I’ve got a question concerning susceptibility distortion correction (SDC) using fMRIprep and a field map acquisition using a Philips scanner.
It is very clear to me, from my own experience and by reading some of the posts here, that neither the DICOM nor the REC/PAR files that come from an acquisition with a Philips scanner give the phase encoding direction parameter, necessary for the SDC. I’m aware that this is not a problem with dcm2niix (as pointed out by Chris Rorden in some of his answers), but a limitation from Philips scanners.
Then, as suggested in this post (BIDS, FMRIPREP: specify phase encoding direction with respect to qform orientation in nifti header? - #4 by Chris_Rorden), we could assign one of the fieldmaps (magnitude and phase) acquired with one echo time with X and the other fieldmaps with the other echo time with X-, being X the axis indicated by the parameter PhaseEncodingAxis, provided by the conversion with dcm2niix. The post suggests that TOPUP will find out how to unwarp the images in both cases.
So I’ve tested that with ec1 images assigned with j and ec2 images with j-. I also tested the other way around, i.e., ec1 images assigned with j- and ec2 images with j. The preprocessed images using one or another approach are slightly different. They are similar, but with different numerical values.
I just would like to understand what is going on here. Does it mean that SDC is working successfully in both cases and the differences are negligible? Or does it mean that are allowing a small mistake in the preprocessing (due to the limitation of not having the PhaseEncodingDirection information) but that doesn’t bring large consequences?
From what you are describing, I think your fieldmap images are two magnitude and two phase images (Fielmap Case 2 in the BIDS Standard). Processing this type of field maps should not require specifying the Phase Encoding (PE) direction. If the BIDS validator requires you to enter one, leaving it as "j" should work. Make sure the filenames end with _magnitude? and _phase? (?: 1 or 2), and not _epi.
You would need to set the PE direction for the functional images you want to correct. If you cannot tell if PE direction (regardless of sign) is "i" or "j", what I would recommend is that you open the images in a viewer (e.g. FSLEyes) and move the cursor in the direction of the distortions while checking which of the indices indicating the location of the cursor moves. If it is the first one, that means that your PE direction should be "i" or "i-". If it is the second index, it will be "j" or "j-".
Unfortunately there is no way of knowing which sign it is, other than trying with both. In one case your images will be correctly undistorted while in the other case the distortions will become worse.
A couple of notes:
Chris Rorden’s post you link refers to EPI fieldmap images collected with opposing PE directions, which is not your case.
You found negligible differences in your tests because your fieldmap images are not distorted to begin with (they were collected using a GRE sequence, not an EPI sequence).
Many thanks for your valuable reply. It really helped me understanding how to solve the problem.
Yes, my case would the Case 2 in the BIDS Standard. Apologies for not telling it in advance, as it was important information. Indeed, I have been specifying the image names as _magnitude? and _phase?, as you pointed out.
The point that the PE direction should be correct for the functional images (and not necessarily for the fieldmaps) was not so clear to me. Thanks for the clarification. Finding out the PE axis shouldn’t be a problem if one also has the DICOM files, as this information can be retrieved from them (if I’m not mistaken), but not from REC/PAR files. Then, the PE direction for the functional image should still be missing.
Makes sense. So I’ve tried the preprocessing with the two PE directions, j and j-. Here is an image of the two preprocessed functional images, overlapped with the preprocessed anatomical file:
So I placed the cursor in the coordinates where see a distortion and compared the intensities (see red rectangles on the bottom). Would it be correct to assume that the images were correctly undistorted in the image on the left and became worse in the image on the right?
Or perhaps comparing the results from the reports? From the same images, one could see a better match between functional and anatomical files (the red line) in the slice on the right?
