Version 0.4.0

Enhancements

• Add comprehensive MEG data support for Magnetometers, Gradiometers, and Optically Pumped Magnetometers (OPM) (#150 by Tomas Ros and Victor Férat).

• Implement MEG-aware SENSAI scoring using singular value projection and adaptive artifact spectrum prescanning (#150 by Tomas Ros and Victor Férat).

• Add support for passing MNE Forward models directly as reference_cov in Gedai, MultibandGedai, and AdaptiveMultibandGedai (#150 by Tomas Ros and Victor Férat).

• Add compute_covariance_from_forward() to compute cortical reference covariance matrices from forward solutions (#150 by Tomas Ros and Victor Férat).

• Add n_pc parameter to MultibandGedai and AdaptiveMultibandGedai (#150 by Tomas Ros and Victor Férat).

• Increase SENSAI epoch cap to 500 epochs for improved optimization convergence (#150 by Tomas Ros).

• Set default verbosity level to INFO to provide informative progress logging out of the box (#150 by Tomas Ros).

• Tighten wavelet band optimization bounds to guard against overcleaning (#150 by Tomas Ros).

Bugs

• Fix reference covariance channel picking and case-insensitive matching in _pick_cov (#150 by Tomas Ros).

• Ensure average reference is never applied to MEG sensor arrays (#150 by Tomas Ros and Victor Férat).

• Preserve raw metadata and sensor coordinates across transformation (#150 by Tomas Ros and Victor Férat).

API and behavior changes

• Automatic signal type detection (_detect_signal_type) from channel types.

• Default verbose=None now defaults to logging.INFO level.

Authors

• Tomas Ros

• Victor Férat