Signals#
A signal is one time series with a sampling rate and a unit, for example the global impedance of an EIT recording or the envelope of an EMG channel. M3Resp stores each signal as a Signal object in session.signals.
Look at a signal#
signal = session.signals.for_modality("eit")[0]
print(signal.name) # Global impedance (raw)
print(signal.sample_frequency) # 20.0
print(signal.unit) # a.u.
print(signal.values.shape) # (1200,)
print(signal.time[:3]) # [0. 0.05 0.1 ]
What a signal holds#
Field |
Contents |
|---|---|
|
The measured values, as a NumPy array |
|
The time of each sample on the clock of its recording, in seconds |
|
The sampling rate, in Hz |
|
The unit of the values, for example |
|
The name of the signal |
|
The device that recorded it: |
|
The physical quantity, for example |
|
The channel of the device, for example an EMG channel name |
|
The package or loader that made the signal, for example |
|
How far the processing is: |
|
For a derived signal: the processing state of the signals that it was calculated from |
|
The calculation that made the signal |
|
Other details, as a dictionary |
The first axis of values is time, so its length is equal to the length of time. values can have more axes. For example, the EIT pixel data has one 32 × 32 image per time point: time × rows × columns.
Signal adds the fields from modality to method to the basic time series, TimeSeries. A TimeSeries has only values, time, sample_frequency, unit, name and metadata.
Units#
M3Resp writes each unit in one form, so units from different devices can be compared. For example, "uV" becomes "µV":
import numpy as np
from m3resp import Signal
signal = Signal(values=np.zeros(3), time=np.arange(3) / 10, sample_frequency=10.0, unit="uV")
print(signal.unit) # µV
To add a spelling of your own, use register_unit_alias in m3resp.data.units. To read a list of spellings from a YAML or JSON file, use load_unit_aliases.
Modality and category#
The modality is the device that recorded the signal. The category is the physical quantity. These are two separate fields, because one device records several quantities, and one quantity can come from several devices:
A ventilator records airway pressure, flow and volume.
Airway pressure can come from a ventilator, from a separate pressure monitor or from an EIT device.
Airway pressure from an EIT device has modality="eit" and category="airway_pressure". You can search on each field:
session.signals.for_modality("ventilator") # all ventilator signals
session.signals.for_category("airway_pressure") # airway pressure from all devices
for_category is also available on session.parameter_results and session.quality.
Both fields accept any name. M3Resp changes a known other name of a category to its main name: category="paw" is stored as "airway_pressure", and for_category("paw") also works. M3Resp keeps an unknown name as it is. To add a name of your own, use register_category_alias or load_category_aliases in m3resp.data.categories.
Processing state#
|
Meaning |
|---|---|
|
The signal as the device or the loader gave it |
|
After one or more processing steps, for example a filter or a slice to a time window. More steps follow. |
|
The final signal of a channel, ready for the calculation of parameters |
|
Calculated from other signals, for example the difference of two signals. |
Two signals can have the same channel and the same processing state, for example two filtered signals that use different filters. The field method then tells them apart. M3Resp writes method as "<package>.<function>", for example "resurfemg.moving_baseline" or "eitprocessing.RateDetection".
Where signals come from#
preprocess_eit, preprocess_emg and preprocess_ventilator add signals to session.signals. Many workflow steps also add signals, for example emg.ecg_gating adds the EMG signal after ECG removal.
Continue with Events, intervals and breaths.