mne-denoise

Denoising methods for EEG and MEG

mne-denoise brings together complementary spatial, spectral, adaptive, and source-informed denoising methods, with NumPy support and integration with MNE-Python objects.

Choose a method based on the artifact and the information available in your recording, and evaluate both what was removed and what was preserved.

Get started

Choose a method

Browse examples

pip

pip install mne-denoise

uv

uv pip install mne-denoise

conda

conda install -c conda-forge mne-denoise

Development status

mne-denoise is under active development. Until version 1.0, the public API may evolve between releases. For reproducible analyses, record the mne-denoise version used in your work.

What do you need to clean?

A useful place to start is the type of artifact or structure you want to address.

Transient high-amplitude artifacts

Reconstruct high-variance periods using a relatively clean calibration reference.

Methods: ASR

Sensor-specific noise

Suppress sensor-specific noise using spatial redundancy or forward-model information.

Methods: SNS · SOUND

Power-line contamination

Address narrowband line contamination with spectral interpolation or DSS-based spatial separation.

Methods: Spectrum interpolation · ZapLine

Reproducible or structured components

Extract or suppress components defined by reproducibility, temporal or spectral structure, or repeated trials.

Methods: DSS · SSA

CCA and reference-informed cleaning

Use shared structure between signals, lagged copies, or reference channels to identify components for cleaning.

Methods: BSS-CCA · iCanClean

TMS-evoked muscle artifact

Suppress a fitted artifact subspace and reconstruct the signal with source-informed geometry.

Method: SSP-SIR

Using mne-denoise in research?

Please cite the software version you used and the primary scientific paper for each method.

Citation · GitHub