Time frequency analysis

Discussions
1

Dear Brainstorm users,

I am running time-frequency analysis (Morlet) on my dataset with these parameters:

frequency definition: 5:20:31 (Log), Central frequency = 1Hz, Time resolution = 3s

I have some questions about the parameters and the analysis steps:

1- To avoid edge artifacts is it alright to extract the epochs between -300 ms and +800 ms (The original time period is -100 to 600)?

2 - To reflect relative changes with respect to the pre-stimulus baseline, is it correct to run standardize > spectrum normalization > Z-score transformation on the TF matrices?

3- Is there any opportunity to define the length of the wavelets? If not what is the default value that brainstorm uses?

Thank you so much in advance.

Regards
Dandil

1 Like
2

Dandil, I'm following this post as I had a similar question. I'm curious about your preferred process for standardizing. Do you standardize the TF transforms for each trial, then average.....or average the unstandardized TF transforms, then standardize the average? I've read justifications for both and can't quite figure out what EEGlab does for ERSP calculations.
Paul

3

1- To avoid edge artifacts is it alright to extract the epochs between -300 ms and +800 ms (The original time period is -100 to 600)?
2 - To reflect relative changes with respect to the pre-stimulus baseline, is it correct to run standardize > spectrum normalization > Z-score transformation on the TF matrices?

This is something you should observe and conclude yourself based on your data, hypotheses, and frequencies of interest. The tutorial teaches you how to do it:
https://neuroimage.usc.edu/brainstorm/Tutorials/TimeFrequency#Normalized_time-frequency_maps

Other examples:
https://neuroimage.usc.edu/brainstorm/Tutorials/VisualSingle#Time-frequency_analysis
https://neuroimage.usc.edu/brainstorm/Tutorials/Epilepsy#Spike_average

3- Is there any opportunity to define the length of the wavelets? If not what is the default value that brainstorm uses?

The way the wavelets are defined is explained in details in the tutorial, and you can explore visually the wavelets that are used:
https://neuroimage.usc.edu/brainstorm/Tutorials/TimeFrequency
Please get back to us if there are pieces of information that you can't understand in there.

Do you standardize the TF transforms for each trial, then average.....or average the unstandardized TF transforms, then standardize the average?

Our recommended pipeline for time-frequency analysis is documented here:
https://neuroimage.usc.edu/brainstorm/Tutorials/Workflows#Time-frequency_maps