Dipole fitting process question

Discussions
1

Dear experts;
Hi;
I have considered two dipoles within the volume of the head and assigned a signal with a wide spectrum of frequencies from 0.5 to 98 Hz to each. Then, I simulated the EEG signal recorded by a 65-channel electrode array and added noise. Using the dipole modeling command, I localized the bipolar sources, and with the Dipole Scanning command, I determined the estimated locations of the dipoles. However, I have encountered several questions for which I would appreciate your response:

  1. What is the relationship between SNR=10logps/pn, where ps is the signal power and pn is the noise power, with the specified SNR in the simulation stages?
    image

  2. Changing the SNR (increasing or decreasing) does not significantly affect the estimated locations of the dipoles.

  3. The dipoles displayed after dipole scanning all have different centers and directions, indicating that all estimates have identified a single location for the dipoles. Is this correct?

    image
    image504×506 46.6 KB

    Did I make a mistake at some stage? if it is needed I will send you the process details.

best regards
Morteza

2
  • The RSN is the inverse of SNR: it indicates a relative amplitude of added noise to the source time series.
  • I recommend you visualize the corresponding sensor time series after you add different levels of noise so that you can better appreciate the actual effect on source localization. Noise added across all sensors will have a greater effect on accuracy than noise added to the respective source time series simulated.
  • I am not sure I understand the image you provided: it looks like the dipoles are all located at two spots, with only their orientations rotating over time I suppose. Again, if you add noise to the simulated EEG data (across all channels) you'll find a greater effect on the consistency of source locations.