+ Post New Thread
Results 1 to 5 of 5
  1. #1
    Newbie level 2
    Points: 20, Level: 1

    Join Date
    Sep 2015
    Posts
    2
    Helped
    0 / 0
    Points
    20
    Level
    1

    denoising question, for EEG-FMRI application

    hello, i'm a researcher doing simultaneous EEG-FMRI recordings. i'm trying to denoise the EEG signal acquired inside the scanner...my question is this: if i have access to a clean signal (EEG data acquired outside the scanner) and a noisy signal (EEG data acquired inside the scanner) is there a way i can use the information from the clean signal to remove the scanner-related noise from the noisy signal? the noise is roughly on the same order of magnitude as the signal.

    the standard approach is to use average artifact subtraction, followed by an ICA decomposition, to first subtract out the majority of the noise, and then remove the rest through subtracting the noisy components from the data using ICA. this doesn't really work so well due to 1) the physiological nature of the noise, average artifact subtraction cannot account for this variability and 2) the artifact is non-stationary so ICA can't handle it cleanly either.

    basically i'm wondering if you have two signals, signal a and signal b, where a was recorded in a noise-free environment, and b was recorded in a noisy environment, is it possible to use the information from signal a to remove the noise from signal b?
    thanks,
    Russell

    •   Alt24th September 2015, 02:26

      advertising

        
       

  2. #2
    Advanced Member level 3
    Points: 5,298, Level: 17

    Join Date
    Jan 2015
    Posts
    767
    Helped
    248 / 248
    Points
    5,298
    Level
    17

    Re: denoising question, for EEG-FMRI application

    Wild guess - can you look at the signal in the frequency domain and tailor a filter (based on the clean signal) to reduce the noise at 'unnecessary' frequencies?
    Susan



    •   Alt24th September 2015, 03:40

      advertising

        
       

  3. #3
    Newbie level 2
    Points: 20, Level: 1

    Join Date
    Sep 2015
    Posts
    2
    Helped
    0 / 0
    Points
    20
    Level
    1

    Re: denoising question, for EEG-FMRI application

    Quote Originally Posted by Aussie Susan View Post
    Wild guess - can you look at the signal in the frequency domain and tailor a filter (based on the clean signal) to reduce the noise at 'unnecessary' frequencies?
    Susan
    yeah i've tried this, doesn't really work because there's too much overlap, ie the noise is strongest in the same frequencies i'm interested in in the clean signal.



    •   Alt24th September 2015, 04:07

      advertising

        
       

  4. #4
    Super Moderator
    Points: 45,413, Level: 52

    Join Date
    Apr 2011
    Location
    Minneapolis, Minnesota, USA
    Posts
    11,305
    Helped
    2233 / 2233
    Points
    45,413
    Level
    52

    Re: denoising question, for EEG-FMRI application

    I don't suppose this is a case where you want only the signal common to both sources? Occasionally a similar question comes up about stereo audio: Is there any way to eliminate instruments which are playing only in the left channel, and only in the right channel? This would let us hear only the instruments (or vocals) playing in the center? I have not seen any truly effective method to do this.



  5. #5
    Super Moderator
    Points: 22,948, Level: 36
    andre_teprom's Avatar
    Join Date
    Nov 2006
    Location
    Portugal
    Posts
    6,975
    Helped
    854 / 854
    Points
    22,948
    Level
    36
    Blog Entries
    5

    Re: denoising question, for EEG-FMRI application

    Quote Originally Posted by russell88 View Post
    the noise is strongest in the same frequencies i'm interested in in the clean signal.
    Seems like matching the Susan's advise.
    If the noise bandwidth is known, you can try performing a notch filtering.
    --------------------------------------------------------------------------------------------------
    Part of the world that you live in, You are the part that you're giving ( Renaissance )



--[[ ]]--