how to solve the “images don’t all have the same dimensions” error?
Thank you very much! I am very grateful for any advice.
Given that you have a Siemens scanner, my best guess is that some of your archiving tools are ignoring the fact that different images from this series have identical instance numbers (0020,0013). While the DICOM standard does not require instance numbers to be unique, they typically are and many tools assume that duplicate instance numbers are repeats of the same image, so one overwrites the other. Unfortunately, Siemens will reuse instance numbers within a series, as seen in this validation dataset. While dcm2niix is aware of this issue, many storage tools ignore this. Make sure that any storage tools use the Media Object Instance UID (0002,0003).
The other idea is to use a different image converter. I assume you mean Jolinda Smith’s MRIconvert, which has not been updated in many years. Alternatively, you might have used the FreeSurfer mri_convert which has recently been updated to use dcm2niix. Regardless, it might be worth seeing if a latest release of dcm2niix is more successful.
I followed your advice and tried MRIcroGL only got the same results as MRIconvert produced.
But it occurred to me that I needed the two magnitude maps to be merged into one to run spm fieldmap correction (I dont know why but these data would work) so I selected “Always Merge Series Regardless of Differences” in MRIcroGL and got a single magnitude map for each participant. Luckily FieldMap worked this time!
I’ve been runing into this trouble for over a month and could not find a proper solution. Your reply really helped me out!