MEG TFR source localization question -- should i use only planars or both grad + mag

Dear colleges!

I have Electa Neuromag dataset that includes both planar gradiometers and magnetometers. And since i have both i wonder what is going to work better for source localization, using mne.minimum_norm.apply_inverse_tfr_epochs()?

Does function counts for different sensor types? Or it is better to use just planars or planar1+planar2 values?

Or covariance matrix should be properly normalized and this is the only thing that really matters?

Thank you!

If both sensor types are producing clean signals (inspect the raw data and see how the signal quality is) you can use both. MNE-Python takes care of the normalization using the noise covariance matrix. If you are studying oscillatory activity, you probably mostly want the magnetometers as they are more sensitive to deeper sources, but the gradiometers are certainly valuable and worth using.

One practical point: apply_inverse_tfr_epochs does not adapt to a sensor subset by itself—your TFR data, forward solution, noise covariance and inverse operator must all use the same channel set. So for planar-only, recompute or subset all of them consistently, and do not average planar1+planar2 before source localization. If you use both, exclude bad channels and compute the noise covariance on that same grad+mag set; then covariance-based whitening handles normalization. Using both is usually fine when both are clean.