2
$\begingroup$

We are plotting the temporal generalization map (binary decoding) using EEG signal, but we obtained a strange result and don’t know how to interpret it. First, after bandpass filter (8-12 Hz), we could obtain a normal result as below. enter image description here As shown in time-course, the classifying values (thick black lines) between red line (first condition) and blue line (second condition) are variable along the time, so that the values in the early time do not easily apply to the late time leading to the normal map of the temporal generalization.

However, using higher frequency band (12-30 Hz) we encountered a strange result enter image description here In this case, the classifying values between red line (first condition) and blue line (second condition) are somewhat constant along the time, so that the values in the early time easily apply to the late time leading to the high decoding accuracy all over the map. The point is that two parallel curves are easily divided by a simple mean (black thick line) between two curves as the classifier. If two curves are not parallel but mixed with each other, the decoding accuracies over the map of the temporal generalization should be low, and we can conclude that the high frequency do not contain decodable information.

Questions, 1. Have you ever seen this type of time-courses of the high frequencies in our result and this strange map the temporal generalization and? 2. We use the matlab function ‘filtfilt.m’ for bandpass filter, could you recommend other function for bandpass filter?

$\endgroup$
1
$\begingroup$

I am unsure to what your paradigm exactly does, but as to your frequency bands a layman's interpretation would be as shown in Table 1. Hence, the frequency bands you investigate could simply be different, because the states associated with them are different.

As to your band filter - the best thing to do is find the rock stars in your field and copy their methods. The filtfilt function in matlab is excellent, as it is a zero-phase digital filter. It will depend on the filter order, however, if the outcome will make any sense. Too low of an order may generate overlap between bands, too high orders may introduce artifacts. But by the looks of it, none is happening here, but I can't be sure and you need to verify yourself.

EEG_bands
Table 1. Simplified overview of the EEG bands and their characteristics. source: New Dawn Magazine

$\endgroup$
  • 1
    $\begingroup$ Oof. I want to burninate the simplification of that table with fire... $\endgroup$ – Bryan Krause Jul 25 at 20:32
  • 1
    $\begingroup$ @BryanKrause haha, it's just to illustrate that different bands are associated with different states of mind. I adapted the legend a bit to accommodate your critique :) $\endgroup$ – AliceD Jul 26 at 7:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.