Hi Peter,

FEAT doesn't provide low-pass filtering options because this should not be used for the standard task-FMRI analyses that FEAT is designed for.  The modelling and statistics in FEAT (e.g. FILM) require and enforce a spectrally flat noise profile.  The high-pass filtering is only for the lowest frequency drifts that are effectively poorly sampled.  Furthermore, the algorithm used is not directly applicable to low pass filtering.

Cheers,

Eugene

On 31 May 2012 22:11, Peter Fried <[log in to unmask]> wrote:

Hi FSL group,

I am trying to understand why FEAT handles low-pass temporal filtering differently from high-pass filtering. For example, if you edit the design.fsf file to enable low pass temporal filtering:

       # Lowpass temporal filtering
       set fmri(templp_yn) 1

Feat uses a fixed sigma of 2.8 sec when calculating the sigma in volumes. This can be seen in the $FSLDIR/tcl/featlib.tcl:

       set lp_sigma_sec 2.8
       set lp_sigma_vol [ expr $lp_sigma_sec / $fmri(tr) ]


Is there a reason Feat doesn't allow you to set the temporal window for low pass filtering? There is no option to set the temporal window in either the gui or by editing the design.fsf file (I tried adding the following line):

       # Low pass filter cutoff
       set fmri(paradigm_lp) 8

but this was ignored by featlib.tcl.

I understand I can edit the featlib.tcl (which I am loathe to do) or do it after the fact by running "fslmaths -bptf>, but I would like to understand why Feat handles low-pass filtering differently. Thanks.

Cheers,
Pete

P.S. I understand that there is a debate as to whether or not to apply a low-pass filter to resting-state data



On May 23, 2012, at 1:58 PM, Peter Fried wrote:

> Thanks!
>
> On May 22, 2012, at 9:27 PM, bettyann wrote:
>
>> To be internally consistent with other FSL conversions from FWHM to sigma, you should probably use 2.
>>
>> See this post for gory details (and my opinion based on empirical observation):
>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=fsl;9ffadc97.1109
>>
>> Best,
>> - BettyAnn