we acquired fieldmaps in our study, but the fieldmaps seemed to rather enhance artefacts. This is why I now used the fieldmap-less approach from fmriprep. So far I can judge, the corrected fmri images look good. However I wonder if it is correct to use this approach in the case of our data. We acquired the T1 structural scans with a 32 channel coil. The fmri data instead were acquired with a 12 channel coil. Are the different channel coils problematic for the fieldmap less approach?
Is the PhaseEncoding direction parameter correct? If so, this could indicate a bug in fMRIPrep. Otherwise, well, you found an important problem in your dataset!
Having fieldmaps I would try to get them to work. The fieldmap-less correction could overfit the distortion pretty easily. Nobody knows the implications of that at this point.
I don’t think that would be a problem - the number of coils do not affects susceptibility distortions as regards the correction.
Yes true! I have resting fmri and task based fmri data and for both data seperate fieldmaps. For the resting fmri data the fieldmaps seem to work well, only for the task based data not. The resting fmri data and the corresponding fieldmaps both have a phase encoding direction of j- and are axial slices. The task based fmri data and the corresponding fieldmaps both have a phase encoding direction of j and are coronal slices. For all 4 of them the .json file says the “InPlanePhaseEncodingDirectionDICOM” is “COL”. Maybe there is a problem with phase encoding in j- direction or with coronal slices?
I will send a link to two example reports. I could not directly post them here because the format svg is not allowed.
I hope you can give me some advice! Thank you!
The Siemens 12-channel is generally a solid performer, and allows scanning people who are unable to fit comfortably in the 32 channel. Just be aware that this coil was designed before the multi band era, so you need to be restrained with multi-band. In contrast, the Siemens 20-channel head(16)/neck(4) coil is a better design for multi-band.
@oesteban oh no… that’s weird. For me the link works perfectly and folders are not zero size… could you try again? Or which homepage/cloud would you recommend to share the data best?
Thank you!!
Hi @anna, I tried again and this time I could see your data.
I’m really sad to carry very bad news for you. Looking at your fieldmaps (with the only exception of sub-con2 and task rest), it seems to me that both your fieldmaps and BOLD data have very strong high-order shim effects. I’m afraid that any manipulation you try to fix your data will be dubious. It’s no wonder why the fieldmap-less correction seems to work better, after all it is overfitting the distortion.
You can clearly tell the differences across your fieldmaps and BOLD - the task-rest shows you how your data should look like. BTW, I’ve also seen really bad FoV prescriptions - your magnitude images clearly show that the frontal lobe is cut out.
I hope this reaches you in time to revise your acquisition protocol.
Hi @oesteban , we used a 12 channel coil and coronal slicing in all conditions except for the rest condition, in which we used a 32 channel coil and axial slicing. The decision for the coronal slicing was a trade off of having reduced distortion in the anterior mesial temporal lobe but increased frontal distortion. But probably coronal slicing must not go in hand with higher order shim effects. Unfortunatly it is to late now to revise the acquisition protocol and we have to deal with the data this way… Would you recommend using the fieldmap-less approach on this data? And for the future, can you tell how to most likely revise the acquisition protocol to decrease the higher order shim effects? Thanks again!!
