Cut a recording

Parts of a recording are described by time intervals (Segment), in seconds from the start of the file. There are three kinds:

Kind	What it does	Example
exclusion	its breaths are never analysed (movement, suction, recalibration)	Segment(600, 640, label="movement")
phase	a part with its own protocol, analysed on its own	Segment(0, 290, "phase", "tidal")
PEEP step placed by hand	the breaths of a step and its PEEP level, when there is no pressure signal or to choose the breaths	Segment(60, 100, category="peep_step", parameters={"peep": 16})

1. Find the times

Look at the whole recording before cutting it: the global signal and the airway pressure, with the segments drawn on top.

from fasteit.models import Segment
from fasteit.parsers.loader import load_data
from fasteit.plotting import plot_segmentation

raw = load_data("P02_T0.bin")
plot_segmentation(raw)                       # read the times on the x axis

segments = [
    Segment(0, 290, "phase", "tidal"),
    Segment(300, 1500, "phase", "peep_trial"),
    Segment(285, 300, label="movement"),
]
plot_segmentation(raw, segments)             # check the cuts before preprocessing

The times read on the figure are precise enough: what matters is which breaths fall inside, not the exact sample.

2. Use them

from fasteit.features import tidal_features
from fasteit.models import Segment
from fasteit.parsers.loader import load_data
from fasteit.preprocessing import preprocess
from fasteit.segmentation import segment_peep_trial

raw = load_data("P02_T0.bin")
segments = [
    Segment(0, 290, "phase", "tidal"),
    Segment(300, 1500, "phase", "peep_trial"),
    Segment(285, 300, label="movement"),
]
recording = preprocess(raw, exclusions=[s for s in segments if s.category == "exclusion"])
baseline = tidal_features(recording, phase=segments[0])
trial = segment_peep_trial(recording, phase=segments[1])

Steps placed by hand go to segment_peep_trial(recording, peep_steps=[...]), with the PEEP of each step and, if known, its plateau or driving pressure and tidal volume in parameters. The levels of the protocol (SegmentationConfig(peep_levels=...)) give each step its peep_set: see analyse a recording.

Only Dräger .bin and Timpel files (reconstructed images) can be cut and analysed.

Rules

• A breath belongs to an interval when it starts and ends inside it: a breath cut by the edge of a phase or of a step is left out.
• Exclusions are widened by 0.5 s on each side (filters spread an artifact over the next samples), and every breath that touches them is excluded. To exclude one breath only, use a short interval around its peak, not the whole breath.
• Excluding is not cutting: the signal stays whole, with its original times, and the excluded breaths stay in the tables marked excluded.
• A step placed by hand is the selection of breaths to analyse: all its valid breaths, for example the ten chosen on the figure.
• One recording is one file. Two parts of the same file share the preprocessing (lung mask, filters); two files of the same patient are two recordings, each with its own.

Details: preprocessing (exclusions) and segmentation (phases and steps).