- MNE version: 1.7
- operating system: Windows 10
Hello,
I am trying to do source analysis identifying the source location for significant frequency bands shown up in time frequency analysis. I want to try LORETA. Based on this article, we can compute the spectral power at source location by
\operatorname{diag}\left[\frac{1}{N_\tau N_{\varepsilon}} \sum_{\forall i, t}\left(J_{\mathrm{i}, t}^{\Omega}\right)\left(J_{\mathrm{i}, \mathrm{t}}^{\Omega}\right)^{\mathrm{T}}\right]=\operatorname{diag}\left\{\mathrm{T} \mathscr{S}_{\Phi}^{\Omega} \mathrm{T}^{\mathrm{T}}\right\}
where
\mathscr{S}_{\Phi}^{\Omega}=\left[\frac{1}{\mathrm{~N}_{\varepsilon}} \sum_{\forall \mathrm{i}, \omega \in \Omega}\left(\Phi_{\mathrm{i}, \omega}^{\Omega}\right)\left(\Phi_{\mathrm{i}, \omega}^{\Omega}\right)^*\right]
I think the cross spectral matrix can be easily computed via mne.time_frequency.csd_fourier. But applying the CSD matrix doesn’t seem to be implemented in mne.minimum_norm.apply_inverse. Is there any straight-forward way to do this in MNE? Thank you.
Kind regards,
Shuangyi