Muscle Artifact

Muscle artifacts generally affect a small portion of the data and can often be excluded. However, due to their large amplitude and broad frequency range, detecting and removing them remains a challenge.

The most common approach for detecting muscle artifacts in MEG data involves annotation during pre-processing. This is accomplished using the annotate_muscle_zscore command from the MNE library. This method works as follows:

1. Frequency Range: Z-scores are calculated for each magnetometer sensor within the typical frequency range of muscle artifacts (110–140 Hz).

2. Aggregation: These z-scores are aggregated and normalized by dividing by the square root of the number of sensors.

3. Thresholding: If the resulting score exceeds a threshold of 15 at any point in the data, that segment is annotated as a muscle artifact.

The parameters for muscle artifact detection in MEGAP are customizable and located in the pipeline_config.cfg file under the "annotate_muscle" section:

"annotate_muscle": {
    "ch_type": "mag",
    "threshold": 15,
    "min_length_good": 1,
    "filter_freq": [110, 140]
}

Magnetometers are specifically selected for detecting muscle artifacts because they are more sensitive to muscle activity compared to other sensor types.

The Z-score example plot for a single subject, saved in the MEGAP results folder

Muscle Duration Warning

The annotation period for muscle artifacts is evaluated to ensure that the duration of the affected segments does not exceed a specified percentage of the total data. This threshold is defined in the "warning" section of the pipeline configuration file.

If the annotated period surpasses this predefined percentage, a warning message is generated. This warning is recorded in a text file located in the /warning folder, alerting the user about the significant presence of muscle artifacts in the data.

If you see a warning for muscle artifacts, first check the muscle z-score plot in the plot_muscle_zscore folder. Then, review the z-score threshold and warning threshold in the configuration file to understand the criteria triggering the warning.