Dear Donald,
thank you very much for pointing me to this error in my processing pipeline.
If I reslice at the realignment step and not again at coregistration, then I guess the coreg-info in the header will automatically be taken into account when warping the EPIs with the DARTEL deformations, right? But will the results from "1st-level analyses" of the un-resliced coregistered EPIs (the FC maps) also be aligned with the native skullstriped T1-image, allowing them to be warped with the DARTEL deformations (native T1 -> MNI space)?

Again, thank you very much for your very helpful input.

Best,
Michel   

Date: Wed, 22 Dec 2010 09:28:52 -0500
Subject: Re: [SPM] co-registration in SPM5: how to make best of it?
From: [log in to unmask]
To: [log in to unmask]
CC: [log in to unmask]

My understanding of SPM is that any mathematical operation requires that data be resliced as SPM operations are performed on voxels (e.g. all time points at voxel 1, .. all time points at voxel N) rather than a specific place in space (e.g. all time points at x1,y1,z1; all time point at x1,y1,z1). Thus, I'd be concerned about step 3 if you don't write out the data. The same concern is true for point number 9; however, if you reslice in step2, you don't need to reslice in step 6 because your not changing the relationship of voxels between slices, so the voxel alignment is maintained. This only works for affine transformation. Anything that is non-linear or requires a deformation requires that the data be written out as the information can't be stored in the header -- one caveat being that you can multiply deformation fields together now, so multiple deformations can be done with a single reslicing step.

Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Postdoctoral Research Fellow, GRECC, Bedford VA
Research Fellow, Department of Neurology, Massachusetts General Hospital and Harvard Medical School
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is intended only for the use of the individual or entity named above. If the reader of the e-mail is not the intended recipient or the employee or agent responsible for delivering it to the intended recipient, you are hereby notified that you are in possession of confidential and privileged information. Any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited and may be unlawful. If you have received this e-mail unintentionally, please immediately notify the sender via telephone at (773) 406-2464 or email.


On Wed, Dec 22, 2010 at 6:31 AM, michel grothe <[log in to unmask]> wrote:
Hi all,
sorry that this seems to be a never-ending story, but now I´ve got a doubt about reslicing as well:

> Reslicing after realignment / coregistration may be necessary if you are
> 1) Applying realignment before slice time correction
> 2) Analysing in native space, so there is no spatial normalisation step
> 3) Using realignment with unwarping

In step 2), does this mean that the header information is not considered in SPM stats (and slice-time correction in step 1)? What if temporal filtering is set between realignment and spatial normalisation?
I think I remember some words along the line "reslicing is virtually only necessary if you´re using Nifti-incompatible software in further processing steps". So I thought that SPM always considers the header-information (like in Display/CheckReg) and therefore there is no need to reslice. Did I get that wrong?
The SPM manual states that reslicing should be used with unwarping because of interpolation issues. How about the DARTEL-normalisation of fMRI data?

It would be great if someone could comment on my processing pipeline for rs-fMRI, which I adopted from various posts on this mailing list:

1. slice-time correction
2. realignment, only mean image resliced
3. temporal filtering (0.1-0.01 Hz), using REST-toolbox (SPM5-routines?)

4. bring structural in rough alignment to MNI space with reorient-script found in the archives (improve starting estimates)
5. Segment the structural and DARTEL warp directly to MNI space using VBM8-toolbox; calculate native skull-stripped T1 with ImCalc nativeT1.*((GM+WM+CSF)>0.5)
6. coregister (estimate only) mean_functional to native skull-stripped T1, move along all slice-time corrected, realigned, temporally filtered EPIs
7. Warp coregistered EPIs using deformation from DARTEL-normalisation of T1-image
8. Extract ROIs for Seed-based functional connectivity from warped EPIs using masks in standard space

9. Run multple regression with time-courses of ROIs on slice-time corrected, realigned, temporally filtered, coregistered EPIs from point 6 (NOT warped) to obtain FCmaps 
10. convert FCmaps to z-maps, warp z-maps using deformation from DARTEL-normalisation of T1-image (as in step 6).
11. Smooth the warped z-maps and run 2nd-level random effects analysis

Following the previous post I am now worried about processing step 9. The images are resliced after the realignment at the temporal filtering step, but what about the co-registering to the T1-image which is only estimated (written to the header)?

Any advice would be highly appreciated.
Best,
Michel

> Date: Wed, 22 Dec 2010 10:34:33 +0000
> From: [log in to unmask]
> Subject: Re: [SPM] co-registration in SPM5: how to make best of it?
> To: [log in to unmask]

>
> Hi John,
>
> Just to add a little to John A's answer:
>
>
> On Tue, 21 Dec 2010 21:26:58 +0200, John Gelburg <[log in to unmask]> wrote:
>
> >Hi,
> >
> >Thank you all for the help.
> >
> >So, if I extend what Paul wrote, the prescription looks as follows:
> >
> >Realign: Estimate & Reslice (Mean image only)
> >Coreg: Estimate; source=SPGR, reference=EPI
> >Coreg: Estimate; source=SPGR, reference=T1, other images = all EPIs
> >(including mean)
> >Segmentation
> >Normalize: Write; parameter file = *_sn.mat, Images to Write = EPIs, voxel
> >size = [3,3,3]
> >Normalize: Write; parameter file = *_sn.mat, Images to Write = SPGR, voxel
> >size = [1,1,1]
> >
> >I just wanted to clarify few points in order to understand them
> >conceptually:
> >1. What is the role of realignment reslice for EPI images?
>
> Usually, we avoid reslicing because it involves an interpolation of the data. Instead, the transformations from realignment and coregistration are applied to the header of the NiFTI image, and included in a single reslice when the normalisation parameters are written. Reslicing after realignment / coregistration may be necessary if you are
> 1) Applying realignment before slice time correction
> 2) Analysing in native space, so there is no spatial normalisation step
> 3) Using realignment with unwarping
> http://imaging.mrc-cbu.cam.ac.uk/imaging/MotionCorrection
> http://imaging.mrc-cbu.cam.ac.uk/imaging/SliceTiming
>
> >All the images
> >have the same voxel size. What exactly SPM reslices here and why I need to
> >do it unless I do normalization afterwards? Is coreg reslice is virtually
> >the same thing as realign reslice?
> >
> >2. As some of you have pointed out that since my mean image had already been
> >realigned with other EPI images, after I coregister SPGR source to mean
> >image reference I do not have to take along the other EPIs. Does it mean
> >that what happened is that after finding transformation parameters they were
> >applied to the EPI images while no transformation should have been made?
>
> Yes
>
> >
> >Many thanks again,
> >John
> >
> >
> >
> >
> >
> >On Tue, Dec 21, 2010 at 1:36 PM, Paul Wright <[log in to unmask]> wrote:
> >
> >> Hi John,
> >>
> >> The reason your images are misaligned after coregistration is that you got
> >> the source and reference images the wrong way around: you move the SPGR
> >> source to fit the mean EPI reference, and the apply the transformation to
> >> the other EPIs. This mean that
> >> a) the structural and EPIs will still be in the same relative positions: if
> >> they began unaligned, they will stay unaligned
> >> b) the EPIs will no longer be in register with the mean EPI
> >>
> >> I think you wanted to move the mean EPI source to fit the SPGR reference,
> >> and then move the other EPIs along with it. However, as others have already
> >> said a more straightforward approach would be to coregister the SPGR with
> >> the mean EPI (leaving the other EPIs alone). You can subsequently use your
> >> normalisation parameters on both sets of images.
> >>
> >> This approach holds also using Marko's additional initial coregistration of
> >> the structural to the T1 template, which is intended to improve
> >> segmentation. This would go in between the SPGR-EPI coregistration and
> >> normalisation:
> >> 1) Realignment
> >> 2) Coregister SPGR source to EPI reference
> >> 3) [Optional] Coregister SPGR source to T1 template reference, taking along
> >> EPIs as other images
> >> 4) Normalisation / segmentation
> >> 5) Write normalisation parameters for SPGR and EPIs.
> >>
> >>
> >> Hope this helps,
> >> Paul
> >>
> >