How to change the EMG processing settings#

The "emg" preset runs the EMG processing in this order: band-pass filter, ECG peak detection, ECG gating, envelope, baseline, breath detection and parameters. Each step has a group in the preset configuration:

Group	Step	Settings go to
`"preprocess"`	Band-pass filter and envelope	`ReSurfEMGAdapter.preprocess`
`"ecg_detect_peaks"`	Detection of the R peaks	The step `emg.ecg_detect_peaks`
`"ecg_gating"`	Removal of the ECG at each R peak	The step `emg.ecg_gating`
`"ecg_removal"`	Turn ECG removal off, or give the R peaks	The preset
`"envelope"`	Envelope of the signal after the ECG removal	The step `emg.envelope`
`"baseline"`	Baseline of the envelope	`ReSurfEMGAdapter.moving_baseline`
`"detect_breaths"`	Breath detection	`ReSurfEMGAdapter.detect_breaths`
`"postprocess"`	Parameters and quality checks	`ReSurfEMGAdapter.postprocess`

For example:

session.run_preset(
    "emg",
    config={
        "preprocess": {"channel": 1, "high_pass_hz": 80.0},
        "baseline": {"window_seconds": 7.5, "step_seconds": 0.2},
    },
)

See Workflows for the same steps in a workflow file.

Select the channel#

To select the EMG channel, set "preprocess": {"channel": n}. The channel numbers start at 0.

If you do not set channel, M3Resp uses the channel names. It never uses a channel with the name ECG or EKG as EMG. If the names do not show the breathing muscle channel, M3Resp raises UnresolvedChannelError and asks for channel.

Change the band-pass filter#

The default band-pass filter is 20 Hz to 500 Hz. If the sampling rate is below 1053 Hz, the upper limit is 95% of the Nyquist frequency (half the sampling rate).

To change the lower limit, set "preprocess": {"high_pass_hz": ...}.
To change the upper limit, set "preprocess": {"low_pass_hz": ...}.

Remove mains hum or EIT interference#

A notch filter removes one frequency and its multiples, for example 50 Hz mains hum. An EIT device that records at the same time can also add a frequency and its multiples. To use the notch filter, set "preprocess": {"notch_base_frequency": 50.0}.

Setting	Default	Meaning
`notch_base_frequency`	None (no notch filter)	The lowest frequency to remove, in Hz
`notch_max_frequency`	None (the Nyquist frequency)	The highest multiple to remove, in Hz
`notch_quality_factor`	30.0	Width of each notch: a higher value gives a narrower notch
`notch_before_bandpass`	False	If True, the notch filter runs before the band-pass filter

Change the envelope#

Setting	Default	Meaning
`envelope_method`	`"rms"`	`"rms"`: root mean square. `"arv"`: average rectified value, the method of `resurfemg`’s `full_rolling_arv`. `"median"`: median of the absolute signal, which ignores short spikes such as parts of the ECG that stay after gating.
`envelope_window_seconds`	0.25	Length of the envelope window, in seconds

Put both settings in the "preprocess" group. After the ECG gating, the step emg.envelope calculates the envelope of the gated signal, with the same settings. To give this envelope other settings, set "envelope": {"window_seconds": ..., "method": ...}.

Change the ECG removal#

The preset finds the R peaks in the EMG channel. Then it removes a part of the signal around each R peak and fills the gap.

If you recorded a separate ECG channel, set "ecg_detect_peaks": {"ecg_channel": n}.
If you already have the R peaks (as sample numbers), set "ecg_removal": {"ecg_peak_indices": [...]}. The preset then skips the peak detection. Do not also set "ecg_detect_peaks".
To change the width of the removed part, set "ecg_gating": {"gate_width_seconds": ...}. The default is 205 samples.
To change how M3Resp fills the gap, set "ecg_gating": {"fill_method": n}: 0 fills with zeros, 1 interpolates (default), 2 uses the signal before the gap, 3 uses the RMS of the signal.

Warning

"ecg_removal": {"enabled": False} turns ECG removal off. Then the ECG stays in the envelope, in the breath detection and in all amplitude parameters. Use it only to look at the data.

To remove the ECG with wavelet denoising, use the step emg.ecg_wavelet_denoising in a workflow file. See Remove the ECG: gating or wavelet denoising.

Change the baseline#

The preset calculates the baseline as the 33rd percentile of the envelope in a moving window. Set the window in the "baseline" group:

Setting	Default	Meaning
`window_seconds`	7.5	Length of the window, in seconds
`step_seconds`	0.2	Step of the window, in seconds
`percentile`	33.0	Percentile of the envelope in each window

Without the preset, calculate the baseline with a session method, and give it to the breath detection:

session.preprocess_emg(channel=1)
baseline = session.emg_moving_baseline()  # 7.5 s window, 0.2 s step, 33rd percentile
session.detect_emg_breaths(baseline=baseline)
session.postprocess_emg()

session.emg_slopesum_baseline() gives the slope-sum baseline of resurfemg in place of the moving baseline. Give it to detect_emg_breaths in the same way. Both methods add the baseline to session.signals. Without a baseline, detect_emg_breaths measures the peaks against zero and gives a warning.

Change the breath detection#

Set the breath detection in the "detect_breaths" group:

Setting	Default	Meaning
`min_breath_width_seconds`	0.5	The shortest breath, in seconds
`merge_close_peaks_within_width`	False	If True, M3Resp keeps only the higher peak of two peaks that are closer than `min_breath_width_seconds`

Call another resurfemg function#

The adapter can call 23 functions of resurfemg.postprocessing by name. available_postprocessing() lists them per module:

print(session.emg_adapter.available_postprocessing()["quality_assessment"])
session.emg_adapter.run_postprocessing_function("quality_assessment", "snr_pseudo", *args, **kwargs)