Version 0.19 (2019-09-24)

Changelog

• Add mne.cuda.set_cuda_device() and config variable MNE_CUDA_DEVICE to select among multiple GPUs (by numeric device ID) by Daniel McCloy.

• Add mne.channels.make_standard_montage() to create mne.channels.DigMontage from templates by Joan Massich and Alex Gramfort.

• Add mne.channels.compute_dev_head_t() to compute Device-to-Head transformation from a montage by Joan Massich and Alex Gramfort.

• Add mne.channels.read_dig_fif() to read digitization coordinates from .fif files by Joan Massich and Alex Gramfort.

• Add mne.channels.read_dig_egi() to read digitization coordinates from EGI .xml files by Joan Massich and Alex Gramfort.

• Add mne.channels.read_dig_polhemus_isotrak() and mne.channels.read_polhemus_fastscan() to read Polhemus data by Joan Massich

• Add mne.channels.read_dig_captrack to read BrainVision CapTrak (BVCT) digitization coordinate files by Stefan Appelhoff and Joan Massich

• Add mne.channels.make_dig_montage() to create mne.channels.DigMontage objects out of np.arrays by Joan Massich

• Add mne.channels.read_custom_montage() to read various EEG electrode locations files by Joan Massich and Alex Gramfort.

• Add support for making epochs with duplicated events, by allowing three policies: “error” (default), “drop”, or “merge” in mne.Epochs by Stefan Appelhoff

• Allow mne.Annotations.crop() to support negative tmin and tmax by Joan Massich

• Unknown events code in GDF are now visible in the event_id by Théodore Papadopoulo

• Now mne.io.read_raw_ctf() populates raw.annotations with the markers in MarkerFile.mrk if any by Joan Massich

• Add options for controlling the use of the -T1 flag and the location of the brainmask output in mne watershed_bem by Eric Larson

• Add support to mne.read_annotations() to read CTF marker files by Joan Massich

• Do not convert effective number of averages (nave attribute of mne.Evoked) to integer except when saving to FIFF file by Daniel McCloy.

• Add automatic fiducial position estimation in mne coreg using MNI Talairach fiducial locations in mne.coreg.get_mni_fiducials() by Jon Houck and Eric Larson

• Add support for mne coreg scaling surrogate subjects without surface reconstructions, such as those created for volumetric analyses only (e.g., with recon-all -autorecon1) by Eric Larson

• Add reader for Curry data in mne.io.read_raw_curry() by Dirk Gütlin

• Butterfly channel plots now possible for mne.Epochs.plot_psd() with average=False. Infrastructure for this function now shared with analogous Raw function, found in mne.viz.utils by Jevri Hanna

• Add option not to orthogonalize power envelopes with orthogonalize=False in mne.connectivity.envelope_correlation by Denis Engemann

• Accept filenames of raw .fif files that end in _meg.fif to enable complicance with the Brain Imaging Data Structure by Stefan Appelhoff

• Add function to check the type of a FIF file using mne.what() and mne what by Eric Larson

• Add support for specifying the initial time and/or position and providing a mne.SourceMorph instead of mne.SourceSpaces in mne.viz.plot_volume_source_estimates() by Eric Larson

• Speed up morph map generation in mne.read_morph_map() by ~5-10x by using numba.jit() by Eric Larson

• Speed up mne.setup_volume_source_space(), especially when volume_label is not None by Eric Larson

• Speed up mne coreg interactive and automated (ICP) alignment by using nearest-neighbor calculations in the MRI coordinate frame, by Eric Larson

• Add mne.dig_mri_distances() to compute the distances between digitized head points and the MRI head surface by Alex Gramfort and Eric Larson

• Add scale bars for data channels in mne.io.Raw.plot() by Eric Larson

• Add mne.viz.plot_brain_colorbar() to plot a colorbar appropriately matched to a mne.viz.plot_source_estimates() plot by Eric Larson

• Add support for showing head surface (to visualize digitization fit) while showing a single-layer BEM to mne.viz.plot_alignment() by Eric Larson

• Add option include_tmax=True to cropping methods mne.io.Raw.crop(), mne.Epochs.crop(), mne.Evoked.crop(), mne.SourceEstimate.crop(), mne.Dipole.crop(), and mne.time_frequency.AverageTFR.crop() by Eric Larson

• Change the behavior of mne.io.Raw.plot() for scalings='auto' and remove_dc=True to compute the scalings on the data with DC removed by Clemens Brunner

• Allow creating annotations within existing annotations in mne.io.Raw.plot() by default (the old snapping behavior can be toggled by pressing ‘p’) by Clemens Brunner

• Add plotting of rank estimates in mne.viz.plot_cov() by Eric Larson

• Add a new mne.viz.plot_sensors_connectivity function to visualize the sensor connectivity in 3D by Guillaume Favelier and Alex Gramfort

• Add support for info['utc_offset'], info['device_info'], and info['helium_info'] components of mne.Info by Eric Larson

• Add control over dipole colors in mne.viz.plot_dipole_locations() when using orthoview mode by Eric Larson

• Use second-order-sections filtering in mne.io.Raw.plot() and mne browse_raw by Eric Larson

• Add re-referencing functionality for ecog and seeg channel types in mne.set_eeg_reference() by Keith Doelling

• Add support for median averaging and for returning unaggregated segments in mne.time_frequency.psd_welch by Richard Höchenberger

• io.read_raw_kit(): add support for NYU New York 2019 system update, by Christian Brodbeck

Bug

• Fix setting montage eeg ref position for captrak by Joan Massich

• Fix saving raw read from BDF file using tmin and tmax using preload=False by Alex Gramfort

• Fix mne.grand_average() to use equal sum-to-one weights (like it used to, before changes to underlying mne.combine_evoked()) by Daniel McCloy

• Fix mne.io.Raw.filter() to deal with instances with no data channels properly by Eric Larson

• Fix one-sample baseline issue in mne.BaseEpochs when using tmin=0 by Milan Rybář

• Fix bug in mne.viz.plot_volume_source_estimates() where 'glass_brain' MRIs were not transformed to MNI space, by Eric Larson

• Fix bug in mne.viz.plot_volume_source_estimates() where MRIs with voxels not in RAS orientation could not be browsed properly, by Eric Larson

• Fix bug in mne.SourceMorph.apply() where output STCs had stc.vertices defined improperly, by Eric Larson

• Fix bug in mne.SourceMorph.apply() where the default was errantly mri_space=False instead of mri_space=None (as documented), by Eric Larson

• Fix mne.VolVectorSourceEstimate.normal and mne.minimum_norm.apply_inverse() to only allow normal extraction/orientation for surface and discrete source spaces by Eric Larson

• Fix mne.io.Raw.set_annotations() for meas_date previous to 1970 by Joan Massich

• Fix horizontal spacing issues in mne.io.Raw.plot_psd() by Jevri Hanna

• Fix reading of dates in BrainVision files if no “New Segment” marker is specified, no date is given, or data is missing, by Stefan Appelhoff

• Fix bug with reading one-channel GDF files by Abram Hindle

• Fix bug with y-axis labeling in mne.io.Raw.plot_psd() by Eric Larson

• Fix side-effect where mne.viz.plot_ica_sources() and mne.preprocessing.ICA.plot_sources() changed the ICA.exclude attribute even when users didn’t interact with the plot by Daniel McCloy.

• Fix scaling of sources in ica.plot_sources(epochs) by Eric Larson

• Fix wrong assumptions about units in BrainVision montages and add test asserting units in “mm” or “auto”, by Stefan Appelhoff

• Fix scaling issue with signals in mV in EDF files read with mne.io.read_raw_edf() by Alex Gramfort

• Fix mne.extract_label_time_course() behavior when using STCs generated with apply_inverse(..., label=label) or stc.in_label by Eric Larson

• Fix bug in mne.io.read_raw_brainvision() so that recording date timestamps are also recognized if channel reference data is negative, by Stefan Appelhoff

• Fix order of info['dig'] that was alphabetical based on channel names and not following the channel order when using mne.io.Raw.set_montage() and a mne.channels.Montage object by Joan Massich and Alex Gramfort.

• Fix reading CNT files larger than 2Gb by Joan Massich

• Fix reading of 4D/BTi data with different channel names and onsets in mne.io.read_raw_bti() by Joshua Bear and Eberhard Eich

• Fix formula for effective number of averages in mne.combine_evoked() when weights='equal' by Daniel McCloy.

• Fix bug in mne.simulation.simulate_stc() to avoid empty stc if label vertices and source space do not intersect, by Kostiantyn Maksymenko

• Fix event_id='auto' in mne.events_from_annotations() to recover Brainvision markers after saving it in .fif by Joan Massich

• Fix mne.read_epochs_eeglab() when epochs are stored as float. By Thomas Radman

• Fix mne.Evoked.resample() and mne.Epochs.resample() not setting inst.info['lowpass'] properly by Eric Larson

• Fix checks when constructing volumetric and surface source spaces with mne.setup_volume_source_space() and mne.setup_source_space(), respectively, by Eric Larson

• Fix bug in handling of mne.Evoked types that were not produced by MNE-Python (e.g., alternating average) by Eric Larson

• Fix bug in mne.read_source_estimate() where vector volumetric source estimates could not be read by Eric Larson

• Fix bug in mne.inverse_sparse.mixed_norm() and mne.inverse_sparse.tf_mixed_norm() where weights was supplied but weights_min was not, by Eric Larson

• Fix bug in mne.set_eeg_reference() where non-EEG channels could be re-referenced by default if there were no EEG channels present, by Eric Larson

• Fix bug in mne.io.Raw.plot() when using HiDPI displays and the MacOSX backend of matplotlib by Eric Larson

• Fix bug in mne.viz.plot_compare_evokeds() when using Neuromag 122 system by Eric Larson

• Fix bug in mne.Epochs.plot_psd() when some channels had zero/infinite psd values causing erroneous error messages by Luke Bloy

• Fix mne.Evoked.decimate() not setting inst.first and inst.last properly by Marijn van Vliet

• Fix mne.io.read_raw_brainvision() not handling Event markers created by PyCorder correctly by Richard Höchenberger

• Fix support for string-like objects (such as pathlib.Path) by Eric Larson

• Fix mne.Report silently suppressing exceptions when used as a context manager by Marijn van Vliet

API

• Deprecate mne.channels.Montage class, mne.channels.read_montage and mne.channels.read_dig_montage function by Joan Massich.

• Deprecate passing Montage, str as montage parameter in mne.io.Raw.set_montage() by Joan Massich.

• Deprecate set_dig parameter in mne.io.Raw.set_montage() and update_ch_names in mne.io.RawEEGLAB.set_montage when using mne.channels.DigMontage as by Joan Massich.

• Now mne.io.Raw.set_montage() raises an error when mne.channels.DigMontage contains only a subset of the channels in raw.info. It also adds raise_if_subset parameter to ensure backward compatibility (defaults to False in 0.19, to True in 0.20, and will be removed in 0.21) by Joan Massich.

• Minimum dependency versions for the following libraries have been bumped up (by Eric Larson):

  • NumPy: 1.12.1

  • SciPy: 0.18.1

  • matplotlib: 2.0.2

  • scikit-learn: 0.18.2

  • pandas 0.19.2

• New boolean parameter show_scrollbars for mne.io.Raw.plot(), mne.Epochs.plot(), and mne.preprocessing.ICA.plot_sources() (and associated functions) that allows hiding the scrollbars and buttons for a “zen mode” data browsing experience. When the plot window has focus, zen mode can be toggled by pressing z, by Daniel McCloy.

• Deprecate mne.evoked.grand_average in favor of mne.grand_average() (which works on both Evoked and AverageTFR) by Daniel McCloy

• Deprecate exclude parameter in mne.viz.plot_ica_sources() and mne.preprocessing.ICA.plot_sources(), instead always use the exclude attribute of the ICA object by Daniel McCloy.

• Deprecate montage parameter in favor of the set_montage method in all EEG data readers mne.io.read_raw_cnt(), mne.io.read_raw_egi(), mne.io.read_raw_edf(), mne.io.read_raw_gdf(), mne.io.read_raw_nicolet(), mne.io.read_raw_eeglab() and mne.read_epochs_eeglab() by Alex Gramfort

• New parameter clear in mne.viz.plot_epochs_image() for clearing pre-existing axes before plotting into them by Daniel McCloy

• mne.viz.plot_epochs_image() no longer supports group_by='type' — combining by channel type is now the default when picks is a channel type string; to get individual plots for each channel, pass picks as a list of channel names or indices by Daniel McCloy

• New parameter combine in mne.viz.plot_compare_evokeds() for specifying method to combine information across channels by Daniel McCloy

• FIFF constants related to SmartShield (*_SMSH_* and *_SMARTSHIELD) have been renamed to IAS for consistency with MEGIN, by Eric Larson

• The gfp parameter of mne.viz.plot_compare_evokeds() is deprecated; use combine='gfp' instead by Daniel McCloy

• The truncate_yaxis='max_ticks' parameter of mne.viz.plot_compare_evokeds() is deprecated; use truncate_yaxis='auto' instead by Daniel McCloy

• The truncate_xaxis and truncate_yaxis parameters of mne.viz.plot_compare_evokeds() now perform one-sided truncation unless both are True by Daniel McCloy

• The show_legend parameter of mne.viz.plot_compare_evokeds() is renamed to legend by Daniel McCloy

• mne.viz.plot_compare_evokeds() always returns a list of figures even when a single figure is generated by Daniel McCloy

• Deprecate average=True and spatial_colors=False for mne.Epochs.plot_psd() by Jevri Hanna

• mne.io.read_raw_brainvision() no longer raises an error when there are inconsistencies between info['chs'] and montage but warns instead by Joan Massich

• Add update_ch_names parameter to mne.io.RawEEGLAB.set_montage to allow updating the channel names based on the montage by Joan Massich

• Reading annotations contained in GDF files with mne.io.read_raw_gdf() now returns numeric event codes as descriptions (instead of textual descriptions) due to restrictive licensing of the GDF event code table from BioSig by Clemens Brunner

• channels.find_ch_connectivity now returns pre-built neighbor maps for KIT systems when available, by Christian Brodbeck