Source analysis for certain cluster in the imported Time-frequency data

Dear Expert,

I am writing to seek your advice on conducting a source analysis for time-frequency maps.

First, I implemented time-frequency analysis of EEG data using scripts in Matlab. And then I performed SVM analysis to establish a relationship between the time-frequency data and behavior, discovering that certain cluster in the time-frequency map of a specific channel is correlated with behavior. I am interested in conducting a source analysis for this cluster across all electrodes. However, it seems challenging to achieve in brainstorm, as the tutorial mentions that source analysis can be performed on EEG time-domain data but not on certain cluster of time-frequency maps.

The steps I have followed for analysis are as follows:

  1. Extracting the average values (such as 510^(-6), 4.510^(-6), etc) of the cluster for all electrodes.
  2. Converting these values into time-domain signals (In each of the 59 channels, there are 1,500 identical values, e.g., 5*10^(-6)). ps. time points=1500
  3. Conducting source analysis on these reconstructed signals.

I am unsure if extracting fixed values to construct an EEG signal for source analysis is a viable approach or if Brainstorm offers alternative methods for source analysis of certain cluster in the time-frequency data which is imported from the matlab.

I would greatly appreciate any guidance or suggestions provided on this matter.



Indeed source mapping is valid only from actual biophysical time series (EEG, MEG, intracranial recordings), not from analytical derivatives as in the present case. The reason is that the model that maps brain sources from recordings is based on the biophysics of electrophysiology. This model is not valid anymore when the sensor data is transformed. I encourage you to conduct the same time-frequency analysis but with source time series instead.

Hi, expert

Thank you for your prompt response. I appreciate your clarification and your suggestion to utilize source time series instead.

While it is relatively easy to plot the topographical maps of the transformed data, such as 510^(-6), 4.510^(-6), etc., in Matlab, I understand that achieving source mapping results, as per your previous response, may not be feasible in the Brainstorm.

If I were to perform TF analysis using source time series in Brainstorm, that seems viable. However, my intention is to conduct SVM analysis on the TF maps and subsequently perform source mapping on the identified clusters within the maps. It appears that this specific analysis cannot be accomplished using Brainstorm. Therefore, I would like to inquire if there are any alternative methods or if this analysis is simply not possible.

Thank you for your time and assistance.