This post is in regards to the thread below I dug up from October 2016, and is more of a question directed at Philips users who use FSL.
I have inherited this project and am attempting to run topup on the data described below by using the method Jesper suggested, of using a non-EPI T2, in lieu of a reversed b0, to warp the b0 image to. However, Philips does not clearly label a phase-encoding direction in their exam cards. They do report a "fold-over direction" and a "fat shift direction" instead. We had interpreted the fold-over direction (AP for the DTI acquisition, RL for the T2) to be the phase encoding direction. But another thread from August 2013 titled "eddy acqusition parameters", there are "educated guesses" from Philips users that the fold-over direction is just the plane of collection and instead, the fat-shift direction value is the true PE direction. The fat shift direction for our DTI and T2 acquisitions are both "P", which according to a post in the August 2013 thread, corresponds to a P>A PE direction.
My initial question is what do I interpret as the true PE direction of the scans? Secondly, my interpretation of our unusual use of topup is that we are not warping the b0 and T2 together, but rather warping the b0 to the T2 by telling topup the T2 has a very low dwell time (0.00005). Thus, does the RL vs AP direction really matter for these scans? Since I am not looking for the middle ground between the two images, I could really just set the PE of the T2 as the opposite of the b0, (either 0 1 0 or 0 -1 0, depending on the answer to my first question). Is this correct?
Any help is very much appreciated.
On Tue, 25 Oct 2016 20:02:16 +0000, Buckless, Colleen <[log in to unmask]> wrote:
>I tried what was recommending in which I ran topup using our b0_run1 and the T2 with only two rows in the acqparams file. Then I concatenated by DTI acquisitions using fslmerge and concatenated the bvals and bvecs. When I ran eddy, I got "segmentation fault" error. I double checked to make sure that my concatenated files had the same number of volumes in which they do, and tried the index file with 33 (one run) and 66 (two runs) 1's. When I have 33 1's I get "msg=--index must be an 1xN or Nx1 matrix where N is the number of volumes in --imain" and when I have 66 1's I get "segmentation fault". Do you have any ideas on what to do to fix this problem?
>From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Jesper Andersson
>Sent: Thursday, October 13, 2016 10:17 AM
>To: [log in to unmask]
>Subject: Re: [FSL] Can eddy be used on our DTI data?
>> Hi Jasper,
>> How would you recommend combining the two identical DTI acquisitions?
>> Should we:
>> a. Have the file that we use for topup include b0_run1, b0_run2, and
>> T2 and have this reflected in the acqparams.txt file by having three
>> rows instead of two and have this reflected in the rest of the
>> pipeline b. Average the two runs and run topup/eddy as normal c. Run
>> the two runs separately through topup/eddy and then average at the end
>this depends on if the subject left the scanner and/or if there was a re-shim between the two sessions.
>If not I recommend
>1. Use the b0 from the first session for topup, having only two rows in the acqparams files.
>2. Concatenate your DTI acquisitions using fslmerge, and correspondingly concatenate your bvals and bvecs using paste.
>3. Run eddy on the concatenated data using an --index file of only ones.
>If there was a re-shim I recommend running two separate topups for the two sessions and to feed those results into two separate eddy runs.
>> Is there another option that I am missing that makes more sense?
>> Thank you for all of your help,
>> -----Original Message-----
>> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
>> Behalf Of Jesper Andersson
>> Sent: Tuesday, October 11, 2016 4:07 AM
>> To: [log in to unmask]
>> Subject: Re: [FSL] Can eddy be used on our DTI data?
>> Dear Colleen,
>>> Hi Jasper,
>>> We tried your suggestion of using the T2 instead of having a blip-up and blip-down b0. However, we have some concerns because the "eddy" data appears to be distorted. Given that we acquired our images using right to left encoding, our acqparams.txt had to be changed to reflect this. The fourth element was also changed using the formula suggested on FSL website (the "dwell time" multiplied by "number of PE steps -1").
>>> This is what our acqparams.txt file looks like.
>>> -1 0 0 0.113
>>> 1 0 0 0.113
>> Firstly, make sure you put the EPI as the first volume in your 4D file, and the T2 as the second.
>> Secondly, the trick here is to tell topup that the second image has no distortions, and you do that by setting a very low, arbitrary, readout time for it. If you look at my original email you see that I suggested 0.00005. That way, instead of warping the two images towards each other, topup will warp the first image towards the second.
>> So, try for example
>> -1 0 0 0.113
>> 1 0 0 0.00005
>> The signs of the first columns don't really matter in this context.
>>> Is this correct? Should the top row have a positive 1 because we don't have a blip up and down, but have a t2 instead?
>>> Thank you for all of your help,
>>> -----Original Message-----
>>> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
>>> Behalf Of Jesper Andersson
>>> Sent: Friday, October 07, 2016 6:56 AM
>>> To: [log in to unmask]
>>> Subject: Re: [FSL] Can eddy be used on our DTI data?
>>> Dear Colleen,
>>>> I would like to know if it is possible to use "eddy" on our data. Our DTI images were acquired with single-shot EPI (SENSE factor 2.5) on a 3T 'Acheiva' scanner (Philips, Best, the Netherlands). Two runs were collected in each subject, with 32 gradient directions (b = 800 s/mm3) and one b0 in each run. Sixty 2.2-mm axial slices were acquired for each volume, with 0.9 mm in-plane reconstructed resolution. We also acquired a non-EPI axially acquired T2-weighted image. Is it possible to average the two runs together and then run eddy? Can we still use eddy even though we do not have opposite phase encoding directions?
>>> yes, eddy does not require opposing PE directions. How well it works will depend a little on how your b-vecs are distributed. I know that some Phillips protocols have a rather funny distribution of b-vecs and I have seen problems with running eddy on such data sets.
>>> You do need at least one pair of b0 volumes with opposing PE directions if you want to run topup in order to estimate the susceptibility induced off-resonance field. If you have that, you can then pass the topup output (field estimate) into eddy, which means that eddy will correct the data also for susceptibility distortions.
>>> If you don't have that, you can still run eddy and correct your data for eddy currents and movement.
>>> In your case, if the contrast in your T2-image is sufficiently
>>> similar to the b0-images, you may be able to use your T2 in lieu of a
>>> reversed-PE b0. In that case you stick one of your b0:s followed by
>>> the T2 image in a 4D file and make an acqparams.txt file looking
>>> something like
>>> 0 1 0 0.05
>>> 0 -1 0 0.00005
>>> provided that the PE-direction in your EPI images is A<->P (if it is in the L<->R direction you put the 1 and -1 in the first column instead).
>>> The materials in this e-mail are private and may contain Protected Health Information. Please note that e-mail is not necessarily confidential or secure. Your use of e-mail constitutes your acknowledgment of these confidentiality and security limitations. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying, distribution, or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender via telephone or return e-mail.
To unsubscribe from the FSL list, click the following link: