Version 1.8.0.dev124+g3cc01923a (development) (2024-07-26)

Dependencies

• Minimum versions for dependencies were bumped to those ~2 years old at the time of release (by Eric Larson), including:

  • NumPy ≥ 1.23

  • SciPy ≥ 1.9

  • Matplotlib ≥ 3.6

  • scikit-learn ≥ 1.1 (#12554)

• Official support for PySide2 has been dropped in this release (though it might continue to work), by Eric Larson. (#12742)

Bugfixes

• Fix bug where an event that occurred only once was excluded in mne.io.read_raw_egi(), by new contributor Ping-Keng Jao. (#12300)

• Fix scrolling behavior in Report when clicking on a TOC entry multiple times, by Richard Höchenberger. (#12561)

• Fix bug where mne.decoding.get_coef() did not work properly with mne.decoding.CSP, by Eric Larson. (#12562)

• The color scaling of Evoked topomaps added to reports via mne.Report.add_evokeds() was sometimes sub-optimal if bad channels were present in the data. This has now been fixed and should be more consistent with the topomaps shown in the joint plots, by Richard Höchenberger. (#12578)

• Fix error causing mne.Epochs.interpolate_bads() not to work for seeg channels and fix a single contact on neighboring shafts sometimes being included in interpolation, by Alex Rockhill (#12593)

• Fix bug where mne.io.read_raw_fil() could not assign bad channels on import, by George O’Neill. (#12597)

• Fixed a bug where mne.Evoked.animate_topomap() did not work with mne.preprocessing.compute_current_source_density() - modified data, by Michal Žák. (#12605)

• Fix bug where mne.stats.permutation_cluster_test() (and related functions) uses excessive amount of memory for large 2D data when TFCE method is selected, by new contributor Nicolas Fourcaud-Trocmé. (#12609)

• Fix overflow when plotting source estimates where data is all zero (or close to zero), and fix the range of allowed values for the colorbar sliders, by Marijn van Vliet. (#12612)

• Fix adding channels to EpochsTFR objects, by Clemens Brunner. (#12616)

• Fix for new sklearn metadata routing protocol in decoding search_light, by Alex Gramfort (#12620)

• Fix bug where mne.time_frequency.csd_multitaper(), mne.time_frequency.csd_fourier(), mne.time_frequency.csd_array_multitaper(), and mne.time_frequency.csd_array_fourier() would return cross-spectral densities with the fmin and fmax frequencies missing, by Thomas Binns (#12633)

• Output types of sparse arrays were changed from matrix to array in read_ch_adjacency(), find_ch_adjacency(), combine_adjacency(), spatio_temporal_src_adjacency(), and related functions to comply with the pending deprecation of np.matrix. The returned objects now behave like standard ndarray objects, and in particular * now operates element-wise instead of performing matrix multiplication. You can use @ as a backward compatible matrix multiplication for both np.matrix and np.ndarray objects, and if a matrix is desired the outputs can be cast directly, for example as scipy.sparse.csr_matrix(out). Changed by Eric Larson. (#12646)

• Fix incorrect RuntimeWarning (different channel filter settings) in EDF/BDF import, by Clemens Brunner. (#12661)

• In mne.export.export_raw() (fmt='edf'), when padding data to create equal-length data blocks, edge-padding is favored over zero-padding in order to avoid accidentally enlarging physical range, by Qian Chu. (#12676)

• In mne.io.read_raw_eyelink(), gracefully handle missing datetime in file by Scott Huberty. (#12687)

• Fix regression with mne.decoding.CSP where using rank="full" errantly raised an error, by Eric Larson. (#12694)

• Fix in-memory anonymization of data read with mne.io.read_raw_edf() by Eric Larson. (#12720)

• Fix bug with overplotting of butterfly labels in mne.viz.plot_raw() and related functions and methods, by Eric Larson. (#12721)

• Fix a bug where the ylim parameter would sometimes apply to the wrong channel types in mne.viz.plot_evoked_topo(), by Marijn van Vliet. (#12724)

• Fix bug where ad-hoc regularization of a covariance with mne.cov.regularize() did not properly account for bad channels in rank calculations, by Eric Larson. (#12727)

• set_log_file() and set_log_level() can now be correctly detected by static analysis tools like Pylance, by Richard Höchenberger. (#12735)

API changes by deprecation

• A new argument events_as_annotations has been added to mne.io.read_raw_egi() with a default value of False that will change to True in version 1.9, by Scott Huberty and Eric Larson. (#12300)

• mne.Info.ch_names will now return an empty list instead of raising a KeyError if no channels are present, by Richard Höchenberger. (#12583)

• Documented that match_channel_orders() can also work on Epochs, and Evoked objects. Reflecting this, deprecated the raws parameter in favor of an insts parameter, by Stefan Appelhoff. (#12699)

• The block argument to mne.viz.Brain is deprecated and will be removed in 1.9, use matplotlib.pyplot.show() with block=True instead, by Eric Larson. (#12719)

New features

• Adding mne.channels.Layout.copy() and mne.channels.Layout.pick() to copy and select channels from a mne.channels.Layout object. Plotting 2D topographies of evoked responses with mne.viz.plot_evoked_topo() with both arguments layout and exclude now ignores excluded channels from the mne.channels.Layout. By Mathieu Scheltienne. (#12338)

• In Report you can now easily navigate through images and figures connected to a slider with the left and right arrow keys. Clicking on the slider or respective image will focus the slider, enabling keyboard navigation, by Richard Höchenberger (#12556)

• When plotting EOG and ECG artifact scores for ICA in mne.Report.add_ica(), the channel names used for artifact detection are now displayed in the titles of each respective subplot, by Richard Höchenberger. (#12573)

• Use aseg='auto' for mne.viz.Brain.add_volume_labels() and mne.get_montage_volume_labels() to use aparc+aseg by default or if not present use wmparc because freesurfer uses wmparc in the latest version, by Alex Rockhill. (#12576)

• The HTML representations of Raw, Epochs, and Evoked (which you will see e.g. when working with Jupyter Notebooks or Report) have been updated to be more consistent and contain slightly more information, by Richard Höchenberger. (#12624) (#12583)

• When adding Evoked data to a Report via add_evokeds(), we now also include an “Info” section with some basic summary info, as has already been the case for raw and epochs data, by Richard Höchenberger. (#12584)

• We added new installation variants for a full installation with the PySide6 Qt binding ("mne[full-pyside6]"), with the PyQt6 binding ("mne[full-pyqt6]", equivalent to "mne[full]"), and without any Qt binding ("mne[full-no-qt]"), which may be useful in certain situations by Richard Höchenberger. (#12628)

• Adding argument 'random' to equalize_epoch_counts() and to equalize_event_counts() to randomly select epochs or events. By Mathieu Scheltienne. (#12649)

• Add new parameter ignore_marker_types to read_raw_brainvision() to ignore marker types (and only use marker descriptions) when reading BrainVision files, by Clemens Brunner. (#12652)

• Added support for passing axes to mne.viz.plot_head_positions() when mode='field', by Eric Larson. (#12655)

• Added internals to allow modifying single-channel annotations in the Qt raw browser, by new contributor Noah Markowitz. (#12669)

• The HTML representation of Epochs (visible e.g. in Jupyter or in Report) now indicates whether metadata is attached to the epochs, by Richard Höchenberger. (#12686)

• Montage plots created with plot() now scale both the channel dots and channel names with the new scale parameter. The default is scale=1 (factors less than 1 will scale down, whereas factors greater than 1 will scale up). The previous scale_factor parameter only affected marker size, and this parameter is now deprecated. By Clemens Brunner. (#12703)

• Add compute_sme() to compute the analytical standardized measurement error (SME) as a data quality measure for ERP studies, by Clemens Brunner. (#12707)

• Use datetime.date for info["subject_info"]["birthday"] rather than a tuple of (year, month, day) by Eric Larson. (#12720)

• Add default spatial_colors="auto" to mne.viz.plot_evoked_white() and mne.Evoked.plot_white() to enable spatial colors by default, by Eric Larson. (#12725)

• mne.minimum_norm.make_inverse_operator() and related functions now more robustly warn if the whitener computed from the noise covariance has an incorrect rank, by Eric Larson. (#12727)

• When indexing Epochs (e.g. by doing epochs[0]), static code analysis tools like Pylance should now be able to infer that the returned object is an epoch, too, and provide editor support like automated code completions, by Richard Höchenberger. (#12733)

• When using the data_path() in any dataset included in mne.datasets, static analysis tools like Pylance will now correctly infer that a pathlib.Path will be returned and provide editor support like automated code completions, by Richard Höchenberger. (#12734)

Other changes

• Added vulture as a pre-commit hook and removed related dead code, by Eric Larson. (#12569)

• Fixed issue template links by new contributor Michal Žák (#12600)

• Enhance documentation on decimation filtering to prevent aliasing, by new contributor Xabier de Zuazo. (#12650)

• Add link to BEst-Python in the list of related software. by Ilian Azz. (#12659)

• Improved clarity of parameter documentation for mne.decoding.SSD.fit, by Thomas Binns. (#12664)

• Disable the “Back to top” button in the documentation, by Richard Höchenberger. (#12688)

• Improve argument ylim documentation through Evoked plotting function and validate type to dict or None to prevent misuage, by Mathieu Scheltienne. (#12697)