Version 1.9.0.dev107+ga99aa9f79 (development) (2024-11-20)

Dependencies

• • Minimum supported dependencies were updated in accordance with SPEC0, most notably Python 3.10+ is now required. (#12798)

• Importing from mne.decoding now explicitly requires scikit-learn to be installed, by Eric Larson. (#12834)

Bugfixes

• Fix handling of MRI file-path in mne.SourceSpaces and safeguard saving of pathlib.Path with h5io by casting to str, by Mathieu Scheltienne. (#12803)

• Cast fwd["info"] to Info and fwd["src"] to SourceSpaces when loading a forward solution from an HDF5 file, by Mathieu Scheltienne. (#12804)

• Fix bug where mne.Epochs.compute_tfr() could not be used with the multitaper method and complex or phase outputs, by Thomas Binns. (#12842)

• Fixed a bug where split FIF files that were read and then appended to other mne.io.Raw instances had their BAD boundary annotations incorrectly offset in samples by the number of split files, by Eric Larson. (#12843)

• Enforce SI units for Eyetracking data (eyegaze data should be radians of visual angle, not pixels. Pupil size data should be meters). Updated tutorials so demonstrate how to convert data to SI units before analyses, by Scott Huberty. (#12846)

• Prevent the colorbar parameter being ignored in topomap plots such as mne.time_frequency.Spectrum.plot_topomap(), by Thomas Binns. (#12853)

• Fix bug where invalid data types (e.g., np.ndarray``s) could be used in some :class:`mne.io.Info` fields like ``info["subject_info"]["weight"], by Eric Larson. (#12875)

• When creating a SpectrumArray, the array shape check now compares against the total of both ‘good’ and ‘bad’ channels in the provided Info (previously only good channels were checked), by Mathieu Scheltienne. (#12877)

• Fix bug where mne coreg would always show MEG channels even if the “MEG Sensors” checkbox was disabled, by Eric Larson. (#12884)

• mne.Report HDF5 files are now written in mode='a' (append) to allow users to store other data in the HDF5 files, by Eric Larson. (#12901)

• Fix bug in mne.io.read_raw_gdf() when NumPy >= 2 is used, by Clemens Brunner. (#12909)

• Allow exporting edf where a channel contains only constant values, by Florian Hofer. (#12911)

• Fix typos in the Spatio-Spectral Decomposition example, by new contributor Simon M. Hofmann. (#12924)

• Fix a bug in mne.epochs.make_metadata(), where missing values in the columns generated for keep_first and keep_last events were represented by empty strings, while it should have been NA values, by Richard Höchenberger. (#12931)

• Fix decimal places of float mne.Evoked.nave in mne.Evoked.plot() and mne.Evoked.plot_image(), by Gennadiy Belonosov. (#12936)

• Fix duration calculation for the textual (__repr__) and html (_repr_html_, used by e.g. Jupyter) display of mne.io.Raw instances. For example a duration of 1h is now displayed as 00:01:00 rather than 00:59:60. By new contributor Leonardo Rochael Almeida. (#12955)

• Fix displayed units in representations of classes such as mne.io.Raw to correctly use KiB, MiB, GiB, and so on, by Clemens Brunner. (#12962)

• Gracefully handle invalid patient info when reading EDF files by Scott Huberty. (#12968)

API changes by deprecation

• Deprecate average parameter in plot_filters and plot_patterns methods of the mne.decoding.CSP and mne.decoding.SPoC classes, by Thomas Binns. (#12829)

• Deprecate subject parameter in favor of subjects in mne.datasets.eegbci.load_data(), by Stefan Appelhoff. (#12918)

New features

• Add support for dict type argument ref_channels to mne.set_eeg_reference(), to allow flexible re-referencing (e.g. raw.set_eeg_reference(ref_channels={'A1': ['A2', 'A3']}) will set the new A1 data to be A1 - mean(A2, A3)), by Alex Lepauvre and Qian Chu and Daniel McCloy. (#12366)

• Add reader for ANT Neuro files in the *.cnt format with read_raw_ant(), by Mathieu Scheltienne, Eric Larson and Proloy Das. (#12792)

• • Add support for a mne.transforms.Transform in the argument trans of the coregistration GUI called with mne.gui.coregistration(), by Mathieu Scheltienne. (#12801)

• Added support for sensor_scales to mne.viz.Brain.add_sensors() and mne.viz.plot_alignment(), by new contributor Alex Lepauvre. (#12805)

• Raw() and save() now return the path to the saved file(s), by Victor Ferat. (#12811)

• mne.channels.read_custom_montage() may now read a newer version of the .elc ASA Electrode file format, by Stefan Appelhoff. (#12830)

• Added the title argument to mne.viz.create_3d_figure(), and color and position arguments to mne.viz.set_3d_title(), by Eric Larson. (#12871)

• Improved reporting and plotting options:

  • mne.Report.add_projs() can now plot with mne.viz.plot_projs_joint() rather than mne.viz.plot_projs_topomap()

  • mne.Report now has attributes img_max_width and img_max_res that can be used to control image scaling.

  • mne.Report now has an attribute collapse that allows collapsing sections and/or subsections by default.

  • mne.viz.plot_head_positions() now has a totals=True option to show the total distance and angle of the head.

  Changes by Eric Larson. (#12901)

• Added the psd_args argument to mne.viz.plot_ica_sources() and mne.preprocessing.ICA.plot_sources(), by Richard Scholz. (#12912)

• Add convenience mne.io.Raw.duration property to centralize duration calculation for the textual (__repr__) and html (_repr_html_, used by e.g. Jupyter) display of mne.io.Raw instances, by new contributor Leonardo Rochael Almeida. (#12955)

• Add ability to use mne.preprocessing.compute_fine_calibration() with non-Neuromag-style systems, as well as options to control the bad-angle and error tolerances, by Eric Larson. (#12966)

• Add support for uint16_codec argument in mne.io.read_raw_eeglab() when pymatreader (which already supported this argument previously) is not installed, by Clemens Brunner. (#12971)

Other changes

• Use custom code in mne.sys_info() to get the amount of physical memory and a more informative CPU name instead of using the psutil package, by Clemens Brunner. (#12787)

• Improve documentation clarity of fit_transform methods for mne.decoding.SSD, mne.decoding.CSP, and mne.decoding.SPoC classes, by Thomas Binns. (#12827)

• Improve handling of filenames in raw.filenames by using Path instead of str, by Mathieu Scheltienne. (#12843)

• Improve automatic figure scaling of mne.viz.plot_events(), and event_id and count overview legend when a high amount of unique events is supplied, by Stefan Appelhoff. (#12844)

• mne.preprocessing.ICA.find_bads_muscle() can now be run when passing an inst without sensor positions. However, it will just use the first of three criteria (slope) to find muscle-related ICA components, by Stefan Appelhoff. (#12862)

• Improve the Auto-generating Epochs metadata, by Clemens Brunner and Richard Höchenberger. (#12931)

• Mention some gotchas that arise from the fact that by default, we pool across dipole orientations when performing source estimation, by Marijn van Vliet (#12960)