Hi M,
You can create a set by merging scans of the same subject across timepoints, but that set would not be considered independent.
Independence in this context means the occurrence of one element would not have any effect on any other element in the set, but this would not hold in this case.
https://en.wikipedia.org/wiki/Independence_(probability_theory)
But if you decide to do this anyway, and have no intention of doing any sort of longitudinal analysis across timepoints, there would be no need to run the longitudunal pipeline.
-Paul
________________________________________ From: freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edu on behalf of JM mj70481@gmail.com Sent: Saturday, November 9, 2024 10:27 PM To: Freesurfer support list Subject: [Freesurfer] Fwd: recon all data for classification between groups !
External Email - Use Caution
Dear Freesurfer experts/developers,
I would greatly appreciate your help/response with the following questions/concerns.
---------- Forwarded message --------- From: JM <mj70481@gmail.commailto:mj70481@gmail.com> Date: Wed, Oct 30, 2024 at 11:09 AM Subject: recon all data for classification between groups ! To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu>
Dear FS experts,
I have longitudinal data:
Three time-points [T1-T3] with an interval of about 6 months between T1 and T2 and 6 months between T2 and T3 from a control group as well as from a clinical group (again three time-points [T1-T3] with an interval of about 6 months between T1 and T2 and 6 months between T2 and T3).
Sample sizes are: - For control group: 48 (for T1), 36 (for T2) and 33 (for T3) - For clinical group: 38 (for T1), 33 (for T2) and 31 (for T3)
I was wondering if it's fine to consider and combine three time-points within each group as independent samples for classification purposes (i.e., N = 117 (48+36+33) for the control group and N = 102 (38+33+31) for clinical group) between the two groups (based on brain morphometry measures). If so, I was wondering whether the recon-all should be done in the longitudinal way (MailScanner has detected a possible fraud attempt from "secure-web.cisco.com" claiming to be https://surfer.nmr.mgh.harvard.edu/fswiki/LongitudinalProcessinghttps://secure-web.cisco.com/1dNOHCl2MSH_yKn1dLwqkg767H0z716kQLlx7ce7bPg1Ozkk1dPR7PNmtQlIFH4EwEsYAY0-kP691owcbYQ4vBCwtWxBTq7rvvxlMYOHU2QlaYlBKpce-YUIpq7qTVlDPYe8qhx5laHqD6upsfM29ejeLktvhvP-smjf1v7J0RxcCMxOuEPSbnaDg1YBgCVyNaqvSqscpXpZ3AScvpo8r7XXqciHERsnyi3gAzZtcOBErz8NjdIjiKdWbxp8lAKUnVkDL_wQFTMrGkDm95oNfXsOM4byUWEx5Y0IBhHgrMDE2kEZus3Orev6p2uC6cFdewGYpy0LKNl7WR-qPKu8RDw/https%3A%2F%2Fsurfer.nmr.mgh.harvard.edu%2Ffswiki%2FLongitudinalProcessing) i.e., making it unbiased towards a specific time-point, OR, the recon all should be run cross-sectionally so that we can keep all the time-points independent from each other.
Any detailed explanation/justification would be really helpful. Also, I would greatly appreciate the response at earliest convenience.
Thanks, M