Installation¶

Requirements¶

• Python 3.10, 3.11, 3.12, 3.13 or 3.14
• NumPy ≥ 1.26 (NumPy 1.26 and 2.x are both supported) and SciPy ≥ 1.11, installed automatically

ecg-records itself is pure Python (a py3-none-any wheel). The numerical engine, ecg-records-engine, ships as compiled wheels for CPython 3.10–3.14 on Linux (x86_64, aarch64), macOS (x86_64, arm64) and Windows (x86_64).

Install from PyPI¶

pip install "ecg-records[engine]"

This installs the ECG Record read/write/query API, the ecg-record command line, the plotting module and the measurement engine. Both packages depend only on NumPy and SciPy and must share the minor version; since 0.3.2 the shell also needs an engine at least as new as 0.3.2 and the 0.3.2 engine needs the 0.3.2 shell (the staged pipeline moved into the engine). pip enforces both: on a platform that has no 0.3.2 engine wheel yet, pip install "ecg-records[engine]" installs the working pair 0.3.1 + 0.3.0.

pip install ecg-records alone installs everything except the engine: records can be read, queried and plotted, and measuring a signal raises EngineNotInstalledError with the install command. The engine is free for noncommercial use and needs a licence key for commercial use; see Licensing.

Optional extras¶

Extra	Command	Adds	Needed for
engine	pip install "ecg-records[engine]"	ecg-records-engine (compiled)	measuring signals; see Licensing
viz	pip install "ecg-records[viz]"	Matplotlib ≥ 3.7	ecgfeat.viz plots
table	pip install "ecg-records[table]"	pandas, pyarrow	record.to_dataframe(), Parquet output of ecg-record table / batch --table
interactive	pip install "ecg-records[interactive]"	Plotly	ecgfeat.viz.plot_record_interactive
performance	pip install "ecg-records[performance]"	Numba	source checkouts only: the compiled engine already contains the fast kernels
wfdb	pip install "ecg-records[wfdb]"	the wfdb package	WFDB layouts the built-in reader does not decode (multi-frequency, skewed, multi-segment); formats 16, 212, 24, 32, 61, 80 and .mat need nothing extra
interpret	pip install "ecg-records[interpret]"	the separate ecginterpret distribution	rule-based interpretation and reports, and the legacy ECGFeatureExtractor / to_dict path
all	pip install "ecg-records[all]"	viz, table, interactive, interpret	everything above except performance and wfdb

Extras combine: pip install "ecg-records[engine,viz]". The record path (ecg_record, load_record, the queries, ecgfeat.viz, the CLI) never needs ecginterpret.

Verify the installation¶

import ecgfeat

print(ecgfeat.__version__)               # 0.3.2
print(ecgfeat.STANDARD_12_LEADS)         # ('I', 'II', 'III', 'aVR', ..., 'V6')
print(ecgfeat.list_examples())           # ['af', 'lbbb', 'sinus']: bundled example ECGs

ecg-record --help

import ecgfeat is cheap: imports are lazy, so reading and querying records does not load the numerical engine, and nothing in the package imports Matplotlib until you use ecgfeat.viz.

Install from source¶

The source repository is private (the engine's source is not published; see Licensing). Maintainers with access install the package in editable mode from feature_extraction/, where the engine runs as Python source (optionally accelerated by Numba):

python -m pip install -e "./feature_extraction[viz,performance]"

See the contributor guide for the development tools.