Hi AFNI team, I'm using 3dRSFC and had two things I'd appreciate help with. Let's say I'm using 3dRSFC -ort motion_file.1D -band 0.01 0.15.
- Is ALFF automatically calculated in the 0.01-0.1Hz range (without user input), irrespective of the specified -band option? That is, despite bandpassing at 0.01 0.15, ALFF still represents power in the 0.01-0.1Hz range?
- Just to clarify, the outputted RSFC_LFF file represents what many refer to as the 'denoised' time series? That is whatever signal is left after bandpassing and removing variance associated with motion_file.1D. Analogous to the output of 3dTproject (though I know processing steps differ between these functions)?
Greatly appreciate any advice!
Howdy-
To your questions:
-
No, there isn't an automatic bandpass range for 3dRSFC. What the user provides with -band .. is the bandpassing that is applied, and defines the range for amplitude of low frequence fluctuation (ALFF) estimation.
-
The output *_LFF* dset is a processed version of the input that has been bandpassed, likely detrended, and then also possibly had other things regressed out (as in your case above). Necessarily calling that "denoised" depends how confident you are that those steps have removed all noise, so perhaps is a bit philosophical to answer. But yes, analogous to 3dTproject, indeed.
Note that using 3dRSFC assumes that your input time series have not been censored/scrubbed, because that would effectively create non-uniformly sampled data. The standard frequency processing in that program uses the Fourier Transform, which requires uniform sampling. If you have censored/scrubbed data, then you should use a combination of 3dLombScargle+3dAmpToRSFC instead. (This might not be relevant, just thought I'd mention it, in case.)
--pt
1 Like
Hi Paul, appreciate the response!
For the same data, we're running two analyses. Analysis #1 is run later and uses the output time series (_LFF.nii) that's been bandpassed at 0.01-0.15Hz. Analysis #2 uses ALFF calculated in the 0.01-0.1Hz range. In this instance, I assume we cannot achieve both simultaneously with 3dRSFC. That is, I cannot specify separate parameters for the bandpass filtering vs the band at which ALFF is calculated. Instead, we would need to run it twice, once with -band 0.01-0.15 (analysis #1) and once with -band 0.01-0.1 (analysis #2). Is that correct?
Thank you!
Pete
Indeed, I think it would be best to run the analysis twice, generating the separate LFFs (from un-bandpassed data) in separates call to 3dRSFC: once for 0.01-0.10 Hz, and a second time for 0.01-0.15 Hz.
Trying to generate the smaller-bandpass-range LFF from the larger-bandpass-range one is possible in theory, but seems like it could lend itself to trimming issues. Also, if you are at all interested in fALFF, then you would be unable to get that in each case (unless you ran them separately).
--pt
1 Like