Version 1.11.0.dev109+g7ed5e27dc (development) (2025-11-12)

Dependencies

• New reader for Neuroscan Curry files, using the curry-python-reader module, by Dominik Welke. (#13176)

Bugfixes

• Handle scenario where an Eyelink recording switched from binocular to monocular mode during a trial by Scott Huberty (#12847)

• Fixed ICA getting sources for concatenated raw instances, by new contributor Beige Jin. (#13068)

• Fix reading annotations with mne.read_annotations() from .csv files containing nanoseconds in times, and drop nanoseconds from times returned from mne.Annotations.to_data_frame() and saved in .csv files by mne.Annotations.save(), by Thomas Binns. (#13109)

• Fix bug with mne.io.read_raw_egi() where info["dev_head_t"] was an identity matrix instead of None, by Eric Larson. (#13112)

• Read Nihon Kohden annotation file accurately (using sublog parsing), by Tom Ma. (#13251)

• Fix bug with mne.open_report() not working with old saved files, by Eric Larson. (#13341)

• Restore clipping=3 in plot() (which was mistakenly changed to 1.5) to reduce excessive clipping of raw data, by Clemens Brunner. (#13354)

• Make read_raw_eyelink() work with ASCII files collected with Eye Event Mode set to “HREF” by Scott Huberty. (#13357)

• model parameter of mne.decoding.LinearModel will not be modified, use model_ attribute to access the fitted model. To be compatible with all MNE-Python versions you can use getattr(clf, "model_", getattr(clf, "model")) The provided model is expected to be a supervised predictor, i.e. classifier or regressor (or sklearn.multiclass.OneVsRestClassifier), otherwise an error will be raised. by Gennadiy Belonosov. (#13361)

• Fix the check in mne.make_forward_solution() that all MEG sensors are outside a spherical BEM model, by Marijn van Vliet (#13371)

• Fix bug with mne.grand_average() not working with mne.time_frequency.Spectrum objects, by Thomas Binns. (#13375)

• Ensure auto-computed plot scalings are always non-zero, by Daniel McCloy. (#13376)

• Fix plot_evoked_joint() docstring to correctly document exclude parameter default as 'bads' instead of None, by new contributor Tharupahan Jayawardana. (#13391)

• Make decoding classes sklearn-compliant, by Gennadiy Belonosov. (#13393)

• Fix loading problems caused by attempting to iterate floats in read_epochs_eeglab() for specific EEGLab data, by new contributor Emrecan Çelik. (#13395)

• Fix bug with mne.make_forward_solution() where sources were not checked to make sure they’re inside the inner skull for spherical BEMs, by Eric Larson. (#13407)

• Preserve event-to-epoch mapping when exporting EEGLAB .set files by Scott Huberty (#13428)

• Fix and update osf.io links to new format, by new contributor Johannes Herforth. (#13440)

• Fix bug with mne.add_reference_channels() not working correctly when passing more than one channel name, by Michael Straube. (#13466)

• Correctly set the calibration factor in Nihon Kohden reader (which affects channel amplitudes), by Tom Ma. (#13468)

• Make mne.preprocessing.eyetracking.read_eyelink_calibration() robust to files with blank lines, by Scott Huberty. (#13469)

• Fix bug with mne.viz.plot_topomap() that caused unexpected channel name plotting when using a mask, by Michael Straube. (#13470)

• Fix bug with mne.preprocessing.ICA.apply() not working when the ICA instance was fitted including marked bad channels, by Michael Straube. (#13478)

API changes by deprecation

• Add on_few_samples parameter to mne.compute_covariance() and mne.compute_raw_covariance() for controlling behavior when there are fewer samples than channels, which can lead to inaccurate covariance estimates, by new contributor Emmanuel Ferdman. (#13350)

• The default for mne.make_field_map() will change to "auto" in MNE-Python 1.12 (from (0., 0., 0.04)), changes by new contributor Paul Anders. (#13398)

• Add encoding parameter to mne.io.read_raw_nihon() for better handling of annotation decoding, by Tom Ma. (#13458)

New features

• Support for preloading=False when reading .set files, by Bruno Aristimunha. (#13096)

• Added support for file like objects in read_raw_bdf, read_raw_edf and read_raw_gdf, by new contributor Santi Martínez. (#13156)

• Read impedances and montage from Neuroscan Curry files, by Dominik Welke. (#13176)

• Implement GEDTransformer superclass that generalizes mne.decoding.CSP, mne.decoding.SPoC, mne.decoding.XdawnTransformer, mne.decoding.SSD and fix related bugs and inconsistencies, by Gennadiy Belonosov. (#13259)

• Added on_inside="raise" parameter to mne.make_forward_solution() and mne.make_forward_dipole() to control behavior when MEG sensors are inside the outer skin surface. This is useful for forward solutions that are computed with sensors just inside the outer skin surface (e.g., with some OPM coregistrations), by Eric Larson. (#13307)

• Implement mne.decoding.SpatialFilter class returned by mne.decoding.get_spatial_filter_from_estimator() for visualisation of filters and patterns for mne.decoding.LinearModel and additionally eigenvalues for GED-based transformers such as mne.decoding.XdawnTransformer, mne.decoding.CSP, by Gennadiy Belonosov. (#13332)

• Add more options for the sphere parameter of mne.viz.plot_sensors(), by Marijn van Vliet (#13400)

• Add support for BDF export in mne.export.export_raw(), by Clemens Brunner (#13435)

• Improve precision of fiducial clicking in mne coreg, by Eric Larson. (#13445)

• Add support for Nihon Kohden EEG-1200A V01.00, by Eric Larson. (#13448)

• The rename_channels method now has an on_missing parameter to control behavior on channel mismatch, by Stefan Appelhoff. (#13456)

• Added the annotation_regex parameter to mne.viz.plot_raw() and mne.viz.plot_ica_sources(), allowing automatic hiding of annotations that do not match the specified regular expression. The same functionality is available via mne.io.Raw.plot() and mne.preprocessing.ICA.plot_sources(). Contributed by Johannes Herforth. (#13460)

• The combine_channels method now has an on_missing parameter to control behavior on missing event ids, by new contributor Michael Straube. (#13463)

Other changes

• Changed “enhancement” to “new feature” in the contributing guide as enhancement is not a recognized change type, by Shristi Baral (#13373)

• Add a SPEC0-like policy in which minimum-supported versions for select dependencies are set to the latest minor release that was available 2 years prior, by Thomas Binns (#13451)

• Removed development dependencies from user-visible “extras”; they’re now dependency groups only visible to developers (and can be installed for example via pip install --group dev with pip version 25.1 or later), by Richard Höchenberger (#13452)