Dear Rashmi,
> I have some questions about how I can set up the First Level Analysis
> (seed-based connectivity analysis) for resting state FMRI data using the
> GUI. As I see it, this is how I need to go about it:
> 1. Preprocess the data (slice timing correction, realignment,
> coregistration, normalisation, smoothing)
> 2. Extract time series from a given seed
> 3. Use that as a regressor in the GLM
>
> My questions are as follows:
> 1. How should I extract the time series from a given seed mask? I have
> used fslmeants in FSL for this before, but I wanted to know what SPM
> script I need to use to get this.
If you want to compute the mean (raw) time series within a mask, you can
simply use spm_summarise:
>> F = spm_select(Inf,'image');
>> Y = spm_summarise(F,'mask.nii',@mean);
If you also want to temporally filter the time series, adjust for other
covariates and use the first eigenvariate of a local PCA instead of the
mean, you could specify a GLM (e.g. using a DCT basis set as mentioned
recently on this list) and use the "eigenvariate" button when looking at
results or use the batch interface:
SPM > Util > Volume of interest
The time series with be stored in xY.u in the VOI.mat file.
> 2. In my GLM, what should I input as the conditions? Can this be left
> blank? Or should it be Onset:0 and Duration:till the end of the scan
SPM estimation should run fine even if you didn't specify any condition.
> 3. How does SPM differentiate between regressors of interest and nuisance
> variables? For example, where do I need to specify the time series of my
> seed, and where should I specify movement parameters and maybe White
> Matter and CSF-associated timeseries as confounding variables?
There is no particular distinction between them and their order in the
design matrix does not matter. Just include all these explanatory
variables in the "regressors" or "multiple regressors" entries of the
model specification batch interface.
[if you wanted to be very precise, as it is used for the correction for
temporal autocorrelation, you could manually adjust the indices in
SPM.xX.iC and SPM.xX.iG from the SPM.mat file or provide an F-contrast
defining regressors of interest in SPM.xVi.Fcontrast]
> 4. How should I define the contrasts taking these regressors into account?
Just do a t/F-contrast over the regressor encoding your seed time series
[0 ... 0 1 0 ... 0].
Note that there are a number of toolboxes for SPM that might streamline
this type of analyses:
http://rfmri.org/DPARSF
http://web.mit.edu/swg/software.htm
http://mialab.mrn.org/software/
Best regards,
Guillaume.
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|