What’s new#

Enhancements#

• Added new built-in montage R-Net 128 Ch (Brain Products) (#10974 by new contributor Ilias Machairas)

• DigMontage objects can now be checked for equality via == (#10942 by new contributor Tziona NessAiver)

• Add a note to clarify the inclusion of average referencing gain matrix during whitening (#10922 by new contributor Mingjian He)

• Add time axis scale bar for mne.Epochs.plot() using matplotlib backend (#10736 by new contributor Jordan Drew)

• Add mne.bem.distance_to_bem() to find depth of source positions (#10632 by new contributor Matt Courtemanche)

• Add support for ahdr files in mne.io.read_raw_brainvision() (#10515 by new contributor Alessandro Tonin)

• Add support for reading data from Gowerlabs devices to mne.io.read_raw_snirf() (#10555 by new contributor Samuel Powell and Robert Luke)

• Add built-in sensor layout files for Geodesic Head Web 130 and 280 devices (#10627 by Evan Hathaway)

• Add mne-icalabel to mne.sys_info() (#10615 by Adam Li)

• Add support for overview_mode in raw.plot() and related functions/methods (#10501 by Eric Larson)

• Add mne.io.Raw.crop_by_annotations() method to get chunks of Raw data based on mne.Annotations. (#10460 by Alex Gramfort)

• The pick_channels method gained a verbose parameter, allowing e.g. to suppress messages about removed projectors (#10544 by Richard Höchenberger)

• Add mne.viz.plot_projs_joint() for joint plotting of projectors and Evoked (#10720 by Eric Larson)

• The mne.make_forward_dipole() function can now take a list of dipoles to make a multi-dipole forward models (#10464 by Marijn van Vliet)

• Add mode='outlines' support to mne.Dipole.plot_locations() and mne.viz.plot_dipole_locations() (#10699 by Eric Larson)

• Add mne.preprocessing.ICA.find_bads_muscle() to find muscle-related ICA components with an example, Removing muscle ICA components (#10534 by Alex Rockhill)

• Add example of Xfit-style ECD modeling using multiple dipoles (#10464 by Marijn van Vliet)

• Add head_source argument to mne.make_field_map() to allow selecting which head source to use (#10568 by Eric Larson)

• Add support for n_jobs=None to support joblib.parallel_backend() for more precise control over parallelization (#10567 by Eric Larson)

• It is now possible to compute inverse solutions with restricted source orientations using discrete forward models (#10464 by Marijn van Vliet)

• The new function mne.preprocessing.maxwell_filter_prepare_emptyroom() simplifies the preconditioning of an empty-room recording for our Maxwell filtering operations (#10533 by Richard Höchenberger and Eric Larson)

• Add keyboard shortcuts to toggle mne.preprocessing.ICA.plot_properties() topomap channel types (‘t’) and power spectral density log-scale (‘l’) (#10557 by Alex Rockhill)

• Add --mri, and --threshold options to mne make_scalp_surfaces to improve head surface mesh extraction (#10591 by Eric Larson)

• Add mne.preprocessing.compute_bridged_electrodes() to detect EEG electrodes with shared spatial sources due to a conductive medium connecting two or more electrodes, add Identify EEG Electrodes Bridged by too much Gel for an example and mne.viz.plot_bridged_electrodes() to help visualize (#10571 by Alex Rockhill)

• Add 'nearest' as an option for the image_interp argument in mne.viz.plot_topomap() to plot a topomap without interpolation using a Voronoi parcelation (#10571 by Alex Rockhill)

• Add mne.preprocessing.interpolate_bridged_electrodes() to use the spatially smeared signal to get a better interpolation rather than dropping those channels (#10587 by Alex Rockhill)

• Add support for hemoglobin type fNIRS data to temporal derivative distribution repair (TDDR) mne.preprocessing.nirs.temporal_derivative_distribution_repair() (#10125 by Johann Benerradi)

• mne.viz.plot_evoked_topomap() and mne.Evoked.plot_topomap() now display the time range the map was averaged over if average was passed (#10606 by Richard Höchenberger)

• mne.viz.plot_evoked_topomap() and mne.Evoked.plot_topomap() can now average the topographic maps across different time periods for each time point. To do this, pass a list of periods via the average parameter (#10610 by Richard Höchenberger)

• mne.viz.plot_evoked() and mne.Evoked.plot() gained a new parameter, highlight, to visually highlight time periods of interest (#10614 by Richard Höchenberger)

• Added fNIRS support to mne.Info.get_montage() (#10611 by Robert Luke)

• Add mne.viz.plot_ch_adjacency() to visualize and manually edit the channel adjacency matrix (#10633 by Mikołaj Magnuski)

• Add mne.viz.Brain.get_view() to get the current camera parameters such that they can be passed to mne.viz.Brain.show_view() (#10661 by Alex Rockhill)

• Added support for Aurora version 2021.9.0.6 to mne.io.read_raw_nirx() (#10668 by Robert Luke)

• Added BrainVision format (.vhdr, .vmrk, .eeg) to mne.export.export_raw() (#10681 by Stefan Appelhoff)

• Added export method to mne.Evoked (#10681 by Stefan Appelhoff)

• Add timeformat YYYY/MM/DD to mne.io.ctf.info._convert_time (#10674 by Simon Kern)

• mne.Report.add_figure() and mne.Report.add_image() gained a new parameter, section, allowing to put multiple figures or images into the same “content block”. Previously, the only way to group figures and images was by means of a slider, which sometimes made content hard to discover. The slider remains available as an alternative (#10694 by Richard Höchenberger)

• All functions and methods that plot topographic maps for EEG (2D projections of the EEG sensor locations) now accept the parameter value sphere='eeglab' to lay out the sensors with respect to the head circle in a similar way to how EEGLAB does, i.e., T7/T8 and Fpz/Oz are placed directly on the circle for template montages (#10572 by Richard Höchenberger)

• mne.channels.get_builtin_montages() gained a new parameter, descriptions, which allows to retrieve the descriptions of the montages in addition to their names (#10373 by Richard Höchenberger)

• Add include option to mne.io.read_raw_edf(), mne.io.read_raw_bdf() and mne.io.read_raw_gdf(). (#10734 by Michiru Kaneda)

• mne.gui.coregistration() gained a new parameter, fullscreen, to start the GUI in fullscreen mode (#10284 by Richard Höchenberger)

• The built-in FieldTrip channel adjacency matrices, which can be read via read_ch_adjacency(), have been synchronized with FieldTrip again. This means that numerous new adjacencies have been added, most notably those for standard 10-05 and 10-20 montages (#10749 by Richard Höchenberger)

• mne.Evoked.plot_field() gained a new parameter, interaction, to control the rotation axes when interacting with the head (#10788 by Richard Höchenberger)

• Add mne.read_evoked_besa() for reading evokeds from BESA .avr and .mul files. (#10892 by Marijn van Vliet)

• Add mne.time_frequency.EpochsTFR.decimate() to reduce size of time-frequency epochs objects (#10940 by Alex Rockhill)

• Add an alias that maps EGI reference channel names (VREF or Vertex Reference) to the electrode name in the standard EGI montages (Cz). Now setting standard EGI montages with set_montage() with match_alias=True will avoid raising an error (#10823 by Scott Huberty)

• Add provenance in mne.time_frequency.EpochsTFR.apply_baseline() (#10979 by Alex Rockhill)

Bugs#

• Fix bug in mne.surface._project_onto_surface for the condition method != 'accurate' and return_nn (#10930 by Christian O’Reilly)

• Fix bug in mne.export.export_raw() to ignore None value in filenames attribute of mne.io.RawArray (#10927 by new contributor Reza Nasri)

• Fix bug in mne.io.read_raw_edf() to allow reading in all Hypnodyne ZMax EDFs to be read in without issues (#10754 by new contributor Frederik Weber)

• Fix bug in mne.Epochs.drop_bad() where data was loaded although no rejection had to be performed (#10718 by new contributor Lukas Gemein)

• Pick also bad channels when selecting T1T2 magnetometers in mne.preprocessing.maxwell_filter() (#10639 by new contributor Matti Toivonen)

• Make color parameter check in in mne.viz.plot_evoked_topo() consistent (#10217 by new contributor T. Wang and Stefan Appelhoff)

• Fix bug in mne.io.read_raw_brainvision() when BrainVision data are acquired with the Brain Products “V-Amp” amplifier and disabled lowpass filter is marked with value 0 (#10517 by new contributor Alessandro Tonin)

• Fix bug in mne.pick_types() and related methods where csd=True was not passed handled properly (#10470 by new contributor Matthias Dold)

• Fix bug where plots produced using the 'qt' / mne_qt_browser backend could not be added using mne.Report.add_figure() (#10485 by Eric Larson)

• Fix bug where theme was not handled properly in mne.io.Raw.plot() (#10487, #10500 by Mathieu Scheltienne and Eric Larson)

• Fix bug in raw.crop(start, stop) that would cause annotations to be erroneously shifted when start != 0 and no measurement date was set. (#10491 by Eric Larson)

• Fix bug in mne.io.read_raw_bti() where unknown electrode locations were not handled properly (#10662 by Eric Larson)

• Fix bug in mne.io.read_raw_ctf() on Windows where large files could not be read (#10866 by Eric Larson)

• Fix bug in mne.io.read_raw_ctf() where invalid measurement dates were not handled properly (#10957 by Jean-Remi King and Eric Larson)

• Rendering issues with recent MESA releases can be avoided by setting the new environment variable``MNE_3D_OPTION_MULTI_SAMPLES=1`` or using mne.viz.set_3d_options() (#10513 by Eric Larson)

• Fix behavior for the pyvista 3D renderer’s quiver3D function so that default arguments plot a glyph in arrow mode (#10493 by Alex Rockhill)

• Retain epochs metadata when using mne.channels.combine_channels() (#10504 by Clemens Brunner)

• Fix epochs indexing with metadata containing boolean type and missing values (#10705 by Clemens Brunner and Alex Gramfort)

• Fix reading of fiducial locations in mne.io.read_raw_eeglab() (#10521 by Alex Gramfort)

• Prevent creation of montage with invalid [x, y, z] coordinates with mne.channels.make_dig_montage() (#10547 by Mathieu Scheltienne)

• Fix bug in coregistration GUI that prevented it from starting up if only a high-resolution head model was available (#10543 by Richard Höchenberger)

• Fix bug with mne.Epochs.add_reference_channels where attributes were not updated properly so subsequent pick_types calls were broken (#10912 by Eric Larson)

• Fix bug in the mne.viz.Brain tool bar that prevented the buttons to call the corresponding feature (#10560 by Guillaume Favelier)

• Fix issue with saving epochs once compute_current_source_density() has been used if a rejection threshold was used first (#10619 by Alex Rockhill and Richard Höchenberger)

• Fix bug in mne.viz.plot_evoked_image() that would cause incorrect sub-titles when using group_by (#10618 by Reza Shoorangiz)

• Reduce memory usage when loading an EDF file with preload=False (#10638 by Clemens Brunner)

• Fix bug in mne.io.Raw.get_channel_types() and related methods where unique=True would return a set with arbitrary order rather than a list that preserves the same order as inst.ch_names (#10720 by Eric Larson)

• In mne.preprocessing.find_bad_channels_maxwell(), do not re-filter the data if a low-pass filter with the requested frequency has already been applied (#10664 by Richard Höchenberger)

• Fix a problem in mne.Evoked.get_peak(), where under certain circumstances the mode parameters 'pos' and 'neg' were not honored when tmin and/or tmax were passed as well (#10686 by Richard Höchenberger)

• mne.read_evokeds(), mne.channels.read_custom_montage(), mne.channels.read_dig_hpts(), mne.channels.read_dig_polhemus_isotrak(), and mne.channels.read_polhemus_fastscan() now correctly expand ~ in the provided path to the user’s home directory (#10685, #10688 by Richard Höchenberger)

• Fix bug in mne.io.read_raw_eeglab() and mne.read_epochs_eeglab(), where the presence of channels without position in the .set file caused positions incorrectly assigned to all further channels (#8754 by Mikołaj Magnuski)

• Combining channels of mne.Epochs or mne.Evoked objects now properly retains baseline information (#10703 by Clemens Brunner)

• In mne.Report, some figures would have an undesired border added to the edges; this has now been resolved (#10730 by Richard Höchenberger)

• Fix selection of EEG channels and selected sphere when plotting bridged electrodes with mne.viz.plot_bridged_electrodes() (#10753 by Mathieu Scheltienne)

• Fix broken links in Fixing BEM and head surfaces (#10765 by Alex Rockhill)

• The built-in FieldTrip channel adjacency matrix for easycapM1 had a bug, where the channel PO8 was incorrectly labeled as PO6 This has been resolved by updating the adjacency matrix to the latest version provided by FieldTrip (#10749 by Richard Höchenberger)

• Add a YouTube video tutorial for editing Freesurfer surfaces in Blender to Fixing BEM and head surfaces (#10778 by Alex Rockhill)

• The default interaction style of mne.Evoked.plot_field() has been changed such that if behaves like plot_alignment() when trying to rotate the head (#10788 by Richard Höchenberger)

• Add a video on how to operate the intracranial electrode contact location GUI in Locating intracranial electrode contacts (#10800 by Alex Rockhill)

• Add a video on how to operate the coregistration GUI in Source alignment and coordinate frames (#10802 by Alex Rockhill)

• Add show and block arguments to mne.gui.coregistration() and mne.gui.locate_ieeg() to pop up the GUIs and halt execution of subsequent code respectively (#10802 by Alex Rockhill)

• Correctly report the number of available projections when printing measurement info in a Jupyter notebook (#10471 by Clemens Brunner)

• Fix value set in raw.orig_format for readers of BrainVision (ASCII format), EGI and Artemis123 files (#10851 by Mathieu Scheltienne)

• When reading EGI MFF files, channel names are no longer ignored and reference channel information is properly incorporated (#10898 by Scott Huberty and Daniel McCloy)

• Fix bug in mne.time_frequency.psd_array_welch() and related functions/methods where the bias was not accounted for when average='median' (#10990 by Eric Larson)

• Fix bug in mne.decoding.TemporalFilter where filter parameters were not handled properly (#10968 by Eric Larson)

• Fix documentation bug in ica.plot_sources to specify that picks keyword argument is for picking ICA components to plot (#10936 by Adam Li)

• Annotations contained in EDF files are correctly read as UTF-8 according to the EDF specification (#10963 by Clemens Brunner)

• Fix bug where complex-valued topoplots caused an error, convert to amplitude instead (#10978 by Alex Rockhill)

API and behavior changes#

• When creating BEM surfaces via mne.bem.make_watershed_bem() and mne.bem.make_flash_bem(), the copy parameter now defaults to True. This means that instead of creating symbolic links inside the FreeSurfer subject’s bem folder, we now create “actual” files. This should avoid troubles when sharing files across different operating systems and file systems (#10531 by Richard Höchenberger)

• The ordering of channels returned by mne.io.read_raw_nirx() is now ordered by channel name, rather than the order provided by the manufacturer. This enables consistent ordering of channels across different file types (#10555 by Robert Luke)

• For mne.viz.plot_topomap(), mne.viz.plot_evoked_topomap(), mne.viz.plot_arrowmap(), mne.viz.plot_ica_components(), mne.Covariance.plot_topomap(), mne.Evoked.plot_topomap(), mne.Evoked.animate_topomap(), mne.decoding.CSP.plot_patterns(), mne.Projection.plot_topomap() and mne.preprocessing.ICA.plot_components() the topomap image interpolation was previously a cubic interpolation but now can be 'linear' and 'nearest' as well. Unless image_interp='nearest' is passed mne uses a subsequent matplotlib bilinear interpolation to make the interpolated image smoother. Previously, 'image_interp' controlled this second interpolation step and, for the first interpolation, the only option was cubic. To simplify, image_interp is now responsible for the main interpolation and the subsequent matplotlib image interpolation is bilinear but can be changed afterward using im.set_interpolation (#10617 by Alex Rockhill)

• The FLASH BEM surface generation via mne.bem.make_flash_bem() or the mne flash_bem command has now been clarified in terms of expected inputs. In mne.bem.convert_flash_mris() the convert parameter is now deprecated and the flash5 and flash30 now accept list of file names or nibabel images. One breaking change is that now the flash files are looked for as mef05_*.mgz and mef30_*.mgz in mri/flash folder (#10715 by Alex Gramfort)

• plot_compare_evokeds() gets a new parameter time_unit which can be "s" or "ms" (#10950 by Daniel McCloy)

• The to_data_frame methods for Raw, Epochs, Evoked, SourceEstimate, EpochsTFR, and AverageTFR all now have default time_format=None (keep time in seconds) instead of time_format='ms' (convert time to milliseconds) (#10925 by Daniel McCloy)

• mne.Evoked.decimate() now aligns to the sample at time=0 (so that the exact sample when the event occurred is preserved) to be consistent with mne.Epochs.decimate() (#10945 by Alex Rockhill)

Authors#

• Adam Li

• Adeline Fecker+

• Alessandro Tonin+

• Alex Ciok+

• Alex Rockhill

• Alexandre Gramfort

• Archit Singhal+

• Ashley Drew+

• Ben Beasley+

• Carina Forster+

• Christian O’Reilly

• Clemens Brunner

• Cora Kim

• Daniel McCloy

• Dominik Welke

• Dominique Makowski

• Eric Larson

• Evan Hathaway

• Frederik D. Weber+

• Guillaume Favelier

• Hamid Maymandi+

• Ilias Machairas+

• Jean-Remi King

• Johann Benerradi

• Jon Houck

• Jona Sassenhagen

• Jordan Drew+

• Lukas Gemein+

• Marijn van Vliet

• Martin Schulz

• Mathieu Scheltienne

• Matt Courtemanche+

• Matthias Dold+

• Matti Toivonen+

• Michiru Kaneda+

• Mikołaj Magnuski

• Mingjian He+

• Naveen Srinivasan+

• Phillip Alday

• Reza Nasri+

• Reza Shoorangiz

• Richard Höchenberger

• Rob Luke

• Scott Huberty+

• Simon Kern

• Sondre Foslien+

• Stefan Appelhoff

• 20. Wang+

• Tziona NessAiver+

Enhancements#

• Add use_orig_time option to mne.Annotations.crop(). (#10396 by new contributor Michiru Kaneda)

• Speed up mne.preprocessing.annotate_muscle_zscore(), mne.preprocessing.annotate_movement(), and mne.preprocessing.annotate_nan() through better annotation creation (#10089 by new contributor Senwen Deng)

• Fix some unused variables in time_frequency_erds.py example (#10076 by new contributor Jan Zerfowski)

• Document BAD_ACQ_SKIP annotation behavior (#10054 by new contributor Etienne de Montalivet)

• Add link to related MNE-ARI package (#10174 by new contributor John Veillette)

• mne.time_frequency.psd_array_multitaper() can now return complex results per-taper when specifying output='complex' (#10307 by Mikołaj Magnuski)

• mne.time_frequency.tfr_array_multitaper() can now return results for output='phase' instead of an error (#10281 by Mikołaj Magnuski)

• Add show local maxima toggling button to mne.gui.locate_ieeg() (#9952 by Alex Rockhill)

• Show boundaries in mne.gui.locate_ieeg() (#10379 by Eric Larson)

• Add argument cval to mne.transforms.apply_volume_registration() to set interpolation values outside the image domain (#10379 by Eric Larson)

• Improve docstring of mne.Info and add attributes that were not covered (#9922 by Mathieu Scheltienne)

• Add an alternate, manual procedure for aligning a CT to an MR procedure to Locating intracranial electrode contacts (#9978 by Alex Rockhill)

• Improve docstring of export functions mne.export.export_raw(), mne.export.export_epochs(), mne.export.export_evokeds(), mne.export.export_evokeds_mff() and issue a warning when there are unapplied projectors (#9994 by Mathieu Scheltienne)

• Add support for reading haemoglobin fNIRS data to mne.io.read_raw_snirf() (#9929 by Robert Luke)

• Add fill_hole_size keyword argument to mne.viz.Brain.add_volume_labels() to close holes in the mesh (#10024 by Alex Rockhill)

• Changed mne.Epochs and mne.Evoked to have a more concise __repr__ to improve interactive MNE usage in Python Interactive Console, IDEs, and debuggers when many events are handled. (#10042 by Jan Sosulski)

• Improve docstring of events arguments and cross-referencing to events (#10056 by Mathieu Scheltienne)

• Speed up repeated surface-smoothing operation (e.g., in repeated calls to stc.plot()) (#10077 by Eric Larson)

• Add verbose parameter to mne.io.Raw.load_bad_channels() and log information on how bad channels are updated (#10102 by Stefan Appelhoff)

• Add infer_type argument to mne.io.read_raw_edf() and mne.io.read_raw_bdf() to automatically infer channel types from channel labels (#10058 by Clemens Brunner)

• Reduce the time it takes to generate a mne.io.Raw, Epochs, or ICA figure if a scalings parameter is provided (#10109 by Richard Höchenberger and Eric Larson)

• mne.Report.add_raw() gained a new scalings parameter to provide custom data scalings for the butterfly plots (#10109 by Richard Höchenberger)

• The butterfly parameter of mne.Report.add_raw() now also accepts numbers to specify how many segments to plot (#10115 by Richard Höchenberger)

• The psd parameter of mne.Report.add_epochs() now also accepts numbers to specify the signal duration used for PSD calculation (#10119 by Richard Höchenberger)

• Drastically speed up butterfly plot generation in mne.Report.add_raw(). We now don’t plot annotations anymore; however, we feel that the speed improvements justify this change, also considering the annotations were of limited use in the displayed one-second time slices anyway (#10114, #10116 by Richard Höchenberger)

• In mne.Report, limit the width of automatically generated figures to a maximum of 850 pixels (450 pixels for mne.SourceEstimate plots), and the resolution to 100 DPI to reduce file size, memory consumption, and – in some cases like mne.Report.add_stc() – processing time (#10126, #10129, #10135, #10142 by Richard Höchenberger)

• Epochs metadata tables are now included in mne.Report (#10166 by Richard Höchenberger)

• mne.get_head_surf() and mne.dig_mri_distances() gained a new parameter, on_defects, controlling how to handle surfaces with topological defects (#10175 by Richard Höchenberger)

• mne.Report.add_epochs() gained a new parameter, drop_log_ignore, to control which drop reasons to omit when creating the drop log plot (#10182 by Richard Höchenberger)

• mne.Epochs.plot_drop_log() now also includes the absolute number of epochs dropped in the title (#10186 by Richard Höchenberger)

• Add a button to show the maximum intensity projection in mne.gui.locate_ieeg() (#10185 by Alex Rockhill)

• Annotations from a Raw object are now preserved by the Epochs constructor and are supported when saving Epochs (#9969 and #10019 by Adam Li)

• Add a checkbox to show and hide the MEG helmet in the coregistration GUI (#10200 by Guillaume Favelier)

• The coregistration GUI (instantiated via mne.gui.coregistration()) gained a new attribute, coreg, to access the mne.coreg.Coregistration instance used by the GUI (#10255 by Richard Höchenberger)

• Add marching cubes display of head if mne.bem.make_scalp_surfaces() has not computed or the recon-all hasn’t finished (#10202 by Alex Rockhill)

• The default interaction style of mne.gui.coregistration() and mne.viz.plot_alignment() has been changed to 'terrain', which keeps one axis fixed and should make interactions with the 3D scene more predictable (#9972, #10206 by Richard Höchenberger)

• mne.gui.coregistration() now uses the proper widget style for push buttons, making for a more native feel of the application (#10220 by Richard Höchenberger and Guillaume Favelier)

• mne.coreg.Coregistration, mne.scale_bem(), and mne.scale_mri() gained a new parameter, on_defects, controlling how to handle topological defects (#10230, #10249 by Richard Höchenberger)

• mne.coreg.Coregistration gained a new attribute, fiducials, allowing for convenient retrieval of the MRI fiducial points (#10243, by Richard Höchenberger)

• Added plotting points to represent contacts on the max intensity projection plot for mne.gui.locate_ieeg() (#10212 by Alex Rockhill)

• Add lines in 3D and on the maximum intensity projection when more than two electrode contacts are selected to aid in identifying that contact for mne.gui.locate_ieeg() (#10212 by Alex Rockhill)

• Add a block parameter to mne.viz.Brain and the UI of mne.coreg.Coregistration to prevent the windows from closing immediately when running in a non-interactive Python session (#10222 by Guillaume Favelier)

• All methods of mne.Report with a tags parameter now also accept a single tag passed as a string (previously, you needed to pass a tuple of strings, even for a single tag) (#10183, by Richard Höchenberger)

• mne.Report.add_trans() has gained a new parameter, alpha, to control the level of opacity of the rendered head (#10247, by Richard Höchenberger)

• The new convenience function mne.event.match_event_names() allows for straightforward checking if a specific event name or a group of events is present in a collection of event names (#10233 by Richard Höchenberger)

• Epochs.plot() is now supported by the pyqtgraph-backend (#10297 by Martin Schulz)

• The map_surface parameter of mne.viz.Brain.add_foci() now works and allows you to add foci to a rendering of a brain that are positioned at the vertex of the mesh closest to the given coordinates (#10299 by Marijn van Vliet)

• mne.preprocessing.ICA.plot_sources() is now also supported by the qt backend (#10330 by Martin Schulz)

• Added mne.viz.Brain.add_dipole() and mne.viz.Brain.add_forward() to plot dipoles on a brain as well as mne.viz.Brain.remove_dipole() and mne.viz.Brain.remove_forward() (#10373 by Alex Rockhill)

• Made anterior/posterior slice scrolling in mne.gui.locate_ieeg() possible for users without page up and page down buttons by allowing angle bracket buttons to be used (#10384 by Alex Rockhill)

• Add support for theme='auto' for automatic dark-mode support in raw.plot() and related functions and methods when using the 'qt' backend (#10417 by Eric Larson)

• Add support for MNE_3D_OPTION_THEME and MNE_BROWSER_THEME to use ['light', 'dark', 'auto'] or a custom theme path in mne.viz.Brain and raw.plot() (and related functions and methods when using the 'qt' backend(), respectively (#10418 by Eric Larson)

• Add support for passing time-frequency data to mne.stats.spatio_temporal_cluster_test() and mne.stats.spatio_temporal_cluster_1samp_test() and added an example to Spatiotemporal permutation F-test on full sensor data (#10384 by Alex Rockhill)

• Add support for reading optical density fNIRS data to mne.io.read_raw_snirf() (#10408 by Robert Luke)

• Added mne.source_space.get_decimated_surfaces() to extract subsurfaces from a SourceSpaces instance or a .fif file (#10421 by Joshua Teves)

Bugs#

• Fix bug with mne.io.read_raw_nihon() where latin-1 channels could not be read (#10429 by new contributor Matthias Eberlein)

• Fix bug with mne.io.read_raw_nihon() so that it handles duplicates in channel names (#10431 by new contributor Matthias Eberlein)

• Fix bug in mne.io.read_raw_egi() where the meas_date in a rawMFF object was incorrectly assigned. The function now converts the meas_date to UTC and provides a UTC offset key, i.e. raw.info['utc_offset']. (#10304 by new contributor Scott Huberty)

• Fix datetime conversion for tmin/tmax=None cases in mne.Annotations.crop(). Allow the use of float and None simultaneously for mne.Annotations.crop(). (#10361 by new contributor Michiru Kaneda)

• Add Shift_JIST mu in mne.io.read_raw_edf() (#10356 by new contributor Michiru Kaneda)

• Teach mne.io.read_raw_bti() to use its eog_ch parameter (#10093 by new contributor Adina Wagner)

• Fix use of arguments in numpy.loadtxt() (#10189 by new contributor Federico Zamberlan)

• Fix documentation of options in mne.stc_near_sensors() (:gh:`` by new contributor Nikolai Chapochnikov)

• mne.time_frequency.tfr_array_multitaper() now returns results per taper when output='complex' (#10281 by Mikołaj Magnuski)

• Fix default of mne.io.Raw.plot() to be use_opengl=None, which will act like False unless MNE_BROWSER_USE_OPENGL=true is set in the user configuration (#9957 by Eric Larson)

• Fix bug with mne.Report where figures were saved with bbox_inches='tight', which led to inconsistent sizes in sliders (#9966 by Eric Larson)

• When opening a saved report and saving it to a different filename again, don’t change Report.fname to avoid a regression when using open_report() as a context manager (#9998 by Marijn van Vliet)

• Fix bug in mne.make_forward_solution() where sensor-sphere geometry check was incorrect (#9968 by Eric Larson)

• Use single char alphanumeric suffix when renaming long channel names (over 15-characters) when writing to FIF format. (#10002 by Luke Bloy)

• Add argument overwrite to mne.export.export_raw(), mne.export.export_epochs(), mne.export.export_evokeds() and mne.export.export_evokeds_mff() (#9975 by Mathieu Scheltienne)

• mne.gui.coregistration() and the mne coreg command didn’t respect the interaction parameter (#9972 by Richard Höchenberger)

• Fix incorrect projection of source space onto white matter surface instead of pial in Working with ECoG data (#9980 by Alex Rockhill)

• Fix channel type support when reading from EEGLAB .set format with mne.io.read_raw_eeglab() and mne.read_epochs_eeglab() (#9990 by Mathieu Scheltienne)

• Fix suboptimal alignment using mne.transforms.compute_volume_registration() (#9991 by Alex Rockhill)

• Only warn if header is missing in BrainVision files instead of raising an error (#10001 by Clemens Brunner)

• Add argument overwrite to mne.preprocessing.ICA.save to check for existing file (#10004 by Mathieu Scheltienne)

• mne.Report now raises an exception if invalid tags were passed (#9970 by Richard Höchenberger)

• Fix bug in mne.get_montage_volume_labels() that set the maximum number of voxels to be included too low causing unwanted capping of the included voxel labels (#10021 by Alex Rockhill)

• Fix annotation cropping and I/O roundtrip when there is no measurement date available (#10040 by Mathieu Scheltienne and Alex Gramfort).

• Fix bug where raw.crop() from the beginning of the instance could break annotations and raw.first_samp (#10358 by Eric Larson)

• sys_info() output now contains the installed version of pooch, too; this output had been accidentally removed previously (#10047 by Richard Höchenberger)

• Fix VTK version extraction in mne.sys_info() (#10399 by Eric Larson)

• Fix automatic channel type detection from channel labels in mne.io.read_raw_edf() and mne.io.read_raw_bdf() (and disable this functionality from mne.io.read_raw_gdf()) (#10058 by Clemens Brunner)

• Fix permutation_cluster_1samp_test() to properly handle 2-dimensional data in combination with TFCE (#10073 by Richard Höchenberger)

• Fix channel grouping error when using “butterfly mode” with mne.io.Raw.plot() (#10087 by Daniel McCloy)

• Fix inconsistent behavior of mne.preprocessing.annotate_* functions by making them all return mne.Annotations objects with the orig_time attribute set to raw.info["meas_time"] (#10067 and #10118 by Stefan Appelhoff, Eric Larson, and Alex Gramfort)

• Fix bug that appears during automatic calculation of the colormap of mne.viz.Brain when data values of fmin and fmax are too close (#10074 by Guillaume Favelier)

• We now display a scrollbar in the tags dropdown of a Report if many tags have been added, granting access to all tags instead of “hiding” them below the bottom of the page (#10082 by Richard Höchenberger)

• Creating mne.Epochs now provides clearer logging (less ambiguous, no duplicates) when the preload and/or metadata parameters are set (#10112 by Stefan Appelhoff)

• Fix bug with mne.Epochs where save-load round-trip with FIF would cause mne.Epochs.apply_baseline() to no longer work (#10177 by Eric Larson)

• Fix functions by adding missing overwrite parameters: mne.write_events(), mne.write_cov(), mne.write_evokeds(), mne.SourceEstimate.save(), mne.minimum_norm.write_inverse_operator(), mne.write_proj(), and related methods (#10127 by Eric Larson)

• Fix bug with mne.transforms.compute_volume_registration() and mne.compute_source_morph() (volumetric) where the smoothing factors were not scaled based on zooms (#10132 by Eric Larson)

• Remove repeated logging output when overwriting an existing Raw file (#10095 by Richard Höchenberger and Stefan Appelhoff)

• In the plots generated by mne.Report.add_stc(), we now only add 5 labels to the color bar to reduce the chance of overlap, which could previously cause the labels to become unreadable (#10135 by Richard Höchenberger)

• mne.Report.add_trans() now allows you to add sensor alignment plots for head surfaces that have topological defects (#10175 by Richard Höchenberger)

• mne.Report.add_trans() now also works if no digitization points are present in the data (#10176 by Jeff Stout)

• Argument verbose is now respected by dataset fetching (#10210 by Mathieu Scheltienne)

• Fix bug with mne.io.read_raw_hitachi() where empty Comment descriptions were not handled properly (#10235 by Eric Larson)

• Fix bug with input validation of low-level filtering functions (#10267 by Eric Larson)

• mne.gui.coregistration() now works with surfaces containing topological defects (#10230, by Richard Höchenberger)

• Fix bug with mne.io.read_raw_nirx() being unable to read measurement dates recorded on systems with German (de_DE), French (fr_FR), and Italian (it_IT) locales (#10277 by Eric Larson)

• Fix bug with projector normalization checks that were too sensitive, and improve warning (#10292 by Eric Larson)

• Fix bug with mne.viz.plot_alignment() where head-coordinate source spaces (e.g., from a forward solution) were not properly plotted (#10309 by Eric Larson)

• mne.read_trans() and mne.io.read_fiducials() now correctly expand ~ in the provided path (i.e., to the user’s home directory) (#10265, #10415, by Richard Höchenberger)

• mne.find_events() now uses first_samp and not 0 for initial event when using initial_value (#10289, by Alex Gramfort)

• Fix bug with mne.channels.make_standard_montage() for 'standard*', 'mgh*', and 'artinis*' montages where the points were incorrectly scaled and fiducials incorrectly set away from the correct values for use with the fsaverage subject (#10324 by Eric Larson)

• Fix bug with mne.Report.add_figure() where figures generated externally were closed and possibly resized during render (#10342 by Eric Larson)

• Fix bug with mne.viz.plot_sparse_source_estimates() where the return value was incorrect (#10347 by Eric Larson)

• Fix plotting bug in How to convert 3D electrode positions to a 2D image and make view look more natural in Annotate movement artifacts and reestimate dev_head_t (#10313, by Alex Rockhill)

• Fix bug with blank 3D rendering with MESA software rendering (#10400 by Eric Larson)

• Fix a bug in mne.gui.locate_ieeg() where 2D lines on slice plots failed to update and were shown when not in maximum projection mode (#10335, by Alex Rockhill)

• Fix misleading color scale in Non-parametric between conditions cluster statistic on single trial power for the plotting of cluster F-statistics (#10393 by Alex Rockhill)

• Fix baseline removal using remove_dc=True in raw.plot() for data containing np.nan (#10392 by Clemens Brunner)

• Fix misleading color scale in Mass-univariate twoway repeated measures ANOVA on single trial power for plotting F-stats (#10401 by Alex Rockhill)

• Fix misleading T_obs return name for mne.stats.spatio_temporal_cluster_test() when the default returns an F-statistic (#10401 by Alex Rockhill)

API changes#

• The default browser for raw.plot(), epochs.plot(), and ica.plot_sources() has been changed to the 'qt' backend on systems where mne_qt_browser is installed. To change back to matplotlib within a session, you can use mne.viz.set_browser_backend('matplotlib'). To set it permanently on your system, you can use mne.set_config('MNE_BROWSER_BACKEND', 'matplotlib') (#9960 by Martin Schulz and Eric Larson)

• mne.Info.pick_channels has been deprecated. Use inst.pick_channels to pick channels from Raw, Epochs, and Evoked. Use mne.pick_info() to pick channels from mne.Info (#10039 by Mathieu Scheltienne)

• All data_path functions now return pathlib.Path objects rather than strings. Support for string concatenation with plus (+) is thus deprecated and will be removed in 1.2, use the forward-slash / operator instead (#10348 by Eric Larson)

• Argument event_list has been deprecated in favor of events in mne.write_events() (#10056 by Mathieu Scheltienne)

• mne.preprocessing.annotate_flat has been deprecated in favor of mne.preprocessing.annotate_amplitude(), that covers both minimum and maximum peak-to-peak variation. (#10143 by Mathieu Scheltienne)

• The max_ori_out parameter of mne.beamformer.apply_lcmv() and related functions is being removed as only signed estimates are supported. abs(stc) can be used to obtain unsigned estimates (#10366 by Eric Larson)

• The verbose attribute of classes (e.g., mne.io.Raw, mne.Epochs, etc.) has been deprecated. Explicitly pass verbose to methods as necessary instead. (#10267 by Eric Larson)

• In mne.viz.set_browser_backend(), the mne-qt-browser-based backend is now called 'qt' rather than 'pyqtgraph' for simplicity (#10323 by Eric Larson)

Dependencies#

Numerous external dependencies that used to be bundled with MNE-Python are now not shipped with the package anymore and will instead be retrieved automatically from their official sources when you install MNE-Python. This simplifies MNE-Python maintenance and keeps the package smaller. The following new dependencies have been added:

• Jinja2 (replaces Tempita, which is not maintained anymore; #10211 by Richard Höchenberger)

Authors#

• Adam Li

• Adina Wagner+

• Alex Rockhill

• Alexandre Gramfort

• Britta Westner

• Clemens Brunner

• Daniel McCloy

• Eduard Ort

• Eric Larson

• Etienne de Montalivet+

• Federico Zamberlan+

• Guillaume Favelier

• Jan Sosulski

• Jan Zerfowski+

• Jeff Stout

• John Veillette+

• Joshua Teves

• Julia Guiomar Niso Galán

• Luke Bloy

• Marijn van Vliet

• Martin Schulz

• Mathieu Scheltienne

• Matthias Eberlein+

• Michiru Kaneda+

• Mikołaj Magnuski

• Nikolai Chapochnikov+

• Richard Höchenberger

• Robert Luke

• Scott Huberty+

• Senwen Deng+

• Stefan Appelhoff

• Steve Matindi

• Thomas Hartmann

Enhancements#

• Add pooch to system information reports (#9801 by new contributor Joshua Teves)

• Get annotation descriptions from the name field of SNIRF stimulus groups when reading SNIRF files via mne.io.read_raw_snirf (#9575 by new contributor Darin Erat Sleiter)

• Add support for NIRSport and NIRSport2 devices to mne.io.read_raw_nirx (#9348 and #9401 by new contributor David Julien, new contributor Romain Derollepot, Robert Luke, and Eric Larson)

• New function mne.Dipole.to_volume_labels() to get atlas label for MRI coordinates. (#9376 by new contributor Marian Dovgialo)

• Add support for SURE parameter selection in mne.inverse_sparse.mixed_norm() and make alpha parameter now default to 'sure' (#9430 by new contributor Pierre-Antoine Bannier and Alex Gramfort)

• Speed up BCD solver in mne.inverse_sparse.mixed_norm() by adding Anderson acceleration (#9481 by new contributor Pierre-Antoine Bannier and Alex Gramfort)

• Add support for list of channel types for EEG/sEEG/ECoG/DBS referencing (#9637 by new contributor Mathieu Scheltienne)

• Add the ability to set point size and line width in mne.viz.plot_sensors() and related functions (#9872 by new contributor Dominique Makowski)

• Add support for changing the default event id used by mne.make_fixed_length_epochs() (#9782 by new contributor Mathieu Scheltienne)

• Improve documentation for arguments threshold, method and measure of ICA methods find_bads_ecg, find_bads_ref and find_bads_eog (#9858 by new contributor Mathieu Scheltienne)

• Add support for more than 3 source estimates in mne.viz.plot_sparse_source_estimates() (#9640 by new contributor Pierre-Antoine Bannier and Alex Gramfort)

• Add cortical signal suppression (CSS) mne.preprocessing.cortical_signal_suppression() (#6888, #9892 by new contributor John Samuelsson and Eric Larson)

• Show all good channel types and counts when printing a mne.Info in the notebook (#9725 by Valerii Chirkov and Eric Larson)

• Speed up point decimation in mne.io.read_raw_kit() by vectorization and use of scipy.spatial.cKDTree (#9568 by Jean-Remi King and Eric Larson)

• Add ability to export EDF+ files using mne.export.export_raw() (#9643 by Adam Li)

• New function mne.chpi.get_chpi_info() to retrieve basic information about the cHPI system used when recording MEG data (#9369 by Richard Höchenberger)

• New namespace mne.export created to contain functions (such as mne.export.export_raw and mne.export.export_epochs) for exporting data to non-FIF formats (#9427 by Eric Larson)

• Add support for Hitachi fNIRS devices in mne.io.read_raw_hitachi (#9391 by Eric Larson)

• Add support for picks in mne.stc_near_sensors() (#9396 by Eric Larson)

• Add label methods mne.Label.compute_area and mne.Label.distances_to_outside (#9424 by Eric Larson)

• Add projections when printing a mne.Info in the notebook (#9403 by Alex Gramfort)

• Add new function mne.bem.make_scalp_surfaces() to compute scalp surfaces for coregistration (#9463 by Alex Gramfort)

• Add support for interpolating oxy and deoxyhaemoglobin data types (#9431 by Robert Luke)

• Add support for SNIRF files in mne.Report (#9443 by Robert Luke)

• Add support for exporting MFF evoked files using mne.export.export_evokeds and mne.export.export_evokeds_mff (#9406 by Evan Hathaway)

• mne.concatenate_raws(), mne.concatenate_epochs(), and mne.write_evokeds() gained a new parameter on_mismatch, which controls behavior in case not all of the supplied instances share the same device-to-head transformation (#9438 by Richard Höchenberger)

• Add support for multiple datablocks (acquistions with pauses) in mne.io.read_raw_nihon() (#9437 by Federico Raimondo)

• Add new function mne.preprocessing.annotate_break() to automatically detect and mark “break” periods without any marked experimental events in the continuous data (#9445 by Richard Höchenberger)

• Add “time_format” to mne.io.Raw.plot() and mne.preprocessing.ICA.plot_sources() to display acquisition time on x-axis (#9419 by Martin Schulz)

• Speed up mne.decoding.TimeDelayingRidge.predict() by switching to FFT-based convolution (#9458 by Ross Maddox)

• Show multiple colors and linestyles for excluded components with mne.Evoked in mne.preprocessing.ICA.plot_sources() (#9444 by Martin Schulz)

• Add functions for aligning MRI and CT data mne.transforms.compute_volume_registration and mne.transforms.apply_volume_registration (#9503 by Alex Rockhill and Eric Larson)

• Add tutorial for how to processes image (CT and MR) files in order to localize electrode contacts for intracranial recordings Locating intracranial electrode contacts (#9484 by Alex Rockhill)

• Add vertex number in source space for mne.VolSourceEstimate.plot() and mne.viz.plot_volume_source_estimates() (#9697 by Eric Larson)

• Add support for colormap normalization in mne.viz.plot_topomap() (#9468 by Clemens Brunner)

• Update mne.preprocessing.realign_raw() with Numpy-recommended polynomial fitting method (#9514 by Erica Peterson)

• Add a warning to mne.preprocessing.nirs.beer_lambert_law when source-detector distances are zero (#9541 by Eric Larson)

• Add exclude parameter to mne.io.Raw.plot_psd() and mne.Epochs.plot_psd() (#9519 by Clemens Brunner)

• Add mne.Annotations.rename to rename annotation descriptions (#9525 by Robert Luke)

• Add mne.Annotations.set_durations to set annotation durations (#9529 by Robert Luke)

• The exclude parameter in mne.io.read_raw_edf, mne.io.read_raw_bdf, and mne.io.read_raw_gdf now also accepts a regular expression (#9558 by Clemens Brunner)

• Add mne.Evoked.get_data() method to mne.Evoked (#9555 by Stefan Appelhoff)

• The mne.Epochs.get_data() method now has a units parameter (#9553 by Stefan Appelhoff)

• Add mne.warp_montage_volume() to use an anatomical volume with channel locations at high or low insensities to apply a symmetric diffeomorphic registration to a template (e.g. fsaverage) space (#9544 by Alex Rockhill)

• mne.io.Raw.get_data(), mne.Epochs.get_data(), and mne.Evoked.get_data() methods now have tmin and tmax parameters to specify data window in seconds to get (#9556 by Stefan Appelhoff)

• The mne.Evoked.get_data() method now has a units parameter (#9578 by Stefan Appelhoff)

• Add mne.chpi.compute_chpi_snr and mne.viz.plot_chpi_snr for computing and plotting the time-varying SNR of continuously-active HPI coils (#9570 by Daniel McCloy and Jussi Nurminen)

• Add mne.get_montage_volume_labels() to find the regions of interest in a Freesurfer atlas anatomical segmentation for an intracranial electrode montage and mne.viz.plot_channel_labels_circle() to plot them (#9545 by Alex Rockhill)

• Add mne.viz.Brain.add_volume_labels() to plot subcortical surfaces and other regions of interest (#9540 by Alex Rockhill and Eric Larson)

• Add custom cortex curvature colors in mne.viz.Brain via the cortex argument (#9750 by Eric Larson)

• Add mne.channels.DigMontage.apply_trans() to apply a transform directly to a montage (#9601 by Alex Rockhill)

• mne.preprocessing.ICA.fit() now emits a warning if any of the start, stop, reject, and flat parameters are passed when performing ICA on Epochs. These parameters only have an effect on Raw data and were previously silently ignored in the case of Epochs (#9605 by Richard Höchenberger)

• mne.viz.plot_alignment() now emits a warning (instead of aborting with an exception) if the surfaces contain topological defects (#9614 by Richard Höchenberger)

• The coregistration GUI can now display the head surface even if there are topological defects. Previously, a low-resolution standard head unsuitable for individualized coregistration was displayed (#9614 by Richard Höchenberger)

• New mne.coreg.Coregistration object to allow coregistration scripting without a dependency on mayavi (#9516 by Guillaume Favelier).

• Add mne.viz.Brain.add_head() to plot the head surface (#9618 by Alex Rockhill)

• Add mne.channels.read_dig_localite() to read Localite electrode location files (#9658 by Clemens Brunner)

• Add mne.viz.Brain.add_sensors() to plot sensor locations (#9585 by Alex Rockhill)

• Add mne.coreg.estimate_head_mri_t() to estimate the head->mri transform from fsaverage fiducials (#9585 by Alex Rockhill)

• Add mne.channels.DigMontage.add_mni_fiducials() to add fiducials to a montage in MNI coordinates (#9682 by Alex Rockhill)

• Add mne.channels.DigMontage.remove_fiducials() to remove fiducials under rare circumstances when the coordinate frame should not be set to “head” (#9684 by Alex Rockhill)

• Add remove methods for mesh objects in mne.viz.Brain (#9688 by Alex Rockhill)

• Add --show-paths and --developer options to mne sys_info (#9789 by Eric Larson)

• Add ability to export EDF+ files using mne.export.export_raw() (#9643 by Adam Li)

• Reading EDF files via mne.io.read_raw_edf() now can infer channel type from the signal label in the EDF header (#9694 by Adam Li)

• Add mne.gui.locate_ieeg() to locate intracranial electrode contacts from a CT, an MRI (with Freesurfer recon-all) and the channel names from an mne.Info object (#9586 by Alex Rockhill)

• All data fetching code now relies on pooch, which is an added optional requirement for dataset fetchers (#9742 by Adam Li and Daniel McCloy)

• Add phantom FreeSurfer subject fetcher mne.datasets.fetch_phantom() to allow fetching the Otaniemi phantom (#9796 by Eric Larson)

• Add mne.datasets.fetch_dataset() to allow packages to fetch datasets in a similar fashion to MNE (#9763 by Adam Li)

• Add full ECoG dataset to MNE-misc-data and demonstrate its use in How to convert 3D electrode positions to a 2D image and Locating intracranial electrode contacts (#9784 by Alex Rockhill)

• Add options tol and accuracy to mne.fit_dipole() to control optimization (#9810 by Eric Larson)

• Completely revamp the Report experience: new HTML layout, many new methods, more flexibility; the functionality is demonstrated in Getting started with mne.Report (#9754, #9828, #9847, #9860, #9861, #9862, #9863, #9871, #9881, #9891, #9901 by Richard Höchenberger, Eric Larson, and Alex Gramfort)

• Add basic HTML representations of Forward and InverseOperator instances for a nicer Jupyter experience (#9754 by Richard Höchenberger)

• mne.io.read_raw_brainvision now handles ASCII data with comma-separated values, as may be exported from BrainVision Analyzer (#9795 by Richard Höchenberger)

• Add mne.preprocessing.ieeg.project_sensors_onto_brain() to project ECoG sensors onto the pial surface to compensate for brain shift (#9800 by Alex Rockhill)

• Add support for data acquired with NIRx devices using Aurora software version 2021.9 (#9800 by Robert Luke, and Eric Larson)

• All functions for reading and writing files should now automatically handle ~ (the tilde character) and expand it to the user’s home directory. Should you come across any function that doesn’t do it, please do let us know! (#9613, #9845 by Richard Höchenberger)

• All functions accepting a FreeSurfer subjects directory via a subjects_dir parameter can now consume pathlib.Path objects too (used to be only strings) (#9613 by Richard Höchenberger)

• Add support for colormap normalization in mne.time_frequency.AverageTFR.plot() (#9851 by Clemens Brunner)

• Add support for BIDS-compatible filenames when splitting big epochs files via the new split_naming parameter in mne.Epochs.save() (#9869 by Denis Engemann)

• Add by_event_type parameter to mne.Epochs.average() to create a list containing an mne.Evoked object for each event type (#9859 by Marijn van Vliet)

• Add pyqtgraph as a new backend for mne.io.Raw.plot() (#9687 by Martin Schulz)

• Add mne.viz.set_browser_backend(), mne.viz.use_browser_backend() and mne.viz.get_browser_backend() to set matplotlib or pyqtgraph as backend for mne.io.Raw.plot() (#9687 by Martin Schulz)

Bugs#

• Fix bug in mne.io.Raw.pick() and related functions when parameter list contains channels which are not in info instance (#9708 by new contributor Evgeny Goldstein)

• Fix a few typos (#9706 by new contributor Timothy Gates)

• Fix bug with mne.Epochs.crop() and mne.Evoked.crop() when include_tmax=False, where the last sample was always cut off, even when tmax > epo.times[-1] (#9378 by new contributor Jan Sosulski)

• Fix bug with mne.io.read_raw_curry to allow reading Curry 7 and 8 event files with ‘.ceo’ and ‘.cdt.ceo’ extensions (#9381, #9712 by new contributor Xiaokai Xia, Daniel McCloy, and by new contributor Reza Shoorangiz)

• Fix bug with mne.compute_rank, mne.cov.compute_whitener and mne.preprocessing.ICA.fit when explicit picks including bad channels are provided (#9719 by new contributor Mathieu Scheltienne)

• Fix bug with mne.concatenate_epochs() where the concatenated events could end up in non-chronological order. (#9765 by new contributor Jan Sosulski)

• Fix bug where mne.time_frequency.psd_array_welch() and mne.time_frequency.psd_array_multitaper() failed to handle negative integer for n_jobs argument (#9849 by new contributor Riessarius Stargardsky)

• Fix bug where resampling returns empty instances (#9897 by new contributor Mathieu Scheltienne and Eric Larson)

• Fix bug with mne.io.Raw.get_montage now returning both good and bad channels in the montage. (#9920 by new contributor Mathieu Scheltienne)

• Fix bug with mne.io.read_raw_nihon() where latin-1 annotations could not be read (#9384 by Alex Gramfort)

• Fix bug when printing a mne.io.RawArray in the notebook (#9404 by Alex Gramfort)

• Fix bug when computing rank from info for SSS data with only gradiometers or magnetometers (#9435 by Alex Gramfort)

• Fix bug with mne.preprocessing.find_bad_channels_maxwell where all-flat segments could lead to an error (#9531 by Eric Larson)

• Fix bug with mne.io.Raw.set_montage and related functions where the channel coordinate frame was not properly set to head (#9447 by Eric Larson)

• Fix bug with mne.io.read_raw_fieldtrip and mne.read_epochs_fieldtrip where channel positions were not set properly (#9447 by Eric Larson)

• Fix bug with mne.io.read_raw_kit() where omitting HPI coils could lead to an AssertionError on reading (#9612 by Eric Larson)

• Fix bug with mne.io.read_raw_edf() where reading only stimulus channels would lead to an error (#9831 by Eric Larson)

• Fix bug with mne.preprocessing.nirs.optical_density where protection against zero values was not guaranteed (#9522 by Eric Larson)

• mne.concatenate_raws() now raises an exception if raw.info['dev_head_t'] differs between files. This behavior can be controlled using the new on_mismatch parameter (#9438 by Richard Höchenberger)

• Fixed bug in mne.Epochs.drop_bad() where subsequent rejections failed if they only specified thresholds for a subset of the channel types used in a previous rejection (#9485 by Richard Höchenberger).

• Fix bug with mne.simulation.simulate_evoked, mne.apply_forward, and mne.apply_forward_raw where systems with EEG channels that come before MEG channels would have them mixed up in the output evoked or raw object (##9513 by Eric Larson)

• In mne.viz.plot_ica_scores() and mne.preprocessing.ICA.plot_scores(), the figure and axis titles no longer overlap when plotting only a single EOG or ECG channel (#9489 by Richard Höchenberger).

• Ensure mne.io.Raw.get_montage works with SNIRF data (#9524 by Robert Luke)

• Fix bug in mne.setup_volume_source_space() where non-finite positions could be used in a discrete source space (#9603 by Eric Larson)

• Fix bug in mne.viz.plot_topomap() (and related methods like mne.Evoked.plot_topomap()) where large distances between electrodes (higher than head radius) would lead to an error (#9528 by Mikołaj Magnuski).

• Fix bug in mne.viz.plot_topomap (and related methods) where passing axes that are part of a matplotlib figure that uses a constrained layout would emit warnings (#9558 by Eric Larson)

• Fix bug in mne.make_forward_solution() where impossible sensor-BEM geometries were not checked (#9893 by Eric Larson)

• Fix bug in mne.concatenate_epochs() when concatenating mne.Epochs objects with 0 events (#9535 by Marijn van Vliet)

• Fix bug in mne.viz.Brain.screenshot() where the RGBA mode was not supported (#9564 by Guillaume Favelier).

• Fix bug in mne.io.read_raw_egi() where reading data from a data segment that is part of an acquisition skip would lead to an error (#9565 by Eric Larson)

• Prevent mne.io.Raw.plot() windows from spawning at sizes too small to properly render (#9629 by Daniel McCloy)

• Fix bug where mne.io.Raw.plot() raised an error when plotting raw dipole data (#9771 by Eric Larson)

• Fix bug in :mne.preprocessing.ICA.find_bads_ecg() where passing start and stop lead to erroneous data windows depending on the combination of Raw, Epochs, Evoked, and the type (int, float, None) of start and stop (#9556 by Stefan Appelhoff)

• Fix bug in mne.viz.set_3d_backend() and mne.viz.get_3d_backend() where the PyVistaQt-based backend was ambiguously named 'pyvista' instead of 'pyvistaqt'; use set_3d_backend('pyvistaqt') and expect 'pyvistaqt' as the output of mne.viz.get_3d_backend() instead of 'pyvista', and consider using get_3d_backend().startswith('pyvista') for example for backward-compatible conditionals (#9607 by Guillaume Favelier)

• Fix bug in mne.viz.plot_compare_evokeds() where confidence bands were not drawn if only one condition was plotted (#9663 by Daniel McCloy)

• Fix bug where setting of a montage with fNIRS data got set to “unknown” coordinate frame when it should have been in “head” (#9630 by Alex Rockhill)

• Fix bug where “seeg”, “ecog”, “dbs” and “fnirs” data had coordinate frame unknown upon loading from a file when it should have been in “head” (#9580 by Alex Rockhill)

• Raise error when no trans is provided to mne.viz.plot_alignment() when required instead of assuming identitiy head->mri transform (#9585 by Alex Rockhill)

• Fix bug where mne.Epochs.equalize_event_counts() failed when only one good epoch existed for one of the event types (#9758 by Daniel McCloy)

• Fix bug where channels with a dollar sign ($) were not being labeled “misc” in mne.io.read_raw_nihon() (#9695 by Adam Li)

• Fix bug where mne.io.read_raw_persyst() was lower-casing events it found in the .lay file (#9746 by Adam Li)

• Fix bug with Qhull when plotting OPM sensors in mne.viz.plot_alignment() (#9799 and #9804 by Eric Larson)

• Fix bug where mne.io.read_raw_snirf() was including the landmark index as a spatial coordinate (#9777 by Robert luke)

• Fix bug where mne.Annotations were not appending channel names when being added together (#9780 by Adam Li)

• mne.viz.plot_evoked_joint now correctly handles a times='peaks' parameter value (#9754 by Richard Höchenberger)

• In mne.viz.plot_compare_evokeds, slightly adjust the x-axis limit calculation to avoid drawing an axis that is too short (#9754 by Richard Höchenberger)

• Fix a bug with mne.viz.snapshot_brain_montage() with the Mayavi backend where electrodes were shown in the wrong positions, and with the PyVista backend where the snapshot always contained the electrodes but could be missing a hemisphere (#9933 by Eric Larson)

• In mne.io.Raw.plot_psd() and mne.Epochs.plot_psd(), avoid overlap between subplot titles and x-axis tick labels (#9864 by Richard Höchenberger)

• In mne.Epochs.plot_psd_topomap(), the data is now scaled to match the output of mne.Epochs.plot_psd() (#9873 by Richard Höchenberger)

• Fix channel locations with NaN values causing all channel locations not to be plotted in mne.viz.Brain.add_sensors() (#9911 by Alex Rockhill)

• Fix infinite loop bug in mne.get_montage_volume_labels() (#9940 by Alex Rockhill)

API changes#

• The mne.Info class is now more strictly checked when attributes are set. Setting most attributes directly will lead to an error in the next release. See the docstring of mne.Info for more information (#9867 by new contributor Mathieu Scheltienne)

• The Mayavi 3D backend has been deprecated (#9904 by Eric Larson):

  • In 3D plotting functions, use PyVista-based backends ('pyvistaqt' and 'notebook') as PySurfer/Mayavi will be removed in the next version

  • To use the mne kit2fiff command-line GUI for fiducial markers, install the separate mne-kit-gui module, which can still be launched via $ mne kit2fiff

  • The mne.gui.fiducials GUI has been deprecated, use mne coreg or mne_kit_gui.fiducials() to set MRI fiducials instead

• In mne.compute_source_morph, the niter_affine and niter_sdr parameters have been replaced by niter and pipeline parameters for more consistent and finer-grained control of registration/warping steps and iteration (#9505 by Alex Rockhill and Eric Larson)

• Split mne.viz.Brain.show_view() argument view into azimuth, elevation and focalpoint for clearer view setting and make the default for row and col apply to all rows and columns (#9596 by Alex Rockhill)

• Deprecate solver='prox' in mne.inverse_sparse.mixed_norm() in favor of solver='cd' and solver='bcd' as coordinate descent solvers consistently outperform proximal gradient descent (#9608 by Pierre-Antoine Bannier)

• All mne.connectivity functions have moved to the mne-connectivity package; they are deprecated in MNE-Python and will be removed in version 0.25 (#9493 by Adam Li).

• mne.inverse_sparse.mixed_norm() now simply warns when source estimates contain no dipole, e.g. if data are too noisy and alpha is based on SURE (#9685 by Alex Gramfort)

• Deprecate functions mne.datasets.DATASET_NAME.has_DATASET_NAME_data() for these datasets: epilepsy_ecog, fnirs_motor, multimodal, opm, phantom_4dbti, refmeg_noise, sample, somato, and ssvep. Use the generic mne.datasets.has_dataset() instead (#9781 by Daniel McCloy and Adam Li)

• mne.Report modernization has led to multiple deprecations (#9754 by Richard Höchenberger):

  • mne.Report.add_figs_to_section and mne.Report.add_slider_to_section have been deprecated in favor of mne.Report.add_figure()

  • mne.Report.add_images_to_section has been deprecated in favor of mne.Report.add_image()

  • mne.Report.add_bem_to_section has been deprecated in favor of mne.Report.add_bem()

  • mne.Report.add_htmls_to_section has been deprecated in favor of mne.Report.add_html()

  • In mne.Report.parse_folder(), the sort_sections parameter has been deprecated in favor of sort_content

  • In mne.Report.remove(), the caption and section parameters have been deprecated in favor of title and tags, respectively

  • The mne.Report.sections attribute has been deprecated in favor of mne.Report.tags

  • The mne.Report.fnames attribute has been deprecated without replacement

• mne.Epochs.plot() and mne.viz.plot_epochs() gained parameter show_scalebars analogous to mne.viz.plot_raw() (#9815 by Daniel McCloy)

• The output folder name for HF_SEF datasets is now hf_sef instead of HF_SEF (#9763 by Adam Li)

• Deprecate mne.viz.utils.center_cmap (#9851 by Clemens Brunner)

• The default partial pathlength factor of mne.preprocessing.nirs.beer_lambert_law() will change from 0.1 in 0.24 to 6.0 in the next release (#9843 by Robert Luke)

• mne.preprocessing.ICA.detect_artifacts has been deprecated. Please use find_bads_eog and find_bads_ecg instead (#9909 by Richard Höchenberger)

Authors#

People who contributed to this release in alphabetical order (people with a + are first time contributors):

• Adam Li

• Alex Rockhill

• Alexandre Gramfort

• Britta Westner

• Clemens Brunner

• Daniel McCloy

• Darin Erat Sleiter+

• David Julien+

• Denis A. Engemann

• Dominique Makowski+

• Eric Larson

• Erica Peterson

• Evan Hathaway

• Evgeny Goldstein+

• Fede Raimondo

• Gansheng Tan+

• Guillaume Favelier

• Hubert Banville

• Jan Sosulski+

• Jean-Remi King

• Jeff Stout

• Johann Benerradi

• John Samuelsson+

• Josh Koen+

• Joshua Teves+

• Lukas Hecker+

• Luke Bloy

• Marian Dovgialo+

• Marijn van Vliet

• Martin Schulz

• Mathieu Scheltienne+

• Mikołaj Magnuski

• Pierre-Antoine Bannier+

• Reza Shoorangiz+

• Richard Höchenberger

• Riessarius Stargardsky+

• Robert Luke

• Romain Derollepot+

• Ross Maddox

• Ryan Law+

• Stefan Appelhoff

• Steve Matindi+

• Timothy Gates+

• Valerii Chirkov

• Xiaokai Xia+

Enhancements#

• Add support for exporting to EEGLAB’s set format with eeglabio with new methods mne.io.Raw.export() and mne.Epochs.export(). (#9192 by new contributor Jack Zhang)

• mne.viz.plot_evoked_topo() now uses a Minimum/Maximum of peaks for each channel type to calculate ylim when input param is None (#9285 by new contributor Ram Pari)

• Add exclude parameter to mne.viz.plot_evoked_topo() (#9278 by new contributor Ram Pari)

• Add mne.time_frequency.EpochsTFR.to_data_frame() and mne.time_frequency.AverageTFR.to_data_frame() methods (#9124 by new contributor Felix Klotzsche)

• Add selection and drop_log to mne.time_frequency.EpochsTFR (#9124 by new contributor Felix Klotzsche)

• Add CSV, TSV, and XYZ support to mne.channels.read_custom_montage() (#9203 by new contributor Jack Zhang)

• Add HTML representation for Epochs in Jupyter Notebooks (#9174 by new contributor Valerii Chirkov)

• Speed up mne.viz.plot_ica_properties() by refactoring (#9174 by new contributor Valerii Chirkov)

• Add apply_function method to epochs and evoked objects (#9088 by new contributor Erica Peterson and Victoria Peterson)

• New tutorial for function mne.make_fixed_length_epochs() (#9156 by new contributor Erica Peterson)

• Add different colors for each volume source space in mne.viz.plot_alignment() (#9043 by new contributor Valerii Chirkov)

• Add overlap parameter to mne.make_fixed_length_epochs() to allow creating overlapping fixed length epochs (#9096 by new contributor Silvia Cotroneo)

• Add mne.Dipole.to_mni() for more convenient dipole.pos to MNI conversion (#9043 by new contributor Valerii Chirkov)

• Update citations in maxwell.py (#9043 by new contributor Valerii Chirkov)

• New Tutorial for analyzing frequency-tagging data (#8867 by new contributor Dominik Welke and kalenkovich)

• Add dbs as new channel type for deep brain stimulation (DBS) recordings (#8739 by new contributor Richard Koehler)

• Add some preprocessing functions to the EEGLAB migration guide (#9169 by new contributor Apoorva Karekal)

• Add mne.chpi.extract_chpi_locs_kit() to read cHPI coil locations from KIT/Yokogawa data (#8813 by new contributor Matt Sanderson, Robert Seymour, and Eric Larson)

• Add match_alias parameter to mne.io.Raw.set_montage() and related functions to match unrecognized channel location names to known aliases (#8799 by new contributor Zhi Zhang)

• Update the notebook 3d backend to use ipyvtk_simple for a better integration within Jupyter (#8503 by Guillaume Favelier)

• Remove the 15-character limitation for channel names when writing to FIF format. If you need the old 15-character names, you can use something like raw.rename_channels({n: n[:13] for n in raw.ch_names}, allow_duplicates=True), by Eric Larson (#8346)

• Add channel-specific annotation support to mne.Annotations via ch_names parameter (#8896 by Eric Larson)

• Add toggle-all button to mne.Report HTML and width argument to mne.Report.add_bem_to_section (#8723 by Eric Larson)

• Add infant template MRI dataset downloader mne.datasets.fetch_infant_template() (#8738 by Eric Larson and Christian O’Reilly)

• Add digitizer information to mne.io.read_raw_egi() (#8789 by Christian Brodbeck)

• Add support for reading some incomplete raw FIF files in mne.io.read_raw_fif() (#9268 by Eric Larson)

• Allow reading digitization from files other than *.fif in the coregistration GUI (#8790 by Christian Brodbeck)

• Speed up mne.inverse_sparse.tf_mixed_norm() using STFT/ISTFT linearity (#8697 by Eric Larson)

• Reduce memory consumption of mne.io.Raw and speed up epoching when thousands of events are present for mne.Epochs (#8801 by Eric Larson)

• Speed up import mne by reducing function creation overhead (#8829 by Eric Larson)

• mne.Report.parse_folder now processes supported non-FIFF files by default, too (#8744 by Richard Höchenberger)

• mne.Report has gained the new methods add_custom_js and add_custom_css for adding user-defined JavaScript and styles (#8762, #9037 by Richard Höchenberger)

• Add option to control appearance of opaque inside surface of the head to mne coreg (#8793 by Eric Larson)

• Add option to disable projection using --projoff in mne browse_raw (#9262 by Eric Larson)

• Add keypress to toggle projection using shift+j in mne.io.Raw.plot() and mne browse_raw (#9262 by Eric Larson)

• Add support for non-FIF files in mne browse_raw using mne.io.read_raw() (#8806 by Eric Larson)

• Add mne.io.read_raw_nedf() for reading StarStim / enobio NEDF files (#8734 by Tristan Stenner)

• Add raw.describe() to display (or return) descriptive statistics for each channel (#8760 by Clemens Brunner)

• Add annotations.to_data_frame() to return annotations as a pandas dataframe (#8783 by Robert Luke)

• Add mne.preprocessing.compute_maxwell_basis() to compute the SSS basis function (#8822 by Eric Larson)

• Add the silhouette parameter to mne.viz.Brain to display sharp edges and improve perception (#8771 by Guillaume Favelier)

• Add warning to mne.cov.compute_whitener() when an explicit rank parameter leads to a large increase in condition number (#8805 by Eric Larson)

• Add parameter align=True to mne.viz.Brain.show_view to make views relative to the closest canonical (MNI) axes rather than the native MRI surface RAS coordinates (#8794 by Eric Larson)

• Add auto_close to mne.Report.add_figs_to_section and mne.Report.add_slider_to_section to manage closing figures (#8730 by Guillaume Favelier)

• Add mne.write_head_bem() to support writing head surface files (#8841 by Yu-Han Luo)

• The signal of resp (respiratory) channels is now assumed to be in the unit Volt (#8858 by Richard Höchenberger)

• Static type checkers like Pylance (comes with VS Code) now display the parameters of many more functions correctly, largely improving overall usability for VS Code users (#8862 by Richard Höchenberger)

• Support new EEGLAB file format (#8874 by Clemens Brunner)

• Reading and writing FIFF files whose filenames end with _meg.fif.gz, _eeg.fif(.gz), and _ieeg.fif(.gz) doesn’t emit a warning anymore; this improves interobaility with BIDS-formatted datasets (#8868 by Richard Höchenberger)

• On macOS, we now set the environment variable QT_MAC_WANTS_LAYER to "1" if it hasn’t been set explicitly by the user, in order to ensure that SourceEstimate plots work on macOS 11 with older versions of Qt and PyQt (#8959 by Richard Höchenberger)

• mne.time_frequency.EpochsTFR.average() now allows different ways of averaging, such as “median”, or callable functions (#8879 by Adam Li)

• Epochs metadata can now be generated automatically from events using mne.epochs.make_metadata (#8834 by Richard Höchenberger)

• Interactions with sliders in mne.Report will now continuously update the linked content (it was updated only on mouse button release before) (#9023 by Richard Höchenberger)

• mne.viz.plot_drop_log and mne.Epochs.plot_drop_log() now omit displaying the subject name in the title if subject=None is passed (#9015 by Richard Höchenberger)

• Plot ECoG tutorial now uses a real epilepsy seizure dataset and visualizes the seizure onset (#9087 by Eric Larson, Adam Li, Alex Rockhill and Liberty Hamilton)

• Improve documentation of Report-Class (#9113 by Martin Schulz)

• Add mne.channels.DigMontage.add_estimated_fiducials() which will add LPA, RPA and Nasion fiducial points to the DigMontage object in mri coordinate frame (#9118 by Adam Li)

• mne.io.anonymize_info() now anonymizes also sex and hand fields when keep_his is False (#9103, #9175 by new contributor Rotem Falach and Richard Höchenberger)

• Add parameter theme to mne.viz.Brain for optional Dark-Mode (#9149 by Martin Schulz, Guillaume Favelier)

• Add first_samp support for raw simulations with mne.simulation.simulate_raw and mne.simulation.SourceSimulator (#9166 by Steven Bierer)

• Evoked gained a baseline attribute that is automatically assembled based on the baseline of the averaged Epochs (#9210 by Richard Höchenberger)

• Add units parameter to mne.io.Raw.get_data() to return data in the desired unit (#9136 by Johann Benerradi and Stefan Appelhoff)

• Add mne.preprocessing.equalize_bads() to interpolate bad channels in a list of Evoked, Epochs or Raw having different sets of bad channels (#9241 by Alex Gramfort)

• mne.Epochs.equalize_event_counts() can now be called without providing a list of event names, and will equalize the counts of all event types present in the Epochs (#9261 by Richard Höchenberger)

• mne.preprocessing.find_eog_events() and mne.preprocessing.create_eog_epochs() now accept a list of channel names, allowing you to specify multiple EOG channels at once (#9269 by Richard Höchenberger)

• Improve performance of mne.set_bipolar_reference() (#9270 by Martin Schulz)

• Add support for setting montages on fNIRS data, with built in standard montages for Artinis OctaMon and Artinis Brite23 devices (#9141 by Johann Benerradi, Robert Luke and Eric Larson)

• Enable support for reading SNIRF files with 2D optode positions (#9347 Robert Luke)

Bugs#

• Fix bug with mne.time_frequency.tfr() where plot misbehaved. Plot now always returns a list of figures instead of a single figure handle (#8013 by new contributor Eduard Ort)

• Fix bug with mne.viz.plot_evoked_topo() where set ylim parameters gets swapped across channel types. (#9207 by new contributor Ram Pari)

• Fix bug with mne.io.read_raw_edf() where µV was not correctly recognized (#9187 by new contributor Sumalyo Datta)

• Fix bug with mne.viz.plot_compare_evokeds() did not check type of combine. (#9151 by new contributor Matteo Anelli)

• Fix bug with mne.viz.plot_evoked_topo() where ylim was only being applied to the first channel in the dataset (#9162 by new contributor Ram Pari )

• Fix bug with mne.Epochs.plot_image() allowing interactive zoom to work properly (#9152 by by new contributor Maggie Clarke and Daniel McCloy)

• Fix bug with mne.Epochs.plot_image() where the x_label was different depending on the evoked parameter (#9115 by new contributor Matteo Anelli)

• Fix bug with restricting mne.io.Raw.save() saving options to .fif and .fif.gz extensions (#9062 by new contributor Valerii Chirkov)

• Fix bug with mne.io.read_raw_kit() where missing marker coils were not handled (#8989 by new contributor Judy D Zhu)

• Fix bug with mne.connectivity.spectral_connectivity where time axis in Epochs data object was dropped. (#8839 by new contributor Anna Padee)

• Fix bug with mne.io.Raw.resample to allow passing stim_picks='misc' (#8844 by new contributor Enrico Varano and Eric Larson)

• Fix bugs with mne.io.read_raw_persyst where multiple Comments with the same name are allowed, and Comments with a “,” character are now allowed (#8311 and #8806 by new contributor Andres Rodriguez and Adam Li)

• Fix zen mode and scalebar toggling for raw.plot() when using the macosx matplotlib backend (#8688 by Daniel McCloy)

• Fix bug with mne.viz.snapshot_brain_montage() where the positions were incorrect (#8983 by Eric Larson)

• Fix bug with mne.preprocessing.maxwell_filter() where the eSSS basis had to exactly match the good channels instead of being a superset (#8675 by Eric Larson)

• Fix bug with mne.Report.add_bem_to_section where n_jobs != 1 would cause n_jobs subsets of MRI images in some orientations to be flipped (#8713 by Eric Larson)

• Fix bug with raw.plot() where annotations didn’t immediately appear when changing window duration (#8689 by Daniel McCloy)

• Fix bug with raw.plot() where scalings='auto' did not compute scalings using the full range of data (#8806 by Eric Larson)

• Fix bug with raw.plot() where setting a lowpass could lead to non-data-channels not plotting (#8954 by Eric Larson)

• Fix bug with mne.io.Raw.load_data() and mne.Epochs.drop_bad() where verbose logging was not handled properly (#8884 by Eric Larson)

• Fix bug with mne.io.read_raw_nicolet() where header type values such as num_sample and duration_in_sec where not parsed properly (#8712 by Alex Gramfort)

• Fix bug with mne.preprocessing.read_ica_eeglab() when reading decompositions using PCA dimensionality reduction (#8780 by Alex Gramfort and Eric Larson)

• Fix bug with mne.minimum_norm.make_inverse_operator() where depth was errantly restricted to be less than or equal to 1. (#8804 by Eric Larson)

• Fix bug with mne.stats.permutation_cluster_1samp_test() and related clustering functions when adjacency=None and out_type='indices' (##8842 by Eric Larson)

• Fix bug with mne.viz.plot_alignment() where plotting a sphere model could ignore the brain argument (#8857 by Eric Larson)

• Fix bug with mne.SourceEstimate.plot() where flatmaps were not positioned properly when using hemi='both' (#9315 by Eric Larson)

• Fix bug with mne.Annotations.save() where files could be overwritten accidentally, it can now be controlled via the overwrite argument (#8896 by Eric Larson)

• Fix bug with replace argument of mne.Report.add_bem_to_section and mne.Report.add_slider_to_section (#8723 by Eric Larson)

• Fix bug with mne.chpi.compute_chpi_locs() where all cHPI coils being off would lead to an empty array of the wrong dimensionality (#8956 by Eric Larson)

• Fix bug with mne.extract_label_time_course() where labels, STCs, and the source space were not checked for compatible subject attributes (#9284 by Eric Larson)

• Fix bug with mne.grow_labels() where overlap=False could run forever or raise an error (#9317 by Eric Larson)

• Fix compatibility bugs with mne_realtime (#8845 by Eric Larson)

• Fix bug with mne.viz.Brain where non-inflated surfaces had an X-offset imposed by default (#8794 by Eric Larson)

• Fix bug with mne coreg where nasion values were not updated when clicking (#8793 by Eric Larson)

• Fix bug with matplotlib-based 3D plotting where Axes3D were not properly initialized in mne.viz.plot_source_estimates() (#8811 by Chris Bailey)

• Allow sEEG channel types in mne.Evoked.plot_joint() (#8736 by Daniel McCloy)

• Fix bug where hidden annotations could be deleted interactively in mne.io.Raw.plot() windows (#8831 by Daniel McCloy)

• Function mne.set_bipolar_reference() was not working when passing Epochs constructed with some picks (#8728 by Alex Gramfort)

• Fix anonymization issue of FIF files after IO round trip (#8731 by Alex Gramfort)

• Fix bug in mne.preprocessing.ICA.plot_sources where right-clicking component names could yield plot_properties windows for the wrong component if picks had been specified (#8996 by Daniel McCloy)

• Fix title not shown in mne.viz.plot_montage() (#8752 by Clemens Brunner)

• mne.io.read_raw_egi now correctly handles pathlib.Path filenames (#8759 by Richard Höchenberger)

• mne.viz.plot_evoked and mne.Evoked.plot now correctly plot global field power (GFP) for EEG data when gfp=True or gfp='only' is passed (used to plot RMS). For MEG data, we continue to plot the RMS, but now label it correctly as such (#8775 by Richard Höchenberger)

• Fix bug with mne make_scalp_surfaces where --overwrite was not functional (#8800 by Yu-Han Luo)

• Fix bug with mne.viz.plot_topomap() when plotting gradiometers with a missing channel in a pair (#8817 by Alex Gramfort)

• epochs.crop() now also adjusts the reject_tmin and reject_tmax attributes if necessary (#8821 by Richard Höchenberger)

• When creating Epochs, we now ensure that reject_tmin and reject_tmax cannot fall outside of the epochs’ time interval anymore (#8821 by Richard Höchenberger)

• read_raw_bti erroneously treated response channels as respiratory channels (#8856 by Richard Höchenberger)

• The RMS trace shown in the time viewer of SourceEstimate plots is now correctly labeled as RMS (was GFP before) (#8965 by Richard Höchenberger)

• Fix bug with mne.SourceEstimate.plot() and related functions where the scalars were not interactively updated properly (#8985 by Eric Larson)

• Fix bug with mne.channels.find_ch_adjacency() returning wrong adjacency for Neuromag122-Data (#8891 by Martin Schulz)

• Fix mne.read_dipole() yielding mne.Dipole objects that could not be indexed (#8963 by Marijn van Vliet)

• Fix bug when setting n_jobs > 1 in mne.Report.parse_folder() (#9109 by Martin Schulz)

• Fix bug with mne.Evoked.plot_image() where an incorrect clim parameter did not raise any error (#9115 by new contributor Matteo Anelli)

• Fix bug with mne.io.Raw.pick where incorrect fnirs types were returned (#9178 by Robert Luke)

• Fix bug when passing both axes and picks to mne.viz.plot_compare_evokeds (#9252 by Daniel McCloy)

• Improved string representation of Epochs containing multiple event types; improved (and more mathematically correct) evoked.comment in the mne.combine_evoked output; and better (and often more concise) legend labels in the figures created via plot_compare_evokeds (#9027 by Richard Höchenberger)

• mne.preprocessing.find_ecg_events() now correctly handles situation where no ECG activity could be detected, and correctly returns an empty array of ECG events (#9236 by Richard Höchenberger)

• Fix bug with picks attribute for Epochs after calling mne.Epochs.add_channels() (#9246 by Alex Gramfort)

• Fix bug where backend='notebook' could not be used in mne.SourceEstimate.plot() (#9305 by Jean-Remi King)

• mne.preprocessing.compute_proj_eog and mne.preprocessing.compute_proj_ecg now return empty lists if no EOG or ECG events, respectively, could be found. Previously, we’d return None in these situations, which does not match the documented behavior of returning a list of projectors (#9277 by Richard Höchenberger)

API changes#

• Introduced new 'auto' settings for ICA.max_iter. The old default max_iter=200 will be removed in MNE-Python 0.24 (#9099 by new contributor Cora Kim)

• mne.viz.plot_sensors_connectivity now allows setting the colorbar label via the cbar_label parameter (#9248 by Daniel McCloy)

• mne.read_selection has been deprecated in favor of mne.read_vectorview_selection. mne.read_selection will be removed in MNE-Python 0.24 (#8870 by Richard Höchenberger)

• mne.beamformer.tf_dics has been deprecated and will be removed in MNE-Python 0.24 (#9122 by Britta Westner)

• Fitting ICA on baseline-corrected Epochs, and / or applying it on baseline-corrected Epochs or Evoked data will now display a warning. Users are advised to only baseline correct their data after cleaning is completed (#9033 by Richard Höchenberger)

• Supplying multiple channel names to mne.preprocessing.find_eog_events or mne.preprocessing.compute_proj_eog as a string of comma-separated channel names has been deprecated; please pass a list of channel names instead. Support for comma-separated strings will be removed in MNE-Python 0.24 (#9269 by Richard Höchenberger)

• The default in mne.beamformer.make_dics() of real_filter=False will change to real_filter=True in 0.24 (#9340 by Britta Westner)

Authors#

People who contributed to this release in alphabetical order (people with a + are first time contributors):

• Adam Li

• Alex Rockhill

• Alexandre Gramfort

• Ana Radanovic

• Andres Rodriguez+

• Anna Padee+

• Apoorva Karekal+

• Britta Westner

• Catalina Magalvan

• Christian Brodbeck

• Christian Mista

• Christian O’Reilly

• Christina Zhao

• Christopher J. Bailey

• Clemens Brunner

• Cora Kim+

• Daniel McCloy

• Denis A. Engemann

• Dominik Welke+

• Eduard Ort

• Enrico Varano+

• Eric Larson

• Erica Peterson+

• Evgenii Kalenkovich

• Felix Klotzsche+

• Giorgio Marinato

• Guillaume Favelier

• Jack Zhang+

• Jean-Remi King

• Johann Benerradi

• Joris Van den Bossche

• Judy D Zhu+

• Liberty Hamilton

• Luke Bloy

• Maggie Clarke+

• Mainak Jas

• Manorama Kadwani

• Marijn van Vliet

• Martin Schulz

• Matt Sanderson+

• Matteo Anelli+

• Nicolas Gensollen

• Ram Pari+

• Richard Höchenberger

• Richard Koehler+

• Robert Luke

• Rotem Falach+

• Sebastien Treguer

• Silvia Cotroneo+

• Stefan Appelhoff

• Steven Bierer

• Sumalyo Datta+

• Timon Merk

• Tristan Stenner

• Valerii Chirkov+

• Victoria Peterson

• Yu-Han Luo

• Zhi Zhang+

Enhancements#

• Add mne.read_evokeds_mff() to read averaged MFFs (requires mffpy >= 0.5.7) by new contributor Evan Hathaway (#8354)

• Add mne.decoding.SSD for spatial filtering with spatio-spectral-decomposition (#7070 by new contributor Victoria Peterson and Denis Engemann)

• Add reader for optical imaging data recorded using ISS Imgagent I/II hardware and BOXY recording software in mne.io.read_raw_boxy() (#7717 by new contributor Jonathan Kuziek and Kyle Mathewson)

• Add options to use labels in mne.minimum_norm.get_point_spread() and mne.minimum_norm.get_cross_talk() (#8275 by Olaf Hauk)

• Update surfaces argument in mne.viz.plot_alignment() to allow dict for transparency values, and set default for sEEG data to have transparency (#8445 by Keith Doelling)

• Add support for mri_fiducials='estimated' in mne.viz.plot_alignment() to allow estimating MRI fiducial locations using mne.coreg.get_mni_fiducials() (#8553 by Eric Larson)

• Update default values in mne coreg and mne.viz.plot_alignment() for clearer representation of MRI and digitized fiducial points (#8553 by Alex Gramfort and Eric Larson)

• Add n_pca_components argument to mne.viz.plot_ica_overlay() (#8351 by Eric Larson)

• Add mne.stc_near_sensors() to facilitate plotting ECoG data (#8190 by Eric Larson)

• Add proj argument to mne.make_fixed_length_epochs() (#8351 by Eric Larson)

• Add mne.preprocessing.realign_raw() to realign simultaneous raw recordings in the presence of clock drift (#8539 by Eric Larson)

• Reduce memory usage of volume source spaces (#8379 by Eric Larson)

• Speed up heavy use of mne.SourceMorph.apply() for volumetric source spaces by use of the method mne.SourceMorph.compute_vol_morph_mat() (#8366 by Eric Larson)

• Add support for non-uniform zooms (e.g., when using a surrogate MRI via mne.scale_mri()) in volumetric morphing (#8642 by Eric Larson)

• In mne.compute_source_morph() zooms are no longer required to match the spacing of src_to, which is useful to ensure the morphing is accurate when the src_to spacing is large (e.g., 1 cm) (#8642 by Eric Larson)

• Add volumetric source space support to mne.labels_to_stc() (#8447 by Eric Larson)

• Speed up mne.decoding.TimeDelayingRidge with edge correction using Numba (#8323 by Eric Larson)

• Add mne.Epochs.reset_drop_log_selection() to facilitate writing epochs with many ignored entries in their drop log (#8449 by Eric Larson)

• Add sEEG source visualization using mne.stc_near_sensors() and sEEG working tutorial (#8402 by Eric Larson and Adam Li)

• Add mne.channels.DigMontage.get_positions(), which will return a dictionary of channel positions, coordinate frame and fiducial locations (#8460 by Adam Li)

• Add support for writing digitization points in a coordinate frame other than head in mne.channels.DigMontage.save() (#8532 by Eric Larson)

• Add picks parameter to mne.preprocessing.fix_stim_artifact() to specify which channel needs to be fixed (#8482 by Alex Gramfort)

• Add progress bar support to mne.time_frequency.csd_morlet() (#8608 by Eric Larson)

• Further improved documentation building instructions and execution on Windows (#8502 by kalenkovich and Eric Larson)

• Add option to disable TQDM entirely with MNE_TQDM='off' (#8515 by Eric Larson)

• Add option on_header_missing to mne.channels.read_polhemus_fastscan() (#8622 by Eric Larson)

• Add option window to mne.time_frequency.psd_welch() and related functions (#8862 by Eric Larson)

• mne.preprocessing.ICA.plot_sources now displays an mne.preprocessing.ICA.plot_properties window when right-clicking on component names on the y-axis (#8381 by Daniel McCloy)

• mne.io.read_raw_edf(), mne.io.read_raw_bdf(), and mne.io.read_raw_gdf() now detect and handle invalid highpass/lowpass filter settings (#8584 by Clemens Brunner)

• If a baseline tuple containing one or two None values – e.g. (None, 0), (0, None), or (None, None) – is passed to Epochs or apply_baseline, the None value(s) will be replaced with the actual time (i.e., tmin and tmax, respectively) when populating Epochs.baseline (#8442 by Richard Höchenberger)

• Epochs will now retain the information about an applied baseline correction, even if the baseline period is partially or completely removed through cropping later on (#8442 by Richard Höchenberger)

• Add mne.source_space.compute_distance_to_sensors() to compute distances between vertices and sensors (#8534 by Olaf Hauk and Marijn van Vliet)

• Annotations can now be shown/hidden interactively in raw plots (#8624 by Daniel McCloy)

• Added argument colors to mne.grow_labels (#8519 by Olaf Hauk)

• Added mne.SourceEstimate.apply_baseline method for baseline-correction of source estimates (#8452 by Olaf Hauk)

• New mne.viz.Brain.set_time method to set the displayed time in seconds (#8415 by Daniel McCloy)

• Update the backend parameter of mne.viz.plot_source_estimates() to integrate pyvista (#8395 by Guillaume Favelier)

• Add group_by parameter to mne.viz.plot_epochs and mne.Epochs.plot to allow displaying channel data by sensor position (#8381 by Daniel McCloy)

Bugs#

• Fix orthogonalization of power envelopes in mne.connectivity.envelope_correlation (#8658 by new contributor Qianliang Li and Eric Larson)

• Fix data overwrite of cascading simulation operations :mne.simulation.simulate_raw (#8633 by new contributor Jeff Stout)

• Fix a transpose issue of mne.decoding.CSP.plot_filters() (#8580 by new contributor Hongjiang Ye)

• Fix mne.io.read_raw_curry() to deal with Curry datasets that have channels that are listed in the labels file, but which are absent from the saved data file (e.g. ‘Ref’ channel). Also now populates info[‘meas_date’] if possible (#8400 by new contributor Tod Flak)

• Fix bug with mne.io.egi.tests/test_egi.py where it mandatorily downloaded testing data when it was not necessary (#8474 by new contributor Aniket Pradhan)

• Fix bug with reading split files that have dashes in the filename (#8339 by new contributor Eduard Ort)

• Fix bug with parsing EDF dates and date integers (#8558 by new contributor Austin Hurst and Eric Larson)

• Fix bug with reading EDF and KIT files on big endian architectures such as s390x (#8618 by Eric Larson)

• Fix bug with mne.beamformer.make_dics() where the rank parameter was not properly handled (#8594 by Marijn van Vliet and Eric Larson)

• Fix bug with mne.beamformer.apply_dics() where the whitener was not properly applied (#8610 by Eric Larson)

• Fix bug with plot_epochs_image when order is supplied and multiple conditions are plotted (#8377 by Daniel McCloy )

• Fix bug with mne.viz.plot_source_estimates() when using the PyVista backend where singleton time points were not handled properly (#8285 by Eric Larson)

• Fix bug when passing axes to plotting functions, matplotlib.pyplot.tight_layout() will not be called when the figure was created using a constrained layout (#8344 by Eric Larson)

• Fix bug with compensated CTF data when picking channels without preload (#8318 by Eric Larson)

• Fix bug with plotting MEG topographies where the wrong extrapolation made was used in ICA (#8637 by Eric Larson)

• Fix bug when merging fNIRS channels in mne.viz.plot_evoked_topomap() and related functions (#8306 by Robert Luke)

• Fix bug where events could overflow when writing to FIF (#8448 by Eric Larson)

• mne.io.read_raw_edf() now supports EDF files with invalid recording dates (#8283 by Clemens Brunner)

• Fix bug with mne.io.Raw.save() when using split_naming='bids' where non-split files would still be named name_split-01_meg.fif instead of the requested name_meg.fif (#8464 by Alex Gramfort and Eric Larson)

• Fix bug with mne.preprocessing.ICA where n_pca_components as a float would give the number of components that explained less than or equal to the given variance. It now gives greater than the given number for better usability and consistency with sklearn.decomposition.PCA. Generally this will mean that one more component will be included (#8326 by Eric Larson)

• Fix bug with mne.preprocessing.ICA where projections were not tracked properly (#8343 by Eric Larson)

• Fix bug where extrapolation points created artifacts in mne.viz.plot_evoked_topomap() and related functions (#8425 by Mikołaj Magnuski)

• Fix bug with mne.preprocessing.read_ica_eeglab() where full-rank data were not handled properly (#8326 by Eric Larson)

• Fix bug with Somatosensory where the BEM was not included (#8317 by Eric Larson)

• Fix bug with coordinate frames when performing volumetric morphs via mne.compute_source_morph() and mne.SourceMorph.apply() that could lead to ~5 mm bias (#8642 by Eric Larson)

• Fix bug with volumetric rendering alpha in mne.VolSourceEstimate.plot_3d() and related functions (#8663 by Eric Larson)

• Fix missing documentation of mne.io.read_raw_nihon() in Importing data from EEG devices (#8320 by Adam Li)

• Fix bug with mne.add_reference_channels() when mne.io.Raw.reorder_channels() or related methods are used afterward (#8303, ##8484 by Eric Larson)

• Fix bug where the verbose arguments to mne.Evoked.apply_baseline() and mne.Epochs.apply_baseline() were not keyword-only (#8349 by Eric Larson)

• ICA.max_pca_components will not be altered by calling fit anymore. Instead, the new attribute ICA.max_pca_components_ will be set (#8321 by Richard Höchenberger)

• Fix bug that plot_ica_overlay would sometimes not create red traces (#8341 by Richard Höchenberger)

• Fix bug with ICA where n_components=None, n_pca_components=None could lead to unstable unmixing matrix inversion by making n_components=None also use the lesser of n_components=0.999999 and n_components=n_pca_components (#8351 by Eric Larson)

• The ica.n_pca_components property is no longer be updated during mne.preprocessing.ICA.fit(), instead ica.n_components_ will be added to the instance (#8351 by Eric Larson)

• Pass rank everyhwere in forward preparation for source imaging. This bug affected sparse solvers when using maxfilter data (#8368 by Alex Gramfort)

• Fix bug in mne.viz.plot_alignment() where ECoG and sEEG channels were not plotted and fNIRS channels were always plotted in the head coordinate frame (#8393 by Eric Larson)

• Fix bug in mne.set_bipolar_reference() where ch_info could contain invalid channel information keys (#8416 by Eric Larson)

• When reading BrainVision raw data, the channel units and types were sometimes not inferred correctly (#8434 by Richard Höchenberger)

• Attempting to remove baseline correction from preloaded Epochs will now raise an exception (#8435 by Richard Höchenberger)

• mne.Report.parse_folder() will now correctly handle split FIFF files (#8486, #8491 by Richard Höchenberger)

• Fix bug where BrainVision channel names, event types, and event descriptions containing commas were incorrectly parsed (#8492 by Stefan Appelhoff)

• Fix bug in mne.preprocessing.compute_fine_calibration() where the magnetometer calibration coefficients were computed incorrectly (#8522 by Eric Larson)

• Fix bug in mne.io.read_raw_eeglab() where empty event durations led to an error (#8384 by Mikołaj Magnuski)

• Fix inset sensor plots to always use equal aspect (#8545 by Daniel McCloy)

• Fix bug in mne.viz.plot_compare_evokeds where evokeds with identical comment attributes would not plot properly if passed as a list (#8590 by Daniel McCloy)

• Fix bug in mne.time_frequency.psd_welch() and related functions where the window default errantly changed from 'hamming' to ('tukey', 0.25) (#8862 by Eric Larson)

• Fix bug in mne.io.read_raw_kit() where scale factors for EEG channels could be set to zero (#8542 by Eric Larson)

• Fix reading GDF files with excluded channels in mne.io.read_raw_gdf() (#8520 by Clemens Brunner)

• Fix automatic selection of extrapolation mask type from channel type when plotting field maps (#8589 by Daniel McCloy)

• Fix bug in mne.viz.set_3d_title() where 3D plot could have multiple titles that overlap (#8564 by Guillaume Favelier)

• Fix bug in mne.viz.set_3d_view() where plotter is not updated properly causing camera issues in the doc (#8564 by Guillaume Favelier)

• mne.preprocessing.find_ecg_events() didn’t take the tstart parameter value into account when calculating the average heart rate (#8605 by Richard Höchenberger)

API changes#

• Minimum required versions were increased for core dependencies NumPy (1.15.4), SciPy (1.1.0), and Matplotlib (3.0) and for the optional dependencies scikit-learn (0.20.2) and pandas (0.23.4) (#8374 by Eric Larson)

• The parameter on_split_missing has been added to mne.io.read_raw_fif() and its default will change from 'warn' to 'raise' in 0.23, by (#8357 Eric Larson)

• The max_pca_components argument of ICA has been deprecated, use n_components during initialization and n_pca_components in apply() instead (#8351 by Eric Larson)

• The n_pca_components argument of ICA has been deprecated, use n_pca_components in apply() (#8356 by Eric Larson)

• The trans argument of mne.extract_label_time_course() is deprecated and will be removed in 0.23 as it is no longer necessary (#8389 by Eric Larson)

• Parameter event_colors in mne.viz.plot_epochs and mne.Epochs.plot is deprecated, replaced by event_color which is consistent with mne.viz.plot_raw and provides greater flexibility (#8381 by Daniel McCloy)

Authors#

People who contributed to this release in alphabetical order (people with a + are first time contributors):

• Adam Li

• Alexandre Gramfort

• Aniket Pradhan +

• Austin Hurst +

• Christian Brodbeck

• Clemens Brunner

• Daniel McCloy

• Denis A. Engemann

• Eduard Ort +

• Eric Larson

• Evan Hathaway +

• Evgenii Kalenkovich

• Fede Raimondo

• Guillaume Favelier

• Hongjiang Ye +

• Jean-Remi King

• Jeff Stout +

• Jonathan Kuziek +

• Jussi Nurminen

• Justus Schwabedal

• Keith Doelling

• Kyle Mathewson

• Mads Jensen

• Mainak Jas

• Marijn van Vliet

• Mikolaj Magnuski

• Olaf Hauk

• Qianliang Li +

• Richard Höchenberger

• Robert Luke

• Stefan Appelhoff

• Thomas Hartmann

• Tod Flak +

• Victoria Peterson +

Enhancements#

• Modified mne.Epochs.pick_types() to remove dropped channel types from reject and flat dictionaries by new contributor Rahul Nadkarni

• Add function mne.preprocessing.regress_artifact() to remove artifacts using linear regression by new contributor Kyle Mathewson and Eric Larson

• Add sources and detectors options for fNIRS use of mne.viz.plot_alignment() allowing plotting of optode locations in addition to channel midpoint channels and path between fNIRS optodes by new contributor Kyle Mathewson

• BrainVision data format files are now parsed for EEG impedance values in mne.io.read_raw_brainvision() and provided as a .impedances attribute of raw by new contributor Jan Sedivy and Stefan Appelhoff

• Add function mne.channels.combine_channels() to combine channels from Raw, Epochs, or Evoked according to ROIs (combinations including mean, median, or standard deviation; can also use a callable) by new contributor Johann Benerradi

• Improved documentation building instructions and execution on Windows by new contributor Martin Schulz, kalenkovich, and Eric Larson

• Speed up reading of annotations in EDF+ files by new contributor Jeroen Van Der Donckt

• Add head to mri and mri to voxel space transform details to Source alignment and coordinate frames tutorial, by Alex Rockhill

• Improve memory efficiency of mne.concatenate_epochs() by Eric Larson

• Allow picking channels in raw instances (e.g., mne.io.Raw.pick_types()) without preloading data, by Eric Larson

• mne.preprocessing.ICA.plot_sources() now plots annotation markers similar to mne.io.Raw.plot() by Luke Bloy

• Add support for scalar multiplication and division of mne.time_frequency.AverageTFR instances by Luke Bloy

• Add support for signals in mV for mne.io.read_raw_brainvision() by Clemens Brunner

• mne.Epochs.plot_psd_topomap() and mne.viz.plot_epochs_psd_topomap() now allow joint colorbar limits across subplots, by Daniel McCloy.

• Add support for using loose-orientation surface priors with free-orientation volume priors in mixed source space inverses by Eric Larson

• Add mne.MixedVectorSourceEstimate for vector source estimates for mixed source spaces, by Eric Larson

• Add mixed and volumetric source estimate plotting using volumetric ray-casting to mne.MixedSourceEstimate.plot() and mne.VolSourceEstimate.plot_3d() by Eric Larson

• Add mne.MixedSourceEstimate.surface() and mne.MixedSourceEstimate.volume() methods to allow surface and volume extraction by Eric Larson

• Add mne.VectorSourceEstimate.project() to project vector source estimates onto the direction of maximum source power by Eric Larson

• Add support to mne.extract_label_time_course() for vector-valued and volumetric source estimates by Eric Larson

• Add method mne.VolSourceEstimate.in_label() by Eric Larson

• Add support for mixed source spaces to mne.compute_source_morph() by Eric Larson

• Add support for volume and mixed source spaces to mne.stats.summarize_clusters_stc() by Eric Larson

• Add support for omitting the SDR step in volumetric morphing by passing n_iter_sdr=() to mne.compute_source_morph by Eric Larson

• Add single_volume argument to mne.setup_volume_source_space() to facilitate creating source spaces with many volumes (e.g., all subvolumes of aseg.mgz) by Eric Larson

• Add support for passing a string argument to bg_img in mne.viz.plot_volume_source_estimates by Eric Larson

• Add support for providing the destination surface source space in the src_to argument of mne.compute_source_morph() by Eric Larson

• Add explained variance logging to mne.inverse_sparse.mixed_norm(), mne.inverse_sparse.gamma_map(), and mne.beamformer.rap_music() by Eric Larson

• Add tol_kind option to mne.compute_rank() by Eric Larson

• Allow resampling raw data with mne.io.Raw.resample() without preloading data, by Eric Larson

• Allow using pick_ori='vector' with a fixed-orientation inverse to facilitate visualization with mne.viz.plot_vector_source_estimates() by Eric Larson

• mne.viz.plot_dipole_locations() and mne.Dipole.plot_locations() gained a title argument to specify a custom figure title in orthoview mode by Richard Höchenberger

• Add orientations projected onto the MRI slices in mne.viz.plot_dipole_locations() in 'orthoview' mode by Eric Larson

• Add temporal derivative distribution repair (TDDR) mne.preprocessing.nirs.temporal_derivative_distribution_repair() with shortened alias mne.preprocessing.nirs.tddr by Robert Luke

• Add mne.read_freesurfer_lut() to make it easier to work with volume atlases by Eric Larson

• Add support for overlap-add processing when method='spectrum_fit' in mne.io.Raw.notch_filter() by Eric Larson

• Add functionality to interpolate bad NIRS channels by Robert Luke

• Add ability to interpolate EEG channels using minimum-norm projection in mne.io.Raw.interpolate_bads() and related functions with method=dict(eeg='MNE') by Eric Larson

• Added ability of mne.io.read_raw_nirx() to open data by passing path to header file Robert Luke

• Add mne.channels.DigMontage.rename_channels() to allow renaming montage channels by Eric Larson

• Document mne.channels.compute_native_head_t() to allow computing the native-to-head transformation that will be applied when doing raw.set_montage and related functions by Eric Larson

• Add support to in mne.io.Raw.plot() for passing clipping as a float to clip to a proportion of the dedicated channel range by Eric Larson

• Add function mne.preprocessing.annotate_muscle_zscore() to annotate periods with muscle artifacts. by Adonay Nunes

• Add mne.stats.ttest_ind_no_p() to mirror mne.stats.ttest_1samp_no_p() with hat correction by Eric Larson

• Add mne.stats.combine_adjacency to combine adjacency matrices for independent data dimensions to enable things like spatio-temporal-time-frequency clustering in mne.stats.permutation_cluster_test by Eric Larson

• Speed up raw data reading without preload in mne.io.read_raw_nirx() by Eric Larson

• Speed up mne.Epochs.copy() and mne.Epochs.__getitem__() by avoiding copying immutable attributes by Eric Larson

• Speed up and reduce memory usage of mne.SourceEstimate.plot() and related functions/methods when show_traces=True by Eric Larson

• Reduce memory usage of plot_psd, psd_welch, and psd_array_welch for long segments of data by Eric Larson

• Support for saving movies of source time courses (STCs) with brain.save_movie method and from graphical user interface by Guillaume Favelier

• Add mri and show_orientation arguments to mne.viz.plot_bem() by Eric Larson

• Add mne.viz.centers_to_edges() to help when using matplotlib.axes.Axes.pcolormesh() with flat shading by Eric Larson

• Add on_missing='raise' to mne.io.Raw.set_montage() and related functions to allow ignoring of missing electrode coordinates by Adam Li

• Add on_missing='raise' to mne.viz.plot_events() to allow ignoring missing events when passing event_id by Eric Larson

• Add REST EEG infinity reference scheme to mne.io.Raw.set_eeg_reference() and related functions by Eric Larson

• Add better sanity checking of max_pca_components and n_components to provide more informative error messages for mne.preprocessing.ICA by Eric Larson

• Add plot option to mne.viz.plot_filter() allowing selection of which filter properties are plotted and added option for user to supply axes by Robert Luke

• Add estimation method legend to mne.viz.plot_snr_estimate() by Eric Larson

• Add support to mne.SourceSpaces.export_volume for mri_resolution='sparse' to color only the nearest-neighbor voxels instead of entire regions by Eric Larson

• Add axes argument to mne.viz.plot_evoked_white(), mne.Evoked.plot_white(), and mne.viz.plot_snr_estimate() by Eric Larson

• Change default to surfaces='auto' from surfaces='head' to allow mne.viz.plot_alignment() to work when just passing a mne.Info as plot_alignment(info) by Eric Larson

• Add ECoG misc EDF dataset to the Working with ECoG data tutorial to show snapshots of time-frequency activity by Adam Li

• Add mne.viz.set_3d_options() and MNE_3D_OPTION_ANTIALIAS environment variable to control full-scene antialiasing (FXAA) in 3D functions like mne coreg, mne.viz.plot_alignment(), and mne.viz.plot_source_estimates(); this can be useful on systems where FXAA does not work well, such as some with MESA software rendering, by Eric Larson

• Add config variable MNE_TQDM to support explicitly setting the TQDM backend to use, for example tqdm or tqdm.auto (default) by Eric Larson

• Add better support for reading corrupted FIF files in mne.io.read_raw_fif() by Eric Larson

• BIDS conformity: When saving FIF files to disk and the files are split into parts, the split_naming='bids' parameter now uses a “_split-%d” naming instead of the previous “_part-%d”, by Stefan Appelhoff

• Add support for whitening and multiple channel types in mne.beamformer.make_dics() by Marijn van Vliet

• Add better error message when trying to save incompatible Evoked objects to the same file by Eric Larson

• Add support for loading complex numbers from mat files by Thomas Hartmann

• Add generic reader function mne.io.read_raw() that loads files based on their extensions (it wraps the underlying specific read_raw_xxx functions) by Clemens Brunner

• Add automatic T3 magnetometer detection and application of mne.io.Raw.fix_mag_coil_types() to mne.preprocessing.maxwell_filter() by Eric Larson

• Add extended SSS (eSSS) support to mne.preprocessing.maxwell_filter() by Eric Larson

• Add 'auto' option to mne.preprocessing.ICA.find_bads_ecg() to automatically determine the threshold for CTPS method by Yu-Han Luo

• Add a notebook 3d backend for visualization in jupyter notebook with mne.viz.set_3d_backend() by Guillaume Favelier

• Add support for reading and writing surfaces in Wavefront .obj format to the mne.read_surface() and mne.write_surface() by Marijn van Vliet

• Add tutorial on how to manually fix BEM meshes in Blender by Marijn van Vliet and Ezequiel Mikulan

• mne.write_evokeds() will now accept mne.Evoked objects with differing channel orders in info['bads'], which would previously raise an exception by Richard Höchenberger

• Add support for reading channel names (e.g., for Ricoh data) to mne.io.read_raw_kit() by Alex Gramfort and Eric Larson

• The reject_tmin and reject_tmax parameters of mne.Epochs are now taken into account when using the reject_by_annotation parameter by Stefan Appelhoff

• mne.preprocessing.find_bad_channels_maxwell() now automatically applies a low-pass filter before running bad channel detection. This can be disabled, restoring previous behavior by Richard Höchenberger

• Add proj='reconstruct' to mne.Evoked.plot() and related functions to apply projectors and then undo the signal bias using field mapping by Eric Larson

• Add writing BEM surfaces and solutions in H5 format in mne.write_bem_surfaces() and mne.write_bem_solution() by Eric Larson

• When picking a subset of channels, or when dropping channels from Raw, Epochs, or Evoked, projectors that can only be applied to the removed channels will now be dropped automatically by Richard Höchenberger

• mne.Report now can add topomaps of SSP projectors to the generated report. This behavior can be toggled via the new projs argument by Richard Höchenberger

• When passing a list of Evoked objects to plot_compare_evokeds, each evoked’s .comment attribute will be used to label the trace. If .comment is empty, a 1-based index is assigned as the label by Richard Höchenberger

• Speed up mne.stats.summarize_clusters_stc() using Numba by Yu-Han Luo

• Add reject_by_annotation=True to mne.make_fixed_length_epochs() and mne.preprocessing.ICA.plot_properties() to reject bad data segments based on annotation by Yu-Han Luo

• parse_folder now accepts a path-like folder name (it used to work with strings only) by Alex Gramfort

• Add component_order parameter to mne.decoding.CSP which allows switching between mutual_info (default) and alternate (a simpler and frequently used option) by Martin Billinger and Clemens Brunner

• Add memory size information to the repr of mne.SourceSpaces and mne.SourceEstimate and related classes by Eric Larson

• Add reader for Persyst (.lay + .dat format) data in mne.io.read_raw_persyst() by Adam Li

• Use PyVista as the default backend for 3D visualization instead of Mayavi by Guillaume Favelier

• Add reader for SNIRF NIRS data in mne.io.read_raw_snirf() by Robert Luke

• Evoked has gained tmin and tmax attributes for more consistency with Epochs by Richard Höchenberger

Bugs#

• Fix bug for writing and reading complex evoked data modifying mne.write_evokeds() and mne.read_evokeds() by new contributor Lau Møller Andersen

• Fix bug by adding error message when trying to save complex stc data in a non.-h5 format mne.VolSourceEstimate.save() by new contributor Lau Møller Andersen

• Fix bug with logging in mne.io.Raw.set_eeg_reference() and related functions by new contributor Simeon Wong

• Fix bug with mne.io.read_raw_gdf() where birthdays were not parsed properly, leading to an error by new contributor Svea Marie Meyer

• Fix bug with mne.io.read_raw_edf() where recording ID was not read properly for non-ASCII characters by new contributor Lx37

• Fix bug in mne.set_eeg_reference() and related functions to set info['custom_ref_applied'] to True for ‘ecog’ and ‘seeg’ channels in addition to ‘eeg’ by new contributor Liberty Hamilton

• Fix bug with mne.chpi.compute_chpi_amplitudes() and mne.chpi.filter_chpi() to optimize time window length by new contributor Steven Bierer

• Fix bug with mne.preprocessing.ICA.find_bads_eog() when more than one EOG components are present by Christian O’Reilly

• Fix bug with non-preloaded data when using raw.apply_proj().load_data().get_data() where projectors were not applied by Eric Larson

• Fix bug to permit stc.project('nn', src) to be applied after stc was restricted to an mne.Label by Luke Bloy

• Fix bug with mne.io.Raw.set_meas_date() to support setting meas_date to None, by Luke Bloy

• Fix bug with mne.setup_volume_source_space() when volume_label was supplied where voxels slightly (in a worst case, about 37% times pos in distance) outside the voxel-grid-based bounds of regions were errantly included, by Eric Larson

• Fix bug with mne coreg where reverse scalings were applied to src[0]['src_mri_t'] for volume source spaces, so morphing and plotting did not work correctly by Eric Larson

• Fix bug with mne.io.read_raw_ctf() when reference magnetometers have the compensation grade marked by Eric Larson

• Fix bug with mne.SourceSpaces.export_volume with use_lut=False where no values were written by Eric Larson

• Fix bug with mne.preprocessing.annotate_movement() where bad data segments, specified in raw.annotations, would be handled incorrectly by Luke Bloy

• Fix bug with mne.compute_source_morph() when more than one volume source space was present (e.g., when using labels) where only the first label would be interpolated when mri_resolution=True by Eric Larson

• Fix bug with mne.compute_source_morph() when morphing to a volume source space when src_to is used and the destination subject is not fsaverage by Eric Larson

• Fix bug with mne.compute_source_morph() where outermost voxels in the destination source space could be errantly omitted by Eric Larson

• Fix bug with mne.compute_source_morph() where complex data was cast to real when doing a volumetric morph by Eric Larson

• Fix bug with mne.minimum_norm.compute_source_psd_epochs() and mne.minimum_norm.source_band_induced_power() raised errors when method='eLORETA' by Eric Larson

• Fix bug with mne.minimum_norm.apply_inverse() where the explained variance did not work for complex data by Eric Larson

• Fix bug with mne.preprocessing.compute_current_source_density() where values were not properly computed; maps should now be more focal, by Alex Rockhill and Eric Larson

• Fix bug with mne.combine_evoked() where equal-weighted averages were wrongly computed as equal-weighted sums, by Daniel McCloy

• Fix bug with mne.channels.DigMontage.plot() when reference channels with no defined position are present, by Eric Larson

• Fix bug with mne.channels.read_custom_montage() where "headshape" points in SFP format were not read properly, by Eric Larson

• Fix bug with setting HTML classes when using mne.Report.add_bem_to_section by Eric Larson

• Fix bug with convex-hull based MEG helmet creation where the face area could be covered up, by Eric Larson

• Fix bug with mne.inverse_sparse.mixed_norm() and mne.inverse_sparse.gamma_map() with return_as_dipoles=True and mne.beamformer.rap_music() where dip.gof was the same for all dipoles. GOF values / explained variance across all dipoles now gets split among the dipoles using SVD of the lead fields. To get the total GOF, sum(dip.gof for dip in dipoles) can be used, by Eric Larson

• Fix bug with mne.inverse_sparse.gamma_map() where output data was not assigned properly when using a free-orientation forward and xyz_same_gamma=False by Eric Larson

• Fix to enable interactive plotting with no colorbar with mne.viz.plot_evoked_topomap() by Daniel McCloy

• Fix plotting with mne.viz.plot_evoked_topomap() to pre-existing axes by Daniel McCloy

• Fix bug with mne.viz.plot_vector_source_estimates() using the PyVista backend with time_viewer=True when updating the arrow colormaps by Eric Larson

• Fix bug with mne.viz.plot_vector_source_estimates() where clim='auto' and clim=dict(..., kind='percent') did not take into account the magnitude of the activation, by Eric Larson

• The default plotting mode for mne.io.Raw.plot() and mne browse_raw has been changed to clipping=3. to facilitate data analysis with large deflections, by Eric Larson

• PSD plots will now show non-data channels (e.g., misc) if those channels are explicitly passed to picks, by Daniel McCloy.

• Fix bug with mne.time_frequency.read_tfrs() where info['meas_date'] was not parsed correctly, by Eric Larson

• Fix bug with mne.time_frequency.tfr_array_stockwell() where inputs were not properly validated by Eric Larson

• Fix handling of NaN when using TFCE in clustering functions such as mne.stats.spatio_temporal_cluster_1samp_test() by Eric Larson

• Fix handling of signs when using TFCE by Eric Larson

• The mne.MixedSourceEstimate class has been clarified to contain two cortical surface source spaces, plus at least one other source space. Creating source estimates in other orderings is not supported, by Eric Larson

• Fix bug where VolSourceEstimate.vertices was an instance of ndarray instead of list of one ndarray, by Eric Larson

• Fix default to be foreground=None in mne.viz.plot_source_estimates() to use white or black text based on the background color by Eric Larson

• Fix bug with writing EGI and CTF mne.Info to H5 format, e.g., with mne.time_frequency.AverageTFR.save by Eric Larson

• Fix bug with topomap outlines in mne.time_frequency.AverageTFR.plot_joint() and mne.time_frequency.AverageTFR.plot_topo() by Eric Larson

• Fix bug with mne.io.Raw.plot() where toggling all projectors did not actually take effect by Eric Larson

• Fix bug with mne.write_proj() where the number of columns was incorrectly written by Eric Larson

• Fix bug with mne.read_epochs() when loading data in complex format with preload=False by Eric Larson

• Fix bug with mne.Epochs.save() where the file splitting calculations did not account for the sizes of non-data writes by Eric Larson

• Fix bug with mne.Epochs when metadata was not subselected properly when event_repeated='drop' by Eric Larson

• Fix bug with mne.Epochs where epochs.drop_log was a list of list of str rather than an immutable tuple of tuple of str (not meant to be changed by the user) by Eric Larson

• Fix bug with mne.Report where the BEM section could not be toggled by Eric Larson

• Fix bug when using mne.Epochs.crop() to exclude the baseline period would break mne.Epochs.save() / mne.read_epochs() round-trip by Eric Larson

• Fix bug with mne.Epochs.subtract_evoked where using decimated epochs would lead to an error by Eric Larson

• Fix bug with mne.viz.plot_bem() and mne.Report when plotting BEM contours when MRIs are not in standard FreeSurfer orientation by Eric Larson

• Fix bug with mne.minimum_norm.make_inverse_operator() where it would warn even when an explicit rank was used by Eric Larson

• Fix bugs with mne.beamformer.make_lcmv() and mne.beamformer.make_dics() where:

  • Noise normalization factors weight_norm='unit-noise-gain' and weight_norm='nai' were computed incorrectly

  • pick_ori='max-power' computed the max-power orientation incorrectly

  • pick_ori='normal' did not compute power or noise normalization factors correctly

  • mne.beamformer.apply_lcmv_cov() did not apply whitening and projections properly

• Fix mne setup_forward_model to have it actually compute the BEM solution in addition to creating the BEM model by Eric Larson

• Fix bug with mne.io.read_raw_edf() where null bytes were not properly handled, causing an error when opening a file by Eric Larson

• Fix bug with mne.Report() where unicode characters were not rendered properly (encoding for HTML was not set) by Eric Larson

• Fix bug with mne.preprocessing.nirs.scalp_coupling_index() where filter transition was incorrectly assigned by Robert Luke

• Fix bug with mne.make_forward_dipole() where mne.write_forward_solution() could not be used by Eric Larson

• Fix bug with mne.io.Raw.plot() when scalings='auto' where bad data would prevent channel plotting by Eric Larson

• Default border and extrapolate arguments for mne.Evoked.plot_topomap() and related functions were changed from 0. to 'mean' and 'box' to 'auto', respectively, to help more accurately reflect sensor geometries and boundary conditions. extrapolate='auto' uses extrapolate='local' for MEG data and extrapolate='head' otherwise, by Eric Larson

• Fix bug that prevents n_jobs from being a NumPy integer type, by Daniel McCloy.

• Fix bug with mne.epochs.average_movements() where epoch weights were computed using all basis vectors instead of the internal basis only by Eric Larson

• Fix bug with mne.get_volume_labels_from_aseg() where the returned labels were alphabetical instead of reflecting their volumetric ID-based order by Eric Larson

• Fix bug with mne.preprocessing.find_bad_channels_maxwell() where good data of exactly step duration would lead to an error by Eric Larson

• Fix bug with mne.preprocessing.find_bad_channels_maxwell() where indices were not handled properly when MEG channels were not first in the raw instance, and logging messages incorrectly reported the interval used by Eric Larson

• Make mne.set_config() accept path-like input values by Richard Höchenberger

• Fix bug with mne.write_labels_to_annot() and mne.datasets.fetch_hcp_mmp_parcellation() where label name strings were not properly terminated, leading to problems loading in FreeSurfer by Eric Larson

• Fix bug with mne.beamformer.make_dics() where complex conjugates were not applied properly by Britta Westner and Eric Larson

• Fix bug with mne.bem.make_watershed_bem() where the RAS coordinates of watershed bem surfaces were not updated correctly from the volume file by Yu-Han Luo

• Fix bug with mne.io.Raw.get_channel_types() and related methods where the ordering of picks was not preserved, by Eric Larson

• Fix bug with mne.io.Raw.plot_psd() with average=False and multiple channel types where channel locations were not shown properly by Eric Larson

• Fix bug with mne.viz.plot_sparse_source_estimates() when using 'sphere' mode by Eric Larson and Guillaume Favelier

• Fix bug in FieldTrip reader functions when channels are missing in the info object by Thomas Hartmann

• Throw proper error when trying to import FieldTrip Epochs data with non-uniform time for trials by Thomas Hartmann

• Throw proper error when trying to import FieldTrip data saved by an old, incompatible version by Thomas Hartmann

• Fix bug in mne.read_epochs_fieldtrip() when importing data without a trialinfo field by Thomas Hartmann

• Fix bug in mne.preprocessing.ICA.plot_properties() where time series plot doesn’t start at the proper tmin by Teon Brooks

• Fix bug with mne.preprocessing.ICA.plot_properties() where a mne.io.Raw object with annotations would lead to an error by Yu-Han Luo

• Fix bug with mne.events_from_annotations(raw.annotations) when orig_time of annotations is None and raw.first_time > 0, by Alex Gramfort

API changes#

• Python 3.5 is no longer supported, Python 3.6+ is required, by Eric Larson

• adjacency has replaced connectivity in the names of:

  1. Arguments to clustering functions, such as mne.stats.permutation_cluster_test, and

  2. Function names for defining adjacency, such as mne.spatio_temporal_src_adjacency replacing mne.spatio_temporal_src_connectivity.

  The complete list of changed function names is:

  • mne.channels.find_ch_connectivity → find_ch_adjacency

  • mne.channels.read_ch_connectivity → read_ch_adjacency

  • mne.spatial_dist_connectivity → spatial_dist_adjacency

  • mne.spatial_inter_hemi_connectivity → spatial_inter_hemi_adjacency

  • mne.spatial_src_connectivity → spatial_src_adjacency

  • mne.spatial_tris_connectivity → spatial_tris_adjacency

  • mne.spatio_temporal_dist_connectivity → spatio_temporal_dist_adjacency

  • mne.spatio_temporal_src_connectivity → spatio_temporal_src_adjacency

  • mne.spatio_temporal_tris_connectivity → spatio_temporal_tris_adjacency

  “connectivity” is now reserved for discussions of functional and effective connectivity of the brain, and “adjacency” for source or sensor neighbor definitions for cluster-based analyses, by Eric Larson.

• The default for the standardize_names argument of mne.io.read_raw_kit() will change from True to False in 0.22, by Eric Larson

• The normalize_fwd argument of mne.beamformer.make_dics() has been deprecated in favor of depth, by Eric Larson

• The overwrite parameter of mne.write_bem_surfaces() and mne.write_bem_solution() defaults to True in 0.21 but will change to False in 0.22, by Eric Larson

• Add n_cols parameter to mne.preprocessing.ICA.plot_scores() to allow plotting scores in multiple columns, by Luke Bloy

• In mne.stats.permutation_cluster_test() and mne.stats.permutation_cluster_1samp_test() the default parameter value out_type='mask' has changed to None, which in 0.21 means 'mask' but will change to mean 'indices' in the next version, by Daniel McCloy

• The default window size set by filter_length when method='spectrum_fit' in mne.io.Raw.notch_filter() will change from None (use whole file) to '10s' in 0.22, by Eric Larson

• vmin and vmax parameters are deprecated in mne.Epochs.plot_psd_topomap() and mne.viz.plot_epochs_psd_topomap(); use new vlim parameter instead, by Daniel McCloy.

• The method stc_mixed.plot_surface for a mne.MixedSourceEstimate has been deprecated in favor of stc.surface().plot(...) by Eric Larson

• The method stc.normal for mne.VectorSourceEstimate has been deprecated in favor of stc.project('nn', src) by Eric Larson

• Add use_dev_head_trans parameter to mne.preprocessing.annotate_movement() to allow choosing the device to head transform is used to define the fixed cHPI coordinates by Luke Bloy

• The function mne.channels.read_dig_captrack will be deprecated in version 0.22 in favor of mne.channels.read_dig_captrak() to correct the spelling error: “captraCK” -> “captraK”, by Stefan Appelhoff

• The threshold argument in mne.preprocessing.ICA.find_bads_ecg() defaults to None in version 0.21 but will change to 'auto' in 0.22 by Yu-Han Luo

• The function mne.preprocessing.mark_flat has been deprecated in favor of mne.preprocessing.annotate_flat by Eric Larson

• The default argument meg=True in mne.pick_types() will change to meg=False in version 0.22 by Clemens Brunner

• Added mne.io.Raw.get_montage(), which obtains the montage that an instance has, by Adam Li