Hi Tamara,
there should be no scaling in a longitudinal study. If there is, that is a problem, which needs to be investigated. Where does it come from? Switching scanners, head coils, or (less likely) software upgrades or incorrect scanner calibration could cause this. You need to find out if it affects all participants, or only some, when this change happened (at what day), to find out which subject's time intervals would be affected.
Freesurfer does not test or report such scaling. It is also difficult to find out if there really is scaling. If you -a) visually see that there is scaling after rigid registration (especially for small scaling this is hard to tell) -b) and find a scaling from the output of the affine registration -c) and looking at affinely mapped results it seems like the scaling from a) is gone
Then you can be pretty sure that there is a problem with scaling. The challenge is that simply running b) and looking for a scaling value is not really sufficient as spurious scaling in the affine registration could be caused by atrophy in the images or motion or what not.
If - all time points are scanned with the same scanner and hardware (head coil) - there was no software upgrade - scan sequence / protocol is identical for all time points
you can be pretty sure that there is no real scaling, unless someone did a scanner calibration in the meantime and messed it up. You should be able to find out if during your study - hardware was changed - software was upgraded - callibration(s) were done You should also be able to check if protocols are identical for both time points. If all that is OK, I would not remove any individuals just because of some % scaling from b) above. If all a)b)and c) are true and you are pretty sure there is real scaling you need to do more detective work to find out where it is coming from.
You also mentioned a brightness difference between time points. This could also indicated protocol of scanner differences.
Best, Martin
On 12 Jan 2017, at 16:20, Martin Reuter mreuter@nmr.mgh.harvard.edu wrote:
Hi Tamara,
that information is in the screen output:
Determinant : 1.03358
also Scale = diag([ 1.0109774487584 1.0113715685396 1.0108915357302 ])
so it finds some scaling (approx 1% in each axis direction, so a total of 3%). This can happen in individual cases, but if it is consistent I would be worried.
Best, Martin
On 11 Jan 2017, at 21:21, Tamara Tavares <ttavare@uwo.ca mailto:ttavare@uwo.ca> wrote:
Hi Dr. Reuter.
Thank you again for your help.
I ran the following command: "mri_robust_register --mov rawavg_1.mgz --dst rawavg_2.mgz --lta v1to2.lta --affine --satit" on the two rawavg.mgz images; one from each time point from the cross-sectional outputs. I have attached the information from the terminal and the output. Are these outputs appropriate to find the scaling information? If so, where can I find it? I am not very familiar with transformations, thus any information or directions to resources I can use would be very much appreciated.
Thank you in advance for you help.
Best, Tamara
Message: 6 Date: Tue, 10 Jan 2017 21:18:34 +0100 From: Martin Reuter <mreuter@nmr.mgh.harvard.edu mailto:mreuter@nmr.mgh.harvard.edu> Subject: Re: [Freesurfer] Brighter image at baseline vs. follow up using longitudinal processing stream To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu mailto:freesurfer@nmr.mgh.harvard.edu> Message-ID: <9BCB373D-D368-4B36-B607-69E6A14AF952@nmr.mgh.harvard.edu mailto:9BCB373D-D368-4B36-B607-69E6A14AF952@nmr.mgh.harvard.edu> Content-Type: text/plain; charset="us-ascii"
Hi Tamara,
mvoing in and popping out sounds like a scaling problem. We do only rigid transforms, so the scaling would already be in your original inputs. If that is the case, it could mean that there was a scanner calibration or something that caused scaled images and that would not be good. It should however appear in all images that have the calibration in between time points.
You can try to use mri_robust_register with the affine flag to see if any scaling is present in the rawavg.mgz images between the two time points (it should report scaling separately on the screen output), else use lta_diff with the appropriate flags. You can also look at the affininely registered images and see if it looks better.
Anyway if there is global scaling in the images, I would try to investigate a where this comes from and also drop that image from the study.
Best, Martin
<terminal_output_mri_robustcommand.txt><v1to2.ita.txt>_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer