lunapi 1.6.3

Python interface to the Luna toolset for sleep signal analysis

lunapi

Python interface to the Luna C/C++ sleep-signal analysis engine.

lunapi exposes the full Luna command set to Python, enabling scripted and notebook-based workflows for polysomnographic (PSG) data — using the same analytical core as the Luna command-line tool and the Lunascope desktop application.

---

Statement of Need

Large-scale sleep research increasingly requires processing thousands of PSG recordings programmatically, while maintaining the ability to inspect individual records interactively. Existing Python EEG libraries are general-purpose frameworks that lack dedicated sleep-analysis primitives. lunapi provides direct Python access to Luna's domain-specific command set — covering spectral analysis, sleep staging, annotation handling, artifact detection, connectivity, and more — without requiring users to leave the Python/notebook environment or learn the Luna command-line interface.

---

Features

• Full Luna command set — run any Luna command (PSD, SPINDLES, SOAP, POPS, …) directly from Python and receive structured results as DataFrames
• Project-level workflows — load a sample list and execute commands across an entire cohort in a few lines
• Per-recording access — attach EDF and annotation files, extract raw signals and annotation events, insert or modify signals in memory
• Structured output — results returned as pandas DataFrames keyed by command and stratum
• SQLite result stores — write output to .db files that can be queried as virtual TSV-like tables
• Sleep staging — SOAP and POPS automated staging models accessible directly from Python
• Visualization helpers — hypnogram, PSD, spectrogram, and topographic map plotting for notebooks
• NSRR Moonbeam — download and attach recordings directly from the National Sleep Research Resource
• GPA association modeling — linear association analysis on cohort-level result matrices
• Docker containers — pre-built images with Luna, lunapi, and all dependencies

---

Installation

lunapi is distributed as a binary wheel via PyPI (no compiler required):

pip install lunapi

Supported platforms: macOS (Intel & Apple Silicon), Linux (x86_64), Windows 10/11 (x86_64), Python 3.9–3.14.

Using a virtual environment is recommended:

python3 -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install lunapi

---

Worked Example

Single-recording workflow

import lunapi as lp

# Create a project and attach a recording
p = lp.proj(verbose=False)
rec = p.inst("my_subject")
rec.attach_edf("/path/to/recording.edf")
rec.attach_annot("/path/to/recording.annot")

# List available channels
print(rec.chs())

# Run a Luna command and retrieve results as a DataFrame
rec.eval("PSD sig=EEG dB spectrum")
df = rec.table("PSD", "CH_F")
print(df.head())

Cohort workflow

import lunapi as lp

p = lp.proj(verbose=False)
p.sample_list("study.lst")        # three-column file: ID, EDF path, annot path

# Run PSD across all recordings (returns dict of DataFrames)
results = p.proc("PSD sig=EEG dB spectrum")
df = results["PSD"]["CH_F"]
print(df.groupby("CH")["PSD"].mean())

Automated sleep staging

rec.eval("RUN-POPS sig=C3 ref=M2 path=~/dropbox/pops")   # point to downloaded POPS models
stages = rec.table("RUN_POPS", "E")   # per-epoch predictions
print(stages["PRED"].value_counts())

Full documentation, command reference, and interactive notebooks are at zzz.nyspi.org/luna/.

---

Testing

pip install -e ".[dev]"
pytest

This runs 59 tests covering the core workflow (load instance, run commands, retrieve tables, extract signals, annotation handling, sample-list workflow), visualization helpers, GPA utilities, and package smoke tests. No real EDF files are required — a synthetic in-memory recording is generated by the test fixtures.

---

Documentation

Resource	URL
Luna ecosystem	https://zzz.nyspi.org/luna/
lunapi notebooks	https://github.com/remnrem/luna-api/tree/main/notebooks
Lunascope desktop app	https://github.com/Lorcan7274/lunascope
NSRR	https://sleepdata.org

---

Contributing

See CONTRIBUTING.md for guidelines on bug reports, feature requests, and pull requests.

Questions / Support

Open an issue on GitHub or write to [e-mail].