IRASA Python version?

Mailing List Archive (read-only)
1

External Email - Use Caution

Hi all,
Does anybody implemented the IRASA (Irregular Resampling Auto-Spectral
Analysis) method or something similar in Python? I'm interested in finding
the neural oscillations at lower frequency bands, however, due to the 1/f
problem it is not possible to locate the peaks at low frequency bands
especially at the delta ranges by using the welch or multitaper PSD
estimation methods. Does MNE Python has any advanced methods to overcome
this problem?

Thanks in advance,
Harish

2

I know of no such implementation, but I've opened an issue on Github.

A simpler procedure for PSD that is often effective for removing 1/f is:

1. log transform psd and f. Then psd ~ 1 / f implies log psd ~ -log(f)
2. Fit a regression line and subtract out the fitted values / take the
residuals.
3. Back transform the residuals.

So something like (may need tweaking depending on whether data is
Epochs/Evoked/etc.):

from mne.time_frequency import psd_welch
from sklearn.linear_model import LinearRegression

psds, freqs = psd_welch(data)
line = LinearRegression()
line.fit(np.log(freqs).reshape(-1,1), psd.reshape(-1,1))
pink_noise = line.predict(np.log(freqs).reshape(-1,1))
de_pinked_log = np.log(psd).reshape(-1,1) - pink_noise
psd_depinked = np.exp( de_pinked_log.reshape(-1))

???External Email - Use Caution???

Hi all,
Does anybody implemented the IRASA (Irregular Resampling Auto-Spectral
Analysis) method or something similar in Python? I'm interested in
finding the neural oscillations at lower frequency bands, however, due
to the 1/f problem it is not possible to locate the peaks at low
frequency bands especially at the delta ranges by using the welch or
multitaper PSD estimation methods. Does MNE Python has any advanced
methods to overcome this problem? ?

Thanks in advance,
Harish

*/------------------------------------------------------------------------------------------/**/
/*
*/Harish Gunasekaran,/*
NUMERICS
<http://numerics.cea.fr/Pages/Numerics/Numerics%20program/Numerics-PhD.aspx&gt;
Ph.D.?fellow under Marie Sk?odowska-Curie actions?(2019-22),
Cognition & Brain Dynamics team <https://brainthemind.com/team2/&gt;,
Cognitive Neuroimaging Unit (UNICOG <http://www.unicog.org/site_2016/&gt;\),?
NeuroSpin, CEA/DRF/Joliot, CEA Paris Saclay,?
B?t 145 PC 156 , Gif s/ Yvette F-91190 FRANCE.
*/------------------------------------------------------------------------------------------/*
*/
/*

_______________________________________________
Mne_analysis mailing list
Mne_analysis at nmr.mgh.harvard.edu
Mne_analysis Info Page

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
Url : http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20200113/f39889ab/attachment.bin
-------------- next part --------------
        External Email - Use Caution
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20200113/f39889ab/attachment.html