gastropy.remove_movement_artifacts

gastropy.remove_movement_artifacts#

gastropy.remove_movement_artifacts(data, sfreq, freq=0.05, window=1.0)[source]#

Attenuate movement artifacts using an LMMSE Wiener filter.

Estimates local signal variance within sliding windows and computes a Wiener-like (LMMSE) correction that suppresses time segments dominated by movement noise while preserving the gastric rhythm.

The algorithm:

  1. For each sample, compute the local mean E[y] and variance var_y in a window of length window / freq seconds.

  2. Estimate the signal-of-interest variance var_e as the mean of all local variances.

  3. Compute the predicted noise contribution:

    x_hat = E[y] + clip(var_y - var_e, 0) / max(var_y, var_e)
        * (y - E[y])

  4. Return the residual y - x_hat.

Parameters:

  • data (array_like) – EGG signal(s). Accepts shape (n_samples,) or (n_channels, n_samples).

  • sfreq (float) – Sampling frequency in Hz.

  • freq (float, optional) – Centre frequency of interest in Hz. Used to set the window length (window / freq seconds). Default is 0.05 Hz (normogastric centre, 3 cpm).

  • window (float, optional) – Window length in cycles of freq. Default is 1.0.

Returns:

cleaned – Movement-corrected signal. Same shape as input.

Return type:

np.ndarray

Notes

The filter always subtracts the estimated noise component x_hat, which contains the local mean E[y]. As a result the output is effectively mean-centered within each sliding window, and any DC offset in the original signal will be removed. A constant (zero-variance) signal maps to all zeros. This is the expected behaviour for EGG preprocessing, where a zero-mean signal is assumed, but should be noted when applying to signals with a meaningful baseline.

References

Gharibans, A. A., Smarr, B., Kunkel, D. C., Kriegsfeld, L. J., Mousa, H., & Coleman, T. P. (2018). Artifact rejection methodology enables continuous, noninvasive measurement of gastric myoelectric activity in ambulatory subjects. Scientific Reports, 8, 5019. https://doi.org/10.1038/s41598-018-23302-9

Dalmaijer, E. S. (2025). electrography v1.1.1. esdalmaijer/electrography

Examples

>>> import numpy as np
>>> from gastropy.signal import remove_movement_artifacts
>>> t = np.arange(0, 300, 0.1)
>>> sig = np.sin(2 * np.pi * 0.05 * t)
>>> cleaned = remove_movement_artifacts(sig, sfreq=10.0)
>>> cleaned.shape == sig.shape
True
Contents