External Email - Use Caution
Thank Martin,
I've been looking into this and it appears that the slight changes in the sequence may have occurred software updates to the scanner. So essentially from what you say, this drastically impacts any longitudinal analyses as there will be a systematic bias across the data. Is there any point in trying to do longitudinal analyses with such data as it will be a factor in all output? Is it worth resampling all three time points to 1mm and then rerun the analysis as it will be difficult to say if any patterns of change arent due to the acquistions as you said?
If I want to resample the data to a "standard 1mm3" voxel size, is there a FS command for that or should I use something in FSL or AFNI tools and then feed the resampled data back to the recon-all inistial step in FS.
Thanks for the help and guidance. Much appreciated
Erik
________________________________ From: freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edu on behalf of Martin Reuter mreuter@nmr.mgh.harvard.edu Sent: 15 May 2018 23:39 To: Freesurfer support list Subject: Re: [Freesurfer] Longitudinal errors, Voxel size differs
Hi Erik,
you should investigate why this is the case. Any difference in acquisition will cause effects and it will be impossible to distinguish these from real effects.
Also you should never just resample one time point, as that will introduce processing bias. I would advise to - find out why, and you need to report this in any analysis so that people are aware of this potential bias - resample both time points to some 3rd space. Probably here simply conform them to 1mm isotropic.
Best, Martin
Erik O'Hanlon Postdoctoral researcher
[cid:rcsi-crest-signature_f581c185-1d03-45c8-a786-23d8ece3d391.png]
RCSI Psychiatry Royal College of Surgeons in Ireland Beaumont Road, Beaumont D9 Ireland T: 8093740 E: erikohanlon@rcsi.ie W: www.rcsi.comhttp://www.rcsi.com/
Transforming Healthcare Education, Research and Service: RCSI Strategic Plan 2018-2022http://www.rcsi.ie/strategy2018
RCSI DEVELOPING HEALTHCARE LEADERS WHO MAKE A DIFFERENCE WORLDWIDE
[cid:both3_b9da8717-e91f-4ea7-8729-dc2c4d0a7a5f.png]
On 15. May 2018, at 07:20, Erik O'Hanlon <erikohanlon@rcsi.iemailto:erikohanlon@rcsi.ie> wrote:
Hi FS Experts,
I'm trying to run a longitudinal analysis across three timepoints and have run the cross steps for the data. I then ran the base step but found that my voxel dimensions are slightly different across the timepoints and I got the following error
ERROR: MultiRegistration::loadMovables: images have different voxel sizes. Currently not supported, maybe first make conform? Debug info: size(2) = 0.881117, 0.881117, 0.881117 size(0) = 0.884507, 0.884507, 0.884507 MultiRegistration::loadMovables: voxel size is different /projects/pi-vokeane/HPC_16_00925/MC_FS_testdata_EOH/freesurfer/subjects/abd025_tp3/mri/norm.mgz.
I assume I need to resample the data to the voxel size of the first time point and then rerun. Can you advise on what command to use to resample the images from time 2 and time 3 to those of time 1 before rerunning from scratch. I assume I can't avoid rerunning the initial reconall step once I resample my data.
Any guidance is much appreciated
Best regards
Erik
Erik O'Hanlon Postdoctoral researcher <rcsi-crest-signature_f581c185-1d03-45c8-a786-23d8ece3d391.png> RCSI Psychiatry Royal College of Surgeons in Ireland Beaumont Road, Beaumont D9 Ireland T: 8093740 E: erikohanlon@rcsi.iemailto:erikohanlon@rcsi.ie W: www.rcsi.comhttp://www.rcsi.com/
Transforming Healthcare Education, Research and Service: RCSI Strategic Plan 2018-2022http://www.rcsi.ie/strategy2018
RCSI DEVELOPING HEALTHCARE LEADERS WHO MAKE A DIFFERENCE WORLDWIDE
<both3_b9da8717-e91f-4ea7-8729-dc2c4d0a7a5f.png>
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
