Comodulogram not showing

[sjg2203]

While computing comodulogram, I can only get it to work for the mean vector length(MVL) but not for the modulation index (MI), which is the one that I need.
Would anyone have any idea why this isn't working?

I included my code and example of the comodulogram that doesn't work and the one that works.

input_fname=easygui.fileopenbox(msg="EDF file")
raw=mne.io.read_raw_edf(input_fname,preload=True,verbose=False)
raw.resample(100)
sf=raw.info
raw.filter(0.1,45)
data=raw.get_data(units="uV")
sls=yasa.SleepStaging(raw,eeg_name='C3',eog_name='EOG R',emg_name='CHIN1')
hypno_pred=sls.predict()
hypno_pred=yasa.hypno_str_to_int(hypno_pred)
hypno_up=yasa.hypno_upsample_to_data(hypno_pred,sf_hypno=1/30,data=raw)
_,data_N2=yasa.sliding_window(hypno_up==2,sf,window=15)
data_N2.shape
f_pha=np.arange(0.375,4.25,0.25)
f_amp=np.arange(7.25,25.5,0.5)
from tensorpac import Pac
p=Pac(idpac=(2,0,0),f_pha=f_pha,f_amp=f_amp,verbose='WARNING')
xpac=p.filterfit(sf,data_N2)
p.comodulogram(xpac.mean(-1),title=str(p),vmin=0,plotas='contour',ncontours=25)
p.show()

p=Pac(idpac=(2,0,0),f_pha=f_pha,f_amp=f_amp,verbose='WARNING')

No comodulogram showing for MI

but suddenly works when asking MVL

p=Pac(idpac=(1,0,0),f_pha=f_pha,f_amp=f_amp,verbose='WARNING')

Thanks

[raphaelvallat]

Hi,

It's hard to say without the original EDF data. Are you able to run this notebook in the notebooks/ folder of YASA?

The MVL output looks kind of weird to m. I think there might be an issue in how you select the data. What does data_N2 look like? Seems like it is currently a boolean?

[sjg2203]

I amended data_N2 so that it uses the raw data instead, and also cut the 100 first epochs.

_,data_N2=yasa.sliding_window(data,sf,window=15)
data_N2=data_N2[:100,:]
data_N2.shape
Output: (100,26,1500)

However, it returned this error:

>>xpac=p.filterfit(sf,data_N2)

>>Traceback (most recent call last):
  File "C:\Users\serv_H\AppData\Local\Programs\Python\Python310\lib\site-packages\IPython\core\interactiveshell.py", line 3433, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-83-fc1fe6e5d6a5>", line 1, in <module>
    xpac=p.filterfit(sf,data_N2)
  File "C:\Users\serv_H\AppData\Local\Programs\Python\Python310\lib\site-packages\tensorpac\pac.py", line 478, in filterfit
    pha = self.filter(sf, x_pha, 'phase', **kw)
  File "C:\Users\serv_H\AppData\Local\Programs\Python\Python310\lib\site-packages\tensorpac\pac.py", line 78, in filter
    assert x.ndim == 2, ("x should be a 2d array like (n_epochs, n_times)")
>>AssertionError: x should be a 2d array like (n_epochs, n_times)

[raphaelvallat]

In your latest example data_N2 is a 3D array of shape (100, 26, 1500). Where does the 26 comes from? Is that the number of channels? You can only calculate the Pac on a single channel at a time. Please use something like np.squeeze(data_N2[:, 0, :])

[sjg2203]

26 is indeed the number of channels in my edf file.
I used what you suggested and it now generates a much better plot, though I am not entirely sure why there is some merging happening here.

>>data_N2=np.squeeze(data_N2[:,0,:])
>>data_N2.shape
Output: (2030, 1500)

[raphaelvallat]

What do you mean by merging?

[sjg2203]

The three main blobs are somehow meeting, I mean, normally we do only have two of those which are quite distinct but here, we can see some gradient going from one another