Print

Print
Thank for your reply, but my results are contradictory to your explanation, but am not sure why.  
 
The high pass entry is in seconds - so the higher the entry the lower the Hz value as the boundary so say:
 
Option A:  High Pass 50 seconds - Hz 0.02 - Then all of the signal greater than 0.02  Hz is allowed through.
Option B:  High Pass 1000 seconds - Hz 0.001 - Then all of the signal greater than 0.001 hz is allowed through - which should be presumably more signal(s)
 
Yet, the FEAT output PowerSpectrum graph of the outputed ICA components from each run demonstrates much greater signal in many components in the higher frequency range (> 8 Hz/100) from option A than option B.
 
My understanding of resting state networks is that they are best found in the very low frequency range - 0.008-0.01 Hz or even less.
 
Perhaps the High Pass filter that is given in seconds - when inverted to filter the same setting in Hz - becomes a low pass filter - allowing only frequencies lower than 0.001Hz, for example? 
 
With option A, I have about 180 ICA components outputed - many of which look like regular sharp up and down pattern on the Normalized Response graph, and wide mix of power spectrum frequency range from 0 to 22.22 Hx/100) - which seems very noisy to me.  With option B, I get about 94 components - all of which have a very smooth and slow fluctuation pattern on the normalized response curve with less than 4.44 Hz/100 on the Powerspectrum timecourse - more like that I would expect of RSNs.
 
Advise?
 
Gratefully,
Varina
 
 
 
 
From: FSL - FMRIB's Software Library [[log in to unmask]] On Behalf Of Stephen Smith [[log in to unmask]]
Sent: Sunday, October 16, 2011 2:03 AM
To: [log in to unmask]
Subject: Re: [FSL] FEAT: High Pass Filtering

The longer you make the highpass setting the *less* it will remove - so if you want to get more aggressive, you need to *reduce* the setting.

On 15 Oct 2011, at 19:00, Varina Wolf wrote:

Hello FSL Experts,

I am using FEAT to pre-process (with ICA denoising) prior to MELODIC on a resting state individual analysis prior to group analysis..

I have one patient who had particularly noisy data, and in an effort to filter out more of the noise, I did several runs with incrementally higher values of the High Pass Filter (s) - 100, 250, 500, 750, 1000, 3000, 6750, 9000.

The outcome showed improving noise removal with fewer and less noisy components after ICA in FEAT.  But no difference past 6750 s.  

- I note an output document stating the cap on the filter is 1000, yet the results improved past this value.  Why?
- How does the  Temporal Filtering High Pass effect this?  
- After FEAT, do I need to put the same time cut off for Melodic?
- I suppose I need to use the same High Pass Filter for all of my data to have uniform group analysis later?

Best regards,
Varina



---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------