help > SPM low pass filter information?
Showing 1-2 of 2 posts
Feb 18, 2020 02:02 PM | Sarah Meier - University of New Hampshire
SPM low pass filter information?
My team and I are trying to figure out what the default low pass
filter threshold is for SPM fMRI analyses (if there is one ~ most
searches have come up short) -- we know that the high pass filter
is .0078 Hz.
We thought that CONN may have a code for this, since it is a connectivity-based software, or some users may know from previous reports...
Thanks for the help.
We thought that CONN may have a code for this, since it is a connectivity-based software, or some users may know from previous reports...
Thanks for the help.
Feb 18, 2020 04:02 PM | Alfonso Nieto-Castanon - Boston University
RE: SPM low pass filter information?
Hi Sarah,
Unlike high-pass filtering, SPM does not implement any form of low-pass filtering of the BOLD signal at all. In SPM, frequency-specific properties of the BOLD signal are typically addressed by using an autoregressive AR(1) model to fit the observed level of temporal autocorrelation of the first-level model residuals (and this information is then used as part of ReML when estimating task-related responses, and also when computing single-subject statistics).
In any way, if you need to band-pass filter your data, you may do that, for example, using CONN's syntax:
conn_module('preprocessing','steps',{'functional_bandpass'},'bp_filter',[0.008, 0.09],'functionals',{'/mydata/func.nii'});
That will create a new file (prepending 'b' to the original filename) with the band-pass filtered data (see "help conn_batch" and "help conn_module" for additional details)
Hope this helps
Alfonso
Originally posted by Sarah Meier:
Unlike high-pass filtering, SPM does not implement any form of low-pass filtering of the BOLD signal at all. In SPM, frequency-specific properties of the BOLD signal are typically addressed by using an autoregressive AR(1) model to fit the observed level of temporal autocorrelation of the first-level model residuals (and this information is then used as part of ReML when estimating task-related responses, and also when computing single-subject statistics).
In any way, if you need to band-pass filter your data, you may do that, for example, using CONN's syntax:
conn_module('preprocessing','steps',{'functional_bandpass'},'bp_filter',[0.008, 0.09],'functionals',{'/mydata/func.nii'});
That will create a new file (prepending 'b' to the original filename) with the band-pass filtered data (see "help conn_batch" and "help conn_module" for additional details)
Hope this helps
Alfonso
Originally posted by Sarah Meier:
My team and I are trying to figure out what the
default low pass filter threshold is for SPM fMRI analyses (if
there is one ~ most searches have come up short) -- we know that
the high pass filter is .0078 Hz.
We thought that CONN may have a code for this, since it is a connectivity-based software, or some users may know from previous reports...
Thanks for the help.
We thought that CONN may have a code for this, since it is a connectivity-based software, or some users may know from previous reports...
Thanks for the help.