Can Pre-Processing EEG Signals Affect Correlation Coefficients?

  • Thread starter Thread starter kaushik200
  • Start date Start date
  • Tags Tags
    Signals
Join the discussion
Ask a follow-up here, or get your own question answered by working scientists, mathematicians and engineers — people, not an autocomplete.
Real named experts · corrections over time · the nuance an AI answer skips
Sign up Log in
4 replies · 5K views
kaushik200
Messages
5
Reaction score
0
Hello,
I need to calculate the correlation betweeen a pair of eeg signals in various bands (delta, theta, alpha...). Before I do so, I am filtering out frequencies below 0.5 Hz. (and also notching out 50 Hz.) I wonder whether the phase changes introduced by these filtering operations would affect my correlation coefficient. If yes, what kind of filter should I use? My prime objective is to study the correlation between different electrode pairs and I am concerned about these values getting affected by pre-processing.

Kaushik
 
Engineering news on Phys.org
kaushik200 said:
Hello,
I need to calculate the correlation betweeen a pair of eeg signals in various bands (delta, theta, alpha...). Before I do so, I am filtering out frequencies below 0.5 Hz. (and also notching out 50 Hz.) I wonder whether the phase changes introduced by these filtering operations would affect my correlation coefficient. If yes, what kind of filter should I use? My prime objective is to study the correlation between different electrode pairs and I am concerned about these values getting affected by pre-processing.

Kaushik

At first it sounds like you want to correlate signals of different frequencies, but that doesn't sound right. But then it sounds like you want to correlate signals in the same band, but from spatially separate electrodes (which sounds better). If the two signals are in the same band with the same processing, I think the processing should not have a big effect on the correlation.

Could you please clarify what you will be correlating?
 
berkeman said:
At first it sounds like you want to correlate signals of different frequencies, but that doesn't sound right. But then it sounds like you want to correlate signals in the same band, but from spatially separate electrodes (which sounds better). If the two signals are in the same band with the same processing, I think the processing should not have a big effect on the correlation.

Could you please clarify what you will be correlating?

Hi, sorry for not being clear enough. eeg readings have been taken from 45 different electrode positions on the scalp @ 256 samples/sec. These signals usually do not have components greater than 128 Hz. I intend to calculate time correlation coefficient between each pair of signal (there are 45c2 pairs) in 4 different bands of interest (0.5-4,4-8,8-12,12-16). Before I do so, all signals are bandlimited to within 0.5-64 Hz. (as a pre-processing formality) to remove very low and high freq. I wonder if the phase delay introduced by these digital filters (while pre-processing as well as while choosing the band of interest) would have a strong impact on the calculated correlation value. If so, I am looking for suggestions as to how to combat this.
 
kaushik200 said:
Hi, sorry for not being clear enough. eeg readings have been taken from 45 different electrode positions on the scalp @ 256 samples/sec. These signals usually do not have components greater than 128 Hz. I intend to calculate time correlation coefficient between each pair of signal (there are 45c2 pairs) in 4 different bands of interest (0.5-4,4-8,8-12,12-16). Before I do so, all signals are bandlimited to within 0.5-64 Hz. (as a pre-processing formality) to remove very low and high freq. I wonder if the phase delay introduced by these digital filters (while pre-processing as well as while choosing the band of interest) would have a strong impact on the calculated correlation value. If so, I am looking for suggestions as to how to combat this.

So you are only correlating signals within the same frequency band, correct?
 
berkeman said:
So you are only correlating signals within the same frequency band, correct?

Yes, of course :)