inter trial coherence non time resolved

💬 Support & Discussions
1

Dear all,

I’m a novice in mne. I need to compute inter-trial phrase coherence across epochs not resolved on time. I may have missed something but I did not found a function for this simple task. I wrote my own function (see below) however it is not clear to me which object to use to store the data in order to be correctly dealt by mne. I used mne.time_frequency.SpectrumArray but probably I need to change the info field in order to set the measure as coh (between 0 and 1).

Thanks in advance,

Francesco

  • MNE version: e.g. 1.9.0
  • operating system: Ubuntu 18.04
def itpc_spectrum(epo, fmin=None, fmax=None):
    """Compute INTER TRIAL PHASE COHERENCE from epochs.

    Parameters
    ----------
    epo : Epochs
    fmin : minimum frequency to return [Hz]
    fmax : maximum frequency to return [Hz]
    
    Returns
    -------
    spectrum :  a spectrum mne object with inter trial phase coherence,  it will be plotted with wrong labels and scales.
    """
    data = epo.get_data()
    N = data.shape[2]
    if np.mod(N,2):
        N = N-1
        data = data[:,:,range(0,N)]

    Nf = np.divmod(N,2)[0]

    yf = fft(data)
    xf = fftfreq(N, 1 / epo.info['sfreq'])[0:Nf]
    ind_min = 0
    ind_max = xf.shape[0] 
    if (fmin!=None):
        ind_min= np.where(xf>fmin)[0][0]

    if (fmax!=None):
        ind_max= np.where(xf<fmax)[0][-1]
    xf=xf[ind_min:ind_max]
    yf=yf[:,:,ind_min:ind_max]
        
    cf = np.sum(np.divide(np.real(yf),np.abs(yf)), axis=0)
    sf = np.sum(np.divide(np.imag(yf),np.abs(yf)), axis=0)

    k = np.power(yf.shape[0],2)
    itpc = np.multiply(cf,cf)/k+np.multiply(sf,sf)/k

    spectrum = mne.time_frequency.SpectrumArray(
        data=itpc,
        freqs=xf,
        info=epo.info,
    )
    return spectrum
2

Hi,

Doesn’t spectral_connectivity_epochs do what you want?

Cheers,

3

Thanks,
I’m looking into this but it is probably something more complex than what I need.
Francesco