 |
 |
Dear Marjolein, > We have acquired DWI data with a Philips 3T scanner. I would like to run topup and eddy on these and for that you need to create a file with acquisition parameters. However it’s unclear to me what to put in this file. Philips scanners do not give you the phase encoding direction. Instead, it gives a ‘fold-over direction’ and ‘fat shift direction’. Our scans are all in ‘fold-over direction = “AP”’, and we have 1 b0 and 30 dwi’s with ‘fat shift direction = “A” as well as 1 b0 and 30 dwi’s with ‘fat shift direction = “P”’. > The question is: does ‘fat shift direction = “A” mean A -> P and “P” P -> A or vice versa? > A similar question was asked previously (see https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1308&L=FSL&P=R26762&1=FSL&9=A&I=-3&J=on&d=No+Match%3BMatch%3BMatches&z=4 ). It was suggested that you can find out the answer to this question by running the correction both ways (i.e. with different acquisition parameters) and see which possibility leads to a good correction. > I have tried this by running topup and applytopup with both of the following acquisition parameters. > 0 -1 0 0.19 > 0 1 0 0.19 > and > 0 1 0 0.19 > 0 -1 0 0.19 > Both options however lead to the exact same corrected images. Does anyone have an explanation for this? yes. If you get the signs wrong (in your case) you will still find exactly the same corrected images, BUT the off-resonance field will be sign-swapped. So, if you specified the —fout argument you can look at your estimated off-resonance field and you will find that one of them has positive values (indicating a “higher than expected” field) around the sinuses and the ear canals, while the other one will have negative values in those same regions. It is the former that is the “correct” field. If all you want (which probably is what most people want) is to have your images corrected for distortions and you have pairs of images with opposing PE directions then getting the sign wrong doesn’t matter for either topup or eddy. BUT it will mean that all your estimated off-resonance fields will be sign-swapped, on the off chance you want to do something to them. It DOES matter if instead of just having opposing PE directions you have for example pairs of A->P and L->R, then you need to get it right. It also matter if you have different echo-spacing for the different images. I hope that was clear? Jesper