Version 1.3.0 (2022-12-21)

Enhancements

• Improve default line width used in mne.viz.plot_filter() (#11333 by new contributor Toomas Erik Anijärv)

• Add the argument max_iter to psd_array_multitaper() and to csd_array_multitaper() to increase the maximum number of iterations to reach convergence when using adaptive=True (#11237 by Mathieu Scheltienne)

• Add a warning to the docstring of mne.channels.find_ch_adjacency() to encourage users to validate their outputs (#11236 by Felix Klotzsche and Eric Larson)

• Mixed, cortical + discrete source spaces with fixed orientations are now allowed. (#11241 by Jevri Hanna)

• Add size information to the repr of mne.Report (#11357 by Eric Larson)

• Add support for image_format='webp' to mne.Report when using Matplotlib 3.6+, which can reduce file sizes by up to 50% compared to 'png'. The new default image_format='auto' will automatically use this format if it’s available on the system (#11359 by Eric Larson)

• Add mne.beamformer.apply_dics_tfr_epochs() to apply a DICS beamformer to time-frequency resolved epochs (#11096 by Alex Rockhill)

• Add mne.time_frequency.fwhm() to determine the full-width half maximum for mne.time_frequency.morlet() (#11353 by Britta Westner, Daniel McCloy, and Eric Larson)

• Check whether head radius (estimated from channel positions) is correct when reading EEGLAB data with read_raw_eeglab() and read_epochs_eeglab(). If head radius is not within likely values, warn informing about possible units mismatch and the new montage_units argument (#11283 by Mikołaj Magnuski).

• Add support for a callable passed in combine for mne.time_frequency.AverageTFR.plot and mne.time_frequency.AverageTFR.plot_joint (#11329 by Mathieu Scheltienne)

Bugs

• Fix bug in mne.export.export_raw() when exporting raw to EDF with EDFLib when subject sex is missing (#11375 by new contributor Mark Alexander Henney)

• Fix bugs in documentation of mne.time_frequency.tfr_array_multitaper() and related functions/methods (#11310 by new contributor Santeri Ruuskanen)

• Fix bug to be able to customize legends when using mne.viz.Brain.add_volume_labels() (#11289 by new contributor Mauricio Cespedes Tenorio)

• Fix bug in documentation of mne.channels.make_dig_montage() (#11235 by new contributor Daniel Hasegan)

• Add support for bad channel handling in mne.set_bipolar_reference() (#11245 by Moritz Gerster, new contributor Dinara Issagaliyeva, new contributor Jennifer Behnke, new contributor Hakimeh Aslsardroud, and new contributor Pavel Navratil)

• Fix X coordinates of ear_left and ear_right in head outlines dict (#11255 by new contributor Tom Ma)

• Add type checking for tmin and tmax arguments for crop (#11263 by new contributor Carlos de la Torre-Ortiz).

• Fix bug where trying to setup projection without an EEG average reference projector would fail (#11351 by new contributor Enzo Altamiranda)

• Fix bug where ica.reject_ was not saved to disk, and the ica.reject_ property was not inherited from Epochs when doing ICA.fit(epochs) (#11244 by Eric Larson)

• Fix bug in automatic MESA detection for disabling advanced 3D options (#11271 by Eric Larson)

• Fix bug in mne.head_to_mri() when unscale=True (#11303 by Eric Larson)

• Fix bug in set_montage('mgh60') and related functions where the 'mgh60' sensor positions were incorrectly ordered (#11290 by Eric Larson)

• Fix bug in the .compute_psd() methods where the number of unaggregated Welch segments was wrongly computed for some inputs, leading to an assertion error when computing the PSD (#11248 by Daniel McCloy)

• Fix bug in the plot_evoked_topo() and plot_topo(), where legend colors where shown incorrectly on newer matplotlib versions (#11258 by Erkka Heinila)

• Fix bug where EEGLAB channel positions were read as meters, while they are commonly in millimeters, leading to head outlies of the size of one channel when plotting topomaps. Now montage_units argument has been added to read_raw_eeglab() and read_epochs_eeglab() to control in what units EEGLAB channel positions are read. The default is millimeters, 'mm' (#11283 by Mikołaj Magnuski)

• Fix bug where computing PSD with welch’s method with more jobs than channels would fail (#11298 by Mathieu Scheltienne)

• Fix bug where the default FFT length changed for spectrum plots (#11345 by Daniel McCloy)

• Fix bug with mne.decoding.cross_val_multiscore() where progress bars were not displayed correctly (#11311 by Eric Larson)

• Fix channel selection edge-cases in find_bads_muscle (#11300 by Mathieu Scheltienne)

• Fix bug with mne.io.read_raw_curry() where a dot in the parent folders prevented files from being read (#11340 by Eric Larson)

• Fix bug with mne.Report with replace=True where the wrong content was replaced and section was not respected (#11318, #11346 by Eric Larson)

• Fix bug with unit conversion when setting reference MEG as the channel type in mne.io.Raw.set_channel_types() and related methods (#11344 by Eric Larson)

• Fix bug where reference MEG channels could not be plotted using mne.viz.plot_epochs_image() (#11344 by Eric Larson)

• Fix bug where image_format='gif' was errantly documented as being supported by mne.Report, it is now only supported in mne.Report.add_image() (#11347 by Eric Larson)

• Multitaper spectral estimation now uses periodic (rather than symmetric) taper windows. This also necessitated changing the default max_iter of our cross-spectral density functions from 150 to 250. (#11293 by Daniel McCloy)

• Fix mne.Epochs.plot_image() and mne.viz.plot_epochs_image() when using EMG signals (#11322 by Alex Gramfort)

• Fix selection of ICA components in mne.viz.plot_ica_components() and mne.preprocessing.ICA.plot_components() (#11369 by Mathieu Scheltienne)

• Fix bug where a Freesurfer reconstruction run with both a T1 and a T2 caused mne.viz.Brain not to initialize pial surfaces properly (#11361 by Alex Rockhill)

API changes

• In mne.time_frequency.dpss_windows(), interpolating is deprecated (nowadays SciPy’s computations are fast enough for large N without interpolation). This affects parameters interp_from and interp_kind. Two new parameters of the underlying SciPy dpss() function are also exposed: sym (for choosing symmetric vs. periodic windows) and norm (window normalization method). (#11293 by Daniel McCloy)

• In mne.decoding.CSP.plot_patterns(), mne.decoding.CSP.plot_filters(), mne.preprocessing.ICA.plot_components(), and mne.viz.plot_ica_components(), the parameters vmin and vmax are deprecated in favor of vlim, for consistency with other topomap-plotting functions and methods (#11371 by Daniel McCloy)

• In mne.decoding.CSP.plot_patterns() and mne.decoding.CSP.plot_filters() the title parameter is deprecated and will be removed in version 1.4 (#11371 by Daniel McCloy)

• The APIs of mne.preprocessing.ICA.plot_components() and mne.viz.plot_ica_components() gained new parameters show_names, extrapolate, border, size, cnorm, cbar_fmt, axes, nrows, ncols, for consistency with other topomap-plotting functions and methods (#11371 by Daniel McCloy)

• The APIs of mne.decoding.CSP.plot_patterns() and mne.decoding.CSP.plot_filters() gained new parameters extrapolate, border, cnorm, axes, nrows, ncols, for consistency with other topomap-plotting functions and methods (#11371 by Daniel McCloy)

Authors

• Alex Rockhill

• Alexandre Gramfort

• Britta Westner

• Carlos de la Torre+

• Daniel Hasegan+

• Daniel McCloy

• Dinara Issagaliyeva+

• Enzo Altamiranda+

• Eric Brayet+

• Eric Larson

• Erkka Heinila

• Felix Klotzsche

• Hakimeh Aslsardroud+

• Jennifer Behnke+

• Jevri Hanna

• Lukas Hecker

• Mark Alexander Henney+

• Mathieu Scheltienne

• Mauricio Cespedes+

• Mikołaj Magnuski

• Moritz Gerster

• Omer Shubi+

• Pavel Navratil+

• Richard Höchenberger

• Santeri Ruuskanen+

• Stefan Appelhoff

• Timon Merk

• Tom Ma+

• Toomas Erik Anijärv+