Hi again, Andrew.
If you run recon-all-clinical and then create the base, it won’t work because:
- Some of the images are inherently different (e.g., norm.mgz is a fake image in recon-all-clinical, check it out!)
- Even if you created the base, the subsequent longitudinal processing wouldn’t work.
Hopefully we’ll have a longitudinal version of recon-all-clinical at some point!
Cheers,
/E
--
Juan Eugenio Iglesias
http://www.jeiglesias.com
From:
freesurfer-bounces@nmr.mgh.harvard.edu <freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of Andrew Dedinas Geoly <ageoly@stanford.edu>
Date: Wednesday, May 31, 2023 at 15:40
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Subject: Re: [Freesurfer] recon-all-clinical Longitudinal workflow
External Email - Use Caution
Thank you for your quick response!
>From my understanding of the longitudinal processing stream, the individual recon-all (cross) outputs from each time point are used to create a base model that averages the cross outputs. Finally,
recon-all long is used for each of those cross outputs and the information from the within-subject base template.
My thought was that I could use reconall_clinical simply to create the (cross) outputs from the first step and then those would be used for step 2 and 3.
Would this be a feasible approach?
Best,
Andrew
From: freesurfer-bounces@nmr.mgh.harvard.edu <freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of Iglesias Gonzalez, Juan E. <JIGLESIASGONZALEZ@mgh.harvard.edu>
Sent: Saturday, May 27, 2023 8:41 AM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Subject: Re: [Freesurfer] recon-all-clinical Longitudinal workflow
External Email - Use Caution
Dear Andrew,
These are both great questions!
- The problem is that there’s no longitudinal version of SynthSeg / SynthSR yet so creating a base etc wouldn’t really affect the results…
- If the accuracy of the standard stream is not good in your data, this is probably the way to go.
Sorry for not having better tools at the moment, but we are in a transitional period in which we are releasing machine learning tools that only replace parts of the main stream and some functionalities are
not fully supported yet.
Cheers,
/Eugenio
--
Juan Eugenio Iglesias
http://www.jeiglesias.com
From:
freesurfer-bounces@nmr.mgh.harvard.edu <freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of Andrew Dedinas Geoly <ageoly@stanford.edu>
Date: Friday, May 26, 2023 at 23:38
To: freesurfer@nmr.mgh.harvard.edu <freesurfer@nmr.mgh.harvard.edu>
Subject: [Freesurfer] recon-all-clinical Longitudinal workflow
External Email - Use Caution
Hello Freesurfer experts!
I am conducting a longitudinal GMV and CT analysis of mTBI participants who received an experimental single-dose pharmacological treatment. We collected GE Bravo (0.9mm
isotropic) T1w data pre, immediately after (~ 4 days after intervention) and 1-month following the intervention.
I have been working with the recon-all longitudinal pipeline stream (7.1.1) but have noticed pretty significant errors in terms of skull stripping and segmentation (participants
are veterans with somewhat random lesion / atrophy patterns).
I have tested your new recon-all-clinical program (7.4.0) on some of my worst cases and it is performing remarkably well. I am doing some more digging into the various
synth segmentation packages available in FS newest versions and was wondering what the best course of action would be for integrating recon-all-clinical or synthseg into a longitudinal workflow.
1.Could I use the recon-all-clinical outputs as the "cross" for later base and long template construction?
2. SynthSeg and SCLimbic seem to have very good test-retest reliability-- would it be feasible to use the independent segmentation extractions from those and bypass the
longitudinal stream entirely (not saying I would want to but I am curious)?
Any and all help here would be greatly appreciated, and I thank you very much for your time.
Warmest regards,
Andrew
---
Andrew D. Geoly, MS
Research Data Analyst 1
Brain Stimulation Lab
Department of Psychiatry and Behavioral Sciences
Stanford University School of Medicine