 |
 |
Hi, Jesper Thank you very much for your comments on these questions. I admit that the sample was not prepared well and there is still large space for improvement in our next experiment. The speck issue you pointed out is a valid one and we will take that into consideration (by saving the data with 32bit format?) and matching the susceptibility between the brain and surrounding materials. Thank you again! I will process the data as you suggested and keep you updated on that. Longchuan On Friday, November 8, 2013 4:14 AM, Jesper Andersson <[log in to unmask]> wrote: Dear Longchuan, (1) It seems to you that the mismatch between the corrected b0s and T2w could be due to the large signal void at the globus pallidus (possibly a result of iron deposition). We have scanned macaque DTI data in vivo with the similar signal void and we also used "topup" for correcting the distortion. However, topup worked really well on these data and we never had such problems. This made me suspect that the largely "low spatial frequency" distortion may be caused by some other things. I contacted Bruker engineer and according to him, although the distortions between the two scans with opposite phase encoding directions will be symmetrical, the "regrid" process will be identical in the two scans. I wonder how the distortion will look like due to that and we will keep it in mind in our next experiment. I wasn't thinking about the globus pallidus. That void is similar/identical in both scans and should hence not cause problems. I was thinking about the rather small black specks in some of the gyri that aren't present in both scans. Come to think of it, they actually reminded me of "binary wraps", i.e. values that are too large to be represent by your choice of data type. I may well be wrong, but it might be worth looking into. > >(2) If I use the undistorted T2w as the "fake" b0up and use the distorted b0 image as the b0dn, will that be largely equivalent to nonlinearly register the distorted b0s to an undistorted T2w image using "fnirt"? (since I acquired all the data in only one phase encoding direction and several b0s with the opposite phase encoding direction). If not, what are the possible advantages? The advantage is that topup will only "allow" displacement along the PE axis, which is consistent with what we think is happening. With fnirt there is always a risk of spurious displacements along the other directions. > >(3) Do you have any suggestion on how to tune up the parameters in the .cnf file for postmortem brains with the spatial resolution of 0.5mm isotropic (smaller than the minimal 1mm in the cnf file)? > > I don't think you need to worry about that. I think 1mm is already overkill, it's mainly there "because I can". You don't need to match the image resolution. The relevant thing is the "resolution" of the field and that will be relatively smooth. Jesper Sorry for so many questions and I really appreciate your help on them. > > >Longchuan > > > >On Thursday, November 7, 2013 12:28 PM, Jesper Andersson <[log in to unmask]> wrote: > >Dear Longchuan, > > >Hi, Jesper >> >> >>I have uploaded b0up (data_b0s_up.nii.gz), b0dn (data_b0s_dn.nii.gz) and undistorted T2w (T2_x180deg) images. The orientations are not completely correct as I am just trying to see if the current protocol will work. I doubt that the residual distortion is due to eddy current: they occur in all images. The eddy current + residual gradients induced distortion and shift across volumes are definitely more severe than the images acquired by the spin-echo dti sequences. However, they can be adequately corrected using eddy_correct/Eddy. >> >> >>Currently, I am trying figure out the causes of the residual distortion in the b0 image. I think mostly the Bruker used hybrid 3D EPI, i.e., they acquired slabs along the z-directions with x and y as the phase encoding directions. But I need to confirm that. Another possibility is that since Bruker does not allow to change the phase encoding directions in 3D EPI, we acquired two b0s by having two studies with head/feet positions switched. I am wondering whether that will contribute to the unidentical local field inhomogenieties between two scans. > > am I right in assuming that these are scans on excised brains? I think the problems that topup has is because there are signal dropout (black lacunae) in the images that aren't always present in both scans. Topup will the try to "fix" that by squeezing the lacuna together to remove it. > > >After having seen the original images I was actually quite please to see how relatively "un-horrible" the corrected images were, though you are right that the match to the T2 image is not particularly good. > > >Your T2 image has a very similar contrast to your EPI images so I think your best bet is to not match the two EPI images together, but instead "trick" topup to match them directly to the T2 image. > > >What you need to do is to run topup twice, once for each EPI image. For both runs the "second scan" should be the T2 image. In your --acqp file you specify the EPI image as before, but for the T2 image you put in a very small readout time, say for example 0.0001. That tricks topup into thinking you have two EPI images, one with appreciable distortions and the other one with almost no distortions at all. > > >This means you will end up with two slightly different fields, one for the up and one for the dn images. This is turn means that you will need to apply the topup (susceptibility) correction separately for the two directions before feeding them into eddy (without specifying a --topup). > > >Try it and please let me now how it goes. > > >Jesper > > > > > >> >>Please let me know if you have any comments >> >> >>Thank you! >>Longchuan >> >> >> >>On Wednesday, November 6, 2013 5:34 AM, Jesper Andersson <[log in to unmask]> wrote: >> >>Dear Longchuan, >> >> >>could you please send me two b=0 volumes acquired with the two different blip-directions so I can have a look at them. Please use >>https://oxfile.ox.ac.uk/oxfile/work/extBox?id=72139C7E463068D5F >>for the upload. >> >> >>Also, it is not clear to me if your problem is uncorrected susceptibility or eddy-current distortions. Do you see the residual distortions in the b=0 images or in the diffusion weighted images? When you look across different diffusion weighted images, do they look like they have different distortions? >> >> >>Jesper >> >> >> >> >>On 5 Nov 2013, at 20:52, Longchuan Li <[log in to unmask]> wrote: >> >>Hi, Jesper >>> >>> >>>Sorry for the late reply as the preprocessing takes some time. >>> >>> >>>Thank you very much for your information. which is very helpful. I think the two cases you mentioned are true echo volume imaging and 3D EPI. I am still in the middle of investigating the issue, but my initial "topup" results show that the majority of the distortion is corrected, i.e., the large local distortion due to air bubbles, brain-air interface, but when I registered the corrected images to the T2w image from the same subject that is free of distortion, there is still some residual spatial distortion left uncorrected. This residual distortion seems to be more low spatial frequency concentrated, i.e., they look almost like the brain is squeezed and sheared along x-y and y-z planes and seem not to be symmetrical in the two original acquisitions with the opposite phase encoding directions (see attached figure). I would appreciate if you would let me know your thoughts on this residual spatial distortion. >>> >>> >>>Also, since the spatial resolution of the image is 0.5 mm isotropic, I wonder whether you have additional suggestions regarding the parameters in the cnf file for topup. >>> >>> >>>Many thanks and I will keep you updated regarding the further data analyses. >>> >>> >>>Longchuan >>> >>> >>>PS: the image is blocked because it exceeds the size limit. >>> >>> >>> >>>On Monday, November 4, 2013 5:31 AM, Jesper Andersson <[log in to unmask]> wrote: >>> >>>Dear Longchuan, >>> >>> >>>We are planning to use 'topup' to correct the diffusion data of postmortem brains acquired using segmented 3D EPI sequences. Since there are two phase encoding directions, I feel the distortion will not be confined to each axial slice (am I right that the distortion will be at 45 degree from the axial slices?). So, my question is whether I could use "topup" to correct for the distortion between two averages of data with completely opposite two phase encoding directions? What other points that I need to pay attention to when I am using "topup" to correct for such data? >>> >>> having two phase encode directions may mean having distortions along two directions, but not necessarily. It depends on the specifics of how you acquired your data. Distortions "only" occur when you have multiple phase-encode steps following a single excitation. If you have another phase-encode direction where the phase only changes between excitations you wouldn't get any distortions in that direction. >>> >>> >>>I'm assuming you have the frequency encode in the z-direction so as to avoid wrap-around of the neck/body? If so each shot will presumably consist of a "slice of k-space" in a direction that is parallel with the z-axis. If each shot is a "coronal slice of k-space" your distortions would be in the left-right direction. If each shot is a "sagittal slice of k-space" they would be in the ant-post direction. So in either of those cases I think you would only see distortions in one direction. >>> >>> >>>Please get back to me if your acquisition is different from what I have assumed and let me know some more details so I can help you. >>> >>> >>>Jesper >>> >>> >>> >>> >>> >>> >>> >>>> >>>>Many thanks in advance! >>>> >>>> >>>>Longchuan >>> >>> >>> >> >> >> > > >