Hi Alex,
On 12/06/2012 12:21 PM, Alex Kell wrote:
freesurfers,
i've been using my own scripts to compute tSNR for data i was running through FSL, but i'm now running the data through freesurfer and i wanted to better understand what the fsnr volumes spit out by selxavg3-sess are. i poked around the wiki and mailing archives and couldn't find answers to the following questions:
- how is fsnr computed? tsnr is
<mean_of_vox_time_course>/<std_of_vox_time_course> -- is fsnr the same thing with a different name?
Pretty close, though maybe a little different. It is the offset/std_of_residuals. This all happens after the model is fit. The "offset" here is close to the mean but technically it is the regression coefficient of the regressor of all 1s. After the full model is fit, the estimated time course is subtracted from the actual time course to give the residuals. So this is the Function SNR after removing task signal and nuisance regressors.
- what kind of preprocessing is done on the data before fsnr is
computed? since freesurfer removes linear trends at the modeling stage, not the preprocessing stage, are linear trends removed from the data going into fsnr calculations? is anything else regressed out before fsnr is computed (e.g., other nuisance regressors)?
The full model is regressed out.
- what's the difference between raw.fsnr.nii.gz, fsnr.nii.gz, and
fsnr2.nii.gz?
The raw fsnr does not regress out anything just mean/std
- is fsnr.dat the mean in the whole brain mask (i.e., gm and wm)?
Yes
separately, random other question: what are the h.dat, h.nii.gz, and h-offset.nii.gz file/volumes?
these are somewhat redundant with the beta volume. In the pre-5.0 version, these were the only files produced. They are created for compatibility with some of the older tools.
doug
thanks in advance!
alex kell
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer