m3resp.Signal

class m3resp.Signal(values, time, sample_frequency=None, unit=None, name=None, metadata=<factory>, modality='unknown', category=None, channel=None, source=None, processing_state='raw', derived_from=None, method=None)

A TimeSeries tagged with modality and provenance context.

Provenance fields (source vs method are distinct and often confused):

• modality: which device/technique produced this - "eit", "emg", "ventilator". Orthogonal to category.

• category: what physical quantity these numbers are - an impedance, an airway pressure, a flow. Known aliases are canonicalized via m3resp.data.categories.normalize_category(); an unrecognized value is kept as-is, since the vocabulary is open. None means unspecified.

• channel: which physical/logical channel this is (e.g. an EMG lead name, "global_impedance" for EIT).

• source: where the data came from - the upstream producer/origin, typically the loader or library that emitted it (e.g. "resurfemg", "eitprocessing"). Provenance of origin, not of transformation.

• method: what was done to it - the specific algorithm/function that produced this particular signal (e.g. "resurfemg.postprocessing.moving_baseline"). This is what distinguishes two signals sharing the same channel and processing_state (e.g. two differently-filtered variants).

• processing_state: how far along the raw -> intermediate -> processed workflow this signal is, or "derived" if computed from another signal. See ProcessingState.

• derived_from: for a "derived" signal, the processing_state it was computed from (e.g. a baseline derived from the "processed" envelope has derived_from="processed"). None otherwise.

Methods

to_manifest_row

Lightweight, JSON-serializable metadata row (no raw sample arrays).

Attributes

category
channel
derived_from
method
modality
processing_state
source
values
time
metadata

Parameters:

• values (ndarray)

• time (ndarray)

• sample_frequency (float | None)

• unit (str | None)

• name (str | None)

• metadata (dict[str, Any])

• modality (Literal['eit', 'emg', 'ventilator', 'monitor', 'unknown'] | str)

• category (Literal['impedance', 'electrical_potential', 'airway_pressure', 'esophageal_pressure', 'gastric_pressure', 'transpulmonary_pressure', 'airflow', 'volume', 'tidal_volume'] | str | None)

• channel (str | None)

• source (str | None)

• processing_state (Literal['raw', 'intermediate', 'processed', 'derived'])

• derived_from (Literal['raw', 'intermediate', 'processed', 'derived'] | None)

• method (str | None)

category: Literal['impedance', 'electrical_potential', 'airway_pressure', 'esophageal_pressure', 'gastric_pressure', 'transpulmonary_pressure', 'airflow', 'volume', 'tidal_volume'] | str | None = None

channel: str | None = None

derived_from: Literal['raw', 'intermediate', 'processed', 'derived'] | None = None

method: str | None = None

modality: Literal['eit', 'emg', 'ventilator', 'monitor', 'unknown'] | str = 'unknown'

processing_state: Literal['raw', 'intermediate', 'processed', 'derived'] = 'raw'

source: str | None = None

to_manifest_row()

Lightweight, JSON-serializable metadata row (no raw sample arrays).

Matches the data model doc’s separation of raw arrays (kept in files) from relational/manifest metadata (Sec 8): a signals_manifest.csv/.json row should describe a stream, not embed it.

Return type:

dict[str, object]

Tutorials that use Signal

Process an EIT recording

Process an EIT recording

Process an EMG recording

Process an EMG recording