Extracting epochs from an edf file to save as .m file?

Hey, newbie to coding and all of that. I'm trying to extract 10 random epochs that are 20 seconds each from my EEG recording (.edf file). I want to be able to save these epochs as a .m file. I'm trying to do this without having to use the EEGLAB GUI. I used edfread and it seems that the data within the edf file is a timetable with a size of 5282 x 1.

Any help on how to do this would be greatly appreciated.

[Mathieu NOE]

hello

so what code have you produced so far ? can you share it (and the data )

[pickyourbrain]

This is a screenshot of what I have so far. Thank you!

[Mathieu NOE]

hello again

pictures are nice but not really helpfull

code and data are needed if you want someone to help you

[Walter Roberson]

You are using

[hdr, record] = edfread('C57.edf');

This will return two timetables, with the first one being the primary data and the second one being for annotations.

sampling_rate = hdr.samples(1);

The timetable returned in your variable hdr is quite unlikely to happen to have a variable named samples. The variable names present are "Record Time" together with one variable per signal. You are unlikely to happen to have a signal named samples and if you did, the first value inside it would be unlikely to be the sampling rate.

It looks to me as if you are expecting hdr to be what you would expect by calling edfinfo

[pickyourbrain]

This is the code that I have in the earlier screenshot. The C57 variable has a size of 7743x3327x3x2 if that is useful at all.

% Read the EDF data
[hdr, record] = edfread('C57.edf');

% Parameters
epoch_duration = 20;  % Duration of each epoch in seconds
num_epochs = 10;      % Number of epochs to extract
sampling_rate = hdr.samples(1); 
epoch_samples = epoch_duration * sampling_rate;  % Number of samples per epoch

% Pre-allocate a cell array to store the epochs
epochs = cell(num_epochs, 1);

% Extract 10 epochs 
for epoch_idx = 1:num_epochs
    start_sample = (epoch_idx - 1) * epoch_samples + 1;  
    end_sample = epoch_idx * epoch_samples;  
    
    if end_sample <= size(record, 1)
        % Extract the data for all channels for this epoch
        epochs{epoch_idx} = record(start_sample:end_sample, :); 
    else
        warning('Not enough data for epoch %d. Skipping it.', epoch_idx);
        epochs{epoch_idx} = [];  % If there is not enough data for this epoch
    end
end

% Save the extracted epochs to a .mat file
save('C57_epoch.mat', 'epochs', 'hdr');

Okay so currently hdr is a 6282 x 1 timetable, and record is a 3x2 timetable. I see that the variables should be switched.

So currently for hdr, the first column says "Record Time" and the second column says "EEGEEG1A_B" with each row holding 10000 x 1 double.

Do I change the code to:

sampling_rate = hdr.EEGEEG1A_B(1);

Unsure if this is the right direction and if there is anything else I need to fix. Thank you for all the help so far!

[Walter Roberson]

sampling_rate = hdr.EEGEEG1A_B(1);

No.

https://www.mathworks.com/help/signal/ref/edfread.html#mw_a83171d9-fafc-4332-ad36-8243ddcc75c6 shows an example of using edfinfo() and edfread() to compute the vector of sampling frequencies, fs . sampling_rate is 1 ./ fs

Each different signal could, in theory, have a different sampling rate.

epoch_samples = epoch_duration * sampling_rate;

That might not be an integer.

What you should be doing instead, is repeatedly using timerange to extract epochs, since there could be a different number of samples in each epoch (because the epoch_samples is not certain to be an integer.)

[pickyourbrain]

Sorry, this is still very abstract for me. Could you explain more about how timerange can be implemented? Do I use timerange in a for loop? What is the structure contained in each epoch?