1.7.0 (2024-04-19)

Notable changes

• In this version, we started adding type hints (also known as “type annotations”) to select parts of the codebase. This meta information will be used by development environments (IDEs) like VS Code and PyCharm automatically to provide better assistance such as tab completion or error detection even before running your code.

  So far, we’ve only added return type hints to mne.io.read_raw(), mne.read_epochs(), mne.read_evokeds() and all format-specific read_raw_*() and read_epochs_*() functions. Now your editors will know: these functions return evoked and raw data, respectively. We are planning add type hints to more functions after careful evaluation in the future.

  You don’t need to do anything to benefit from these changes – your editor will pick them up automatically and provide the enhanced experience if it supports it! (#12250)

Dependencies

• defusedxml is now an optional (rather than required) dependency and needed when reading EGI-MFF data, NEDF data, and BrainVision montages, by Eric Larson. (#12264)

• For developers, pytest>=8.0 is now required for running unit tests, by Eric Larson. (#12376)

• pytest-harvest is no longer used as a test dependency, by Eric Larson. (#12451)

• The minimum supported version of Qt bindings is 5.15, by Eric Larson. (#12491)

Bugfixes

• Fix bug where section parameter in mne.Report.add_html() was not being utilized resulting in improper formatting, by new contributor Martin Oberg. (#12319)

• Fix bug in mne.preprocessing.maxwell_filter() where calibration was incorrectly applied during virtual sensor reconstruction, by Eric Larson and new contributor Motofumi Fushimi. (#12348)

• Reformats channel and detector lookup in mne.io.read_raw_snirf() from array based to dictionary based. Removes incorrect assertions that every detector and source must have data associated with every registered optode position, by new contributor Alex Kiefer. (#12430)

• Remove FDT file format check for strings in EEGLAB’s EEG.data in mne.io.read_raw_eeglab() and related functions by new contributor Seyed Yahya Shirazi (#12523)

• Fixes to interactivity in time-frequency objects: the rectangle selector now works on TFR image plots of gradiometer data; and in TFR.plot_joint() plots, the colormap limits of interactively-generated topomaps match the colormap limits of the main plot. By Daniel McCloy. (#11282)

• Allow mne.viz.plot_compare_evokeds() to plot eyetracking channels, and improve error handling, y Scott Huberty. (#12190)

• Fix bug in mne.Epochs.apply_function() where data was handed down incorrectly in parallel processing, by Dominik Welke. (#12206)

• Remove incorrect type hints in mne.io.read_raw_neuralynx(), by Richard Höchenberger. (#12236)

• Fix bug with accessing the last data sample using raw[:, -1] where an empty array was returned, by Eric Larson. (#12248)

• Correctly handle temporal gaps in Neuralynx .ncs files via mne.io.read_raw_neuralynx(), by Kristijan Armeni and Eric Larson. (#12279)

• Fix bug where parent directory existence was not checked properly in mne.io.Raw.save(), by Eric Larson. (#12282)

• Add tol parameter to mne.events_from_annotations() so that the user can specify the tolerance to ignore rounding errors of event onsets when using chunk_duration is not None (default is 1e-8), by Michiru Kaneda (#12324)

• Allow mne.io.Raw.interpolate_bads() and mne.Epochs.interpolate_bads() to work on ecog and seeg data; for seeg data a spline is fit to neighboring electrode contacts on the same shaft, by Alex Rockhill (#12336)

• Fix clicking on an axis of mne.viz.plot_evoked_topo() when multiple vertical lines vlines are used, by Mathieu Scheltienne. (#12345)

• Fix bug in mne.viz.EvokedField.set_vmax() that prevented setting the color limits of the MEG magnetic field density, by Marijn van Vliet (#12354)

• Fix faulty indexing in mne.io.read_raw_neuralynx() when picking a single channel, by Kristijan Armeni. (#12357)

• Fix bug where mne.preprocessing.compute_proj_ecg() and mne.preprocessing.compute_proj_eog() could modify the default reject and flat arguments on multiple calls based on channel types present, by Eric Larson. (#12380)

• Fix bad channels not handled properly in mne.stc_near_sensors() by Alex Rockhill. (#12382)

• Fix bug where mne.preprocessing.regress_artifact() projection check was not specific to the channels being processed, by Eric Larson. (#12389)

• Change how samples are read when using data_format='auto' in mne.io.read_raw_cnt(), by Jacob Woessner. (#12393)

• Fix bugs with mne.Report CSS where TOC items could disappear at the bottom of the page, by Eric Larson. (#12399)

• In plot_compare_evokeds(), actually plot GFP (not RMS amplitude) for EEG channels when global field power is requested by Daniel McCloy. (#12410)

• Fix Working with sEEG data use of mne.stc_near_sensors() to use the mne.VolSourceEstimate positions and not the pial surface, by Alex Rockhill (#12436)

• Fix prefiltering information management for EDF/BDF, by Michiru Kaneda (#12441)

• Fix validation of ch_type in mne.preprocessing.annotate_muscle_zscore(), by Mathieu Scheltienne. (#12444)

• Fix errant redundant use of BIDSPath.split when writing split raw and epochs data, by Eric Larson. (#12451)

• Disable config parser interpolation when reading BrainVision files, which allows using the percent sign as a regular character in channel units, by Clemens Brunner. (#12456)

• • Fix the default color of mne.viz.Brain.add_text() to properly contrast with the figure background color, by Marijn van Vliet. (#12470)

• • Changed default ECoG and sEEG electrode sizes in brain plots to better reflect real world sizes, by Liberty Hamilton (#12474)

• Fixed bugs with handling of rank in mne.decoding.CSP, by Eric Larson. (#12476)

• • Fix reading segmented recordings with mne.io.read_raw_eyelink() by Dominik Welke. (#12481)

• Improve compatibility with other Qt-based GUIs by handling theme icons better, by Eric Larson. (#12483)

• • Fix problem caused by onsets with NaN values using mne.io.read_raw_eeglab() by Jacob Woessner (#12484)

• Fix cleaning of channel names for non vectorview or CTF dataset including whitespaces or dash in their channel names, by Mathieu Scheltienne. (#12489)

• Fix bug with mne.preprocessing.ICA.plot_sources() for evoked data where the legend contained too many entries, by Eric Larson. (#12498)

• Fix bug where using phase="minimum" in filtering functions like mne.io.Raw.filter() constructed a filter half the desired length with compromised attenuation. Now phase="minimum" has the same length and comparable suppression as phase="zero", and the old (incorrect) behavior can be achieved with phase="minimum-half", by Eric Larson. (#12507)

• Correct reading of info["subject_info"]["his_id"] in mne.io.read_raw_snirf(), by Eric Larson. (#12526)

• Calling compute_psd() with method="multitaper" is now expressly disallowed when reject_by_annotation=True and bad_* annotations are present (previously this was nominally allowed but resulted in nan values in the PSD). By Daniel McCloy. (#12535)

• compute_psd() and psd_array_welch() will now use FFT windows aligned to the onsets of good data spans when bad_* annotations are present. By Daniel McCloy. (#12536)

• Fix bug in loading of complex/phase TFRs. By Daniel McCloy. (#12537)

• Fix bug with mne.SourceSpaces.export_volume() where the img.affine was not set properly, by Eric Larson. (#12544)

API changes by deprecation

• The default value of the zero_mean parameter of mne.time_frequency.tfr_array_morlet() will change from False to True in version 1.8, for consistency with related functions. By Daniel McCloy. (#11282)

• The parameter for providing data to mne.time_frequency.tfr_array_morlet() and mne.time_frequency.tfr_array_multitaper() has been switched from epoch_data to data. Only use the data parameter to avoid a warning. Changes by Thomas Binns. (#12308)

• Change mne.stc_near_sensors() surface default from the 'pial' surface to the surface in src if src is not None in version 1.8, by Alex Rockhill. (#12382)

New features

• Detecting Bad EEG/MEG channels using the local outlier factor (LOF) algorithm in mne.preprocessing.find_bad_channels_lof(), by new contributor Velu Prabhakar Kumaravel. (#11234)

• Inform the user about channel discrepancy between provided info, forward operator, and/or covariance matrices in mne.beamformer.make_lcmv(), by new contributor Nikolai Kapralov. (#12238)

• Support partial pathlength factors for each wavelength in mne.preprocessing.nirs.beer_lambert_law(), by new contributor Richard Scholz. (#12446)

• Add picks parameter to mne.io.Raw.plot(), allowing users to select which channels to plot. This makes makes the raw data plotting API consistent with mne.Epochs.plot() and mne.Evoked.plot(), by new contributor Ivo de Jong. (#12467)

• New class mne.time_frequency.RawTFR and new methods mne.io.Raw.compute_tfr(), mne.Epochs.compute_tfr(), and mne.Evoked.compute_tfr(). These new methods supersede functions mne.time_frequency.tfr_morlet(), and mne.time_frequency.tfr_multitaper(), and mne.time_frequency.tfr_stockwell(), which are now considered “legacy” functions. By Daniel McCloy. (#11282)

• Add ability reject mne.Epochs using callables, by Jacob Woessner. (#12195)

• Custom functions applied via mne.io.Raw.apply_function(), mne.Epochs.apply_function() or mne.Evoked.apply_function() can now use ch_idx or ch_name to get access to the currently processed channel during channel wise processing.

• mne.Evoked.apply_function() can now also work on full data array instead of just channel wise, analogous to mne.io.Raw.apply_function() and mne.Epochs.apply_function(), by Dominik Welke. (#12206)

• Allow mne.time_frequency.EpochsTFR as input to mne.epochs.equalize_epoch_counts(), by Carina Forster. (#12207)

• Speed up export to .edf in mne.export.export_raw() by using edfio instead of EDFlib-Python. (#12218)

• Added a helper function mne.preprocessing.eyetracking.convert_units() to convert eyegaze data from pixel-on-screen values to radians of visual angle. Also added a helper function mne.preprocessing.eyetracking.get_screen_visual_angle() to get the visual angle that the participant screen subtends, by Scott Huberty. (#12237)

• We added type hints for the return values of mne.read_evokeds() and mne.io.read_raw(). Development environments like VS Code or PyCharm will now provide more help when using these functions in your code. By Richard Höchenberger and Eric Larson. (#12297) (#12250)

• Add method="polyphase" to mne.io.Raw.resample() and related functions to allow resampling using scipy.signal.upfirdn(), by Eric Larson. (#12268)

• The package build backend was switched from setuptools to hatchling. This will only affect users who build and install MNE-Python from source. By Richard Höchenberger. (#12281) (#12269)

• mne.Annotations.to_data_frame() can now output different formats for the onset column: seconds, milliseconds, datetime objects, and timedelta objects. By Daniel McCloy. (#12289)

• Add method mne.SourceEstimate.save_as_surface() to allow saving GIFTI files from surface source estimates, by Peter Molfese. (#12309)

• mne.Epochs can now be constructed using mne.Annotations stored in the raw object, by specifying events=None. By Alex Rockhill. (#12311)

• Add savgol_filter(), filter(), apply_hilbert(), and apply_function() methods to mne.SourceEstimate and related classes, by Hamza Abdelhedi. (#12323)

• Add ability to export STIM channels to EDF in mne.io.Raw.export(), by Clemens Brunner. (#12332)

• Speed up raw FIF reading when using small buffer sizes by Eric Larson. (#12343)

• Speed up mne.io.read_raw_neuralynx() on large datasets with many gaps, by Kristijan Armeni. (#12371)

• Add ability to detect minima peaks found in mne.Evoked if data is all positive and maxima if data is all negative. (#12383)

• Add ability to remove bad marker coils in mne.io.read_raw_kit(), by Judy D Zhu. (#12394)

• Add option to pass image_kwargs to mne.Report.add_epochs to allow adjusting e.g. vmin and vmax of the epochs image in the report, by Sophie Herbst. (#12443)

• Add support for multiple raw instances in mne.preprocessing.compute_average_dev_head_t() by Eric Larson. (#12445)

• Completing PR 12453. Add option to pass image_kwargs per channel type to mne.Report.add_epochs. (#12454)

• mne.epochs.make_metadata() now accepts strings as tmin and tmax parameter values, simplifying metadata creation based on time-varying events such as responses to a stimulus, by Richard Höchenberger. (#12462)

• Include date of acquisition and filter parameters in raw.info for mne.io.read_raw_neuralynx() by Kristijan Armeni. (#12463)

• Add physical_range="channelwise" to mne.io.Raw.export() for exporting to EDF, which can improve amplitude resolution if individual channels vary greatly in their offsets, by Clemens Brunner. (#12510)

• Added the ability to reorder report contents via mne.Report.reorder() (with helper to get contents with mne.Report.get_contents()), by Eric Larson. (#12513)

• Add exclude_after_unique option to mne.io.read_raw_edf() and mne.io.read_raw_edf() to search for exclude channels after making channels names unique, by Michiru Kaneda (#12518)

Other changes

• Updated the text in the preprocessing tutorial to use mne.io.Raw.pick() instead of the legacy mne.io.Raw.pick_types(), by new contributor btkcodedev. (#12326)

• Clarify in the EEG referencing tutorial that an average reference projector ready is required for inverse modeling, by new contributor Nabil Alibou (#12420)

• Fix dead links in README.rst documentation by new contributor Will Turner. (#12461)

• Replacing percent format with f-strings format specifiers , by new contributor Hasrat Ali Arzoo. (#12464)

• Adopted towncrier for changelog entries, by Eric Larson. (#12299)

• Automate adding of PR number to towncrier stubs, by Eric Larson. (#12318)

• Refresh code base to use Python 3.9 syntax using Ruff UP rules (pyupgrade), by Clemens Brunner. (#12358)

• Move private data preparation functions for BrainVision export from pybv to mne, by Clemens Brunner. (#12450)

• Update the list of sensor types in docstrings, tutorials and the glossary by Nabil Alibou. (#12509)

Authors

• Alex Rockhill

• Alexander Kiefer+

• Alexandre Gramfort

• Britta Westner

• Carina Forster

• Clemens Brunner

• Daniel McCloy

• Dominik Welke

• Eric Larson

• Erkka Heinila

• Florian Hofer

• Hamza Abdelhedi

• Hasrat Ali Arzoo+

• Ivo de Jong+

• Jacob Woessner

• Judy D Zhu

• Kristijan Armeni

• Liberty Hamilton

• Marijn van Vliet

• Martin Oberg+

• Mathieu Scheltienne

• Michiru Kaneda

• Motofumi Fushimi+

• Nabil Alibou+

• Nikolai Kapralov+

• Peter J. Molfese

• Richard Höchenberger

• Richard Scholz+

• Scott Huberty

• Seyed (Yahya) Shirazi+

• Sophie Herbst

• Stefan Appelhoff

• Thomas Donoghue

• Thomas Samuel Binns

• Tristan Stenner

• Velu Prabhakar Kumaravel+

• Will Turner+

• btkcodedev+