Waveforms (ECG / EEG)

pydcm.waveforms reads and writes DICOM Waveform objects — 12-lead ECG, EEG, hemodynamic and audio — and hands the signal arrays to an analysis library. The multiplex_array decode is exact (difference = 0).

Read

from pydcm import waveforms

wf = waveforms.read_waveform("ecg.dcm")
wf["modality"]                          # "ECG"
group = wf["groups"][0]
group["sampling_frequency"]             # Hz
for ch in group["channels"]:
    ch["label"], ch["units"], ch["signal"]   # e.g. "Lead I", "mV", ndarray

For the raw multiplex matrix:

ds = pydcm.dcmread("ecg.dcm")
arr = waveforms.multiplex_array(ds, index=0, as_raw=False)   # [samples, channels], physical units

Write

signals = [lead_i, lead_ii, ...]   # a list of 1-D channel arrays
                                   # (or one array shaped [samples, channels])
waveforms.write_waveform(
    "ecg_out.dcm",
    signals,
    sampling_frequency=500,
    kind="ecg12",            # 12-lead ECG; also "eeg", "hemodynamic", "audio", …
    units="mV",
    patient_id="PID-1",
)

Each channel is quantised via its sensitivity (physical units per LSB — scalar, per-channel, or auto), and ChannelSensitivity is stored so read_waveform reconstructs the physical values.

Hand off to analysis

pydcm produces the signal arrays; analysis stays in the established tools:

import neurokit2 as nk
signals_df, info = nk.ecg_process(group["channels"][0]["signal"],
                                  sampling_rate=group["sampling_frequency"])

For EEG, mne consumes the same arrays. pydcm does not re-implement signal analysis — it gives those libraries clean, physical-unit input.