Nicola,

You should not need to re-run your existing baseline 'cross-sectional' subjects (that is, the subjects already processed by the default recon- all stream in v4.0.3). That data will be used to initialize the subcortical and cortical data used for the necessary 'longitudinal' baseline subject.

For the new longitudinal analysis scheme, you will first need to create a new baseline (timepoint 1) 'longitudinal' subject (following the instructions in the longitudinal processing section of recon-all -- help), this way:

recon-all -all -long tp1subj tp1subj -i path/to/tp1subj/inputs

for example, lets say your existing processed baseline subject is named nick10, and has two input files (which were already converted to mgz) the command would be:

recon-all -all -long nick10 nick10 \ -i $SUBJECTS_DIR/nick10/mri/orig/001.mgz \ -i $SUBJECTS_DIR/nick10/mri/orig/002.mgz

this will create a new baseline longitudinal subject named:

nick10.long.nick10

which will be the subject data considered 'timepoint 1' against which you will compare future timepoints. This recon-all command will take the aseg file, and white and pial surfaces from your existing 'nick10' data, and create nick10.long.nick10 subject data, processed through the same longitudinal stream as your future timepoint data will be processed. By doing this (by having this baseline longitudinal subject), it reduces the small variability that would otherwise be introduced by the difference in processing of the longitudinal stream versus the default cross-sectional stream.

Then, your timepoint 2 data would be processed this way:

recon-all -all -long nick10 nick10_tp2 \ -i path/to/nick10_tp2/dicom1 \ -i path/to/nick10_tp2/dicom2

which will create the subject:

nick10_tp2.long.nick10

and you would compare the aseg or aparc stats of nick_tp2.long.nick10 with nick10.long.nick10. Note that you do not need to process timepoints beyond timepoint 1 (the baseline) using the default (cross- sectional) recon-all stream, only the longitudinal stream is necessary. That is why, in the timepoint 2 command shown above, it is pointing directly at the source structural dicoms for 'subject' nick10_tp2 (which doesnt exist, and is only a placeholder name in this case). This timepoint 2 longitudinal command will also take the aseg and surface data from the baseline cross-sectional 'nick10' subject, and use it as initialization data (which will get 'warped' toward whatever change happened between time 1 and time2).

Hope this is clear. Probably not, so feel free to send along further questions.

Nick

On Thu, 2008-10-16 at 18:21 +0100, Nicola Hobbs wrote:

Hi there,

We would like to do some longitudinal processing. Our baseline scans have been run through v4.0.3, and we spent a long time manually editing the cortical segmentations. Ideally we would like to use these segmentations as the reference for our longitudinal processing (v4.1.0), but we are worried that using different versions for the same study could introduce errors. Could you advise us on whether we need to re-run our baseline data through v4.1.0 before we attempt any longitudinal processing?

Many thanks for your help, Nicola --

Nicola Hobbs

Dementia Research Centre National Hospital for Neurology and Neurosurgery Queen Square London WC1N 3BG

+44 (0)845 155 5000 ext. 723839

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer