DICS source power

✉️ Mailing List Archive (read-only)
1

Hello !

I?m trying to modify the dics_source_power function
(mne/beamformer/_dics.py) and the compute_epochs_csd function
(mne/time_frequency/csd.py) to apply DICS on each source and each epoch
individually. I?m trying to understand how these functions work but I have
a problem with DICS.

Could you please help me to understand why the matrix filter (?Wk?) is
divided by ?Ck? in the case of fixed source orientation (_dics.py , line
398).

Ck = Gk.T.dot(Cm_inv).dot(Gk)

where G is the forward solution and Cm_inv is the pseudoinverse of the CSD.

Any help or suggestion would be appreciated!

Many thanks.

Alexandre F
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20160624/af6a6f67/attachment.html

2

hi,

I?m trying to modify the dics_source_power function
(mne/beamformer/_dics.py) and the compute_epochs_csd function
(mne/time_frequency/csd.py) to apply DICS on each source and each epoch
individually. I?m trying to understand how these functions work but I have a
problem with DICS.

Could you please help me to understand why the matrix filter (?Wk?) is
divided by ?Ck? in the case of fixed source orientation (_dics.py , line
398).

with fixed ori Ck is a float so

Wk[:] = np.dot(linalg.pinv(Ck, 0.1), Wk)

is replaced by

Wk /= Ck

ok?

if you think there is a but please with us on github a gist so we can
replicate the pb.

Alex