BDSP Python Reader

Reader-first Python package for BDSP/Sleep YODA standardized HDF5 data.

import bdsp

record = bdsp.read_h5("/path/to/study.h5")

record.signals["c3-m2"]
record.metadata
record.annotations
record.features
record.channel_info

df = record.to_dataframe(resample_fs=200)

with bdsp.open_h5("/path/to/study.h5") as reader:
    x, t = reader.read_signal("c3-m2", start=60, duration=30)

Scope

Version 0.1 reads BDSP/Sleep YODA H5 files. It intentionally does not include EDF conversion, S3 downloading, GUI code, model inference, report generation, or feature computation.

API

• bdsp.read(path, **kwargs) dispatches to the H5 reader for .h5 and .hdf5.
• bdsp.read_h5(path, raw_digital=False, include_annotations=True, include_features=True) returns a BDSPRecord.
• BDSPRecord.to_dataframe(resample_fs=200) returns synchronized signal and annotation columns.
• bdsp.open_h5(path) returns a lazy BDSPH5Reader for slice-based reads.
• bdsp.compat.sleep_yoda exposes load_standardized_h5 and PSGReader compatibility aliases.

Development

python -m pip install -e ".[test]"
pytest

The project metadata requires Python 3.10+. The reader source itself is kept small and depends only on h5py, numpy, pandas, and scipy.

License

This package is licensed under the Apache License, Version 2.0.