mne.preprocessing.find_bad_channels_maxwell

mne.preprocessing.find_bad_channels_maxwell(raw, limit=7.0, duration=5.0, min_count=5, return_scores=False, origin='auto', int_order=8, ext_order=3, calibration=None, cross_talk=None, coord_frame='head', regularize='in', ignore_ref=False, bad_condition='error', head_pos=None, mag_scale=100.0, skip_by_annotation=('edge', 'bad_acq_skip'), h_freq=40.0, extended_proj=(), verbose=None)[source]

Find bad channels using Maxwell filtering.

Parameters
rawinstance of Raw

Raw data to process.

limitfloat

Detection limit for noisy segments (default is 7.). Smaller values will find more bad channels at increased risk of including good ones. This value can be interpreted as the standard score of differences between the original and Maxwell-filtered data. See the Notes section for details.

Note

This setting only concerns noisy channel detection. The limit for flat channel detection currently cannot be controlled by the user. Flat channel detection is always run before noisy channel detection.

durationfloat

Duration of the segments into which to slice the data for processing, in seconds. Default is 5.

min_countint

Minimum number of times a channel must show up as bad in a chunk. Default is 5.

return_scoresbool

If True, return a dictionary with scoring information for each evaluated segment of the data. Default is False.

Warning

This feature is experimental and may change in a future version of MNE-Python without prior notice. Please report any problems and enhancement proposals to the developers.

New in version 0.21.

originarray_like, shape (3,) | str

Origin of internal and external multipolar moment space in meters. The default is 'auto', which means (0., 0., 0.) when coord_frame='meg', and a head-digitization-based origin fit using fit_sphere_to_headshape() when coord_frame='head'. If automatic fitting fails (e.g., due to having too few digitization points), consider separately calling the fitting function with different options or specifying the origin manually.

int_orderint

Order of internal component of spherical expansion.

ext_orderint

Order of external component of spherical expansion.

calibrationstr | None

Path to the '.dat' file with fine calibration coefficients. File can have 1D or 3D gradiometer imbalance correction. This file is machine/site-specific.

cross_talkstr | None

Path to the FIF file with cross-talk correction information.

coord_framestr

The coordinate frame that the origin is specified in, either 'meg' or 'head'. For empty-room recordings that do not have a head<->meg transform info['dev_head_t'], the MEG coordinate frame should be used.

regularizestr | None

Basis regularization type, must be “in” or None. “in” is the same algorithm as the “-regularize in” option in MaxFilter™.

ignore_refbool

If True, do not include reference channels in compensation. This option should be True for KIT files, since Maxwell filtering with reference channels is not currently supported.

bad_conditionstr

How to deal with ill-conditioned SSS matrices. Can be “error” (default), “warning”, “info”, or “ignore”.

head_posarray | None

If array, movement compensation will be performed. The array should be of shape (N, 10), holding the position parameters as returned by e.g. read_head_pos.

mag_scalefloat | str

The magenetometer scale-factor used to bring the magnetometers to approximately the same order of magnitude as the gradiometers (default 100.), as they have different units (T vs T/m). Can be 'auto' to use the reciprocal of the physical distance between the gradiometer pickup loops (e.g., 0.0168 m yields 59.5 for VectorView).

skip_by_annotationstr | list of str

If a string (or list of str), any annotation segment that begins with the given string will not be included in filtering, and segments on either side of the given excluded annotated segment will be filtered separately (i.e., as independent signals). The default ('edge', 'bad_acq_skip') will separately filter any segments that were concatenated by mne.concatenate_raws() or mne.io.Raw.append(), or separated during acquisition. To disable, provide an empty list.

h_freqfloat | None

The cutoff frequency (in Hz) of the low-pass filter that will be applied before processing the data. This defaults to 40., which should provide similar results to MaxFilter. If you do not wish to apply a filter, set this to None.

extended_projlist

The empty-room projection vectors used to extend the external SSS basis (i.e., use eSSS).

New in version 0.21.

verbosebool, str, int, or None

If not None, override default verbose level (see mne.verbose() and Logging documentation for more). If used, it should be passed as a keyword-argument only.

Returns
noisy_chslist

List of bad MEG channels that were automatically detected as being noisy among the good MEG channels.

flat_chslist

List of MEG channels that were detected as being flat in at least min_count segments.

scoresdict

A dictionary with information produced by the scoring algorithms. Only returned when return_scores is True. It contains the following keys:

  • ch_namesndarray, shape (n_meg,)

    The names of the MEG channels. Their order corresponds to the order of rows in the scores and limits arrays.

  • ch_typesndarray, shape (n_meg,)

    The types of the MEG channels in ch_names ('mag', 'grad').

  • binsndarray, shape (n_windows, 2)

    The inclusive window boundaries (start and stop; in seconds) used to calculate the scores.

  • scores_flatndarray, shape (n_meg, n_windows)

    The scores for testing whether MEG channels are flat. These values correspond to the standard deviation of a segment. See the Notes section for details.

  • limits_flatndarray, shape (n_meg, 1)

    The score thresholds (in standard deviation) above which a segment was classified as “flat”.

  • scores_noisyndarray, shape (n_meg, n_windows)

    The scores for testing whether MEG channels are noisy. These values correspond to the standard score of a segment. See the Notes section for details.

  • limits_noisyndarray, shape (n_meg, 1)

    The score thresholds (in standard scores) above which a segment was classified as “noisy”.

Note

The scores and limits for channels marked as bad in the input data will be set to np.nan.

Notes

All arguments after raw, limit, duration, min_count, and return_scores are the same as maxwell_filter(), except that the following are not allowed in this function because they are unused: st_duration, st_correlation, destination, st_fixed, and st_only.

This algorithm, for a given chunk of data:

  1. Runs SSS on the data, without removing external components.

  2. Excludes channels as flat that have had low variability (standard deviation < 0.01 fT or fT/cm in a 30 ms window) in the given or any previous chunk.

  3. For each channel \(k\), computes the range or peak-to-peak \(d_k\) of the difference between the reconstructed and original data.

  4. Computes the average \(\mu_d\) and standard deviation \(\sigma_d\) of the differences (after scaling magnetometer data to roughly match the scale of the gradiometer data using mag_scale).

  5. Marks channels as bad for the chunk when \(d_k > \mu_d + \textrm{limit} \times \sigma_d\). Note that this expression can be easily transformed into \((d_k - \mu_d) / \sigma_d > \textrm{limit}\), which is equivalent to \(z(d_k) > \textrm{limit}\), with \(z(d_k)\) being the standard or z-score of the difference.

Data are processed in chunks of the given duration, and channels that are bad for at least min_count chunks are returned.

Channels marked as flat in step 2 are excluded from all subsequent steps of noisy channel detection.

This algorithm gives results similar to, but not identical with, MaxFilter. Differences arise because MaxFilter processes on a buffer-by-buffer basis (using buffer-size-dependent downsampling logic), uses different filtering characteristics, and possibly other factors. Channels that are near the limit for a given min_count are particularly susceptible to being different between the two implementations.

New in version 0.20.

Examples using mne.preprocessing.find_bad_channels_maxwell