bisdecode 0.1.0

Decode Aspect/Covidien/Medtronic BIS Advance/Vista data-download files (processed trend, spectrogram, raw EEG) to tidy CSV/NumPy.

bisdecode

Decode Aspect / Covidien / Medtronic BIS Advance / Vista data-download files into analysis-ready tables. The per-case DH export is a vendor-undocumented, nine-file bundle; bisdecode reverse-engineers it into tidy CSV + NumPy.

⚠️ Unofficial / reverse-engineered. Not affiliated with or endorsed by the device manufacturer. Field mappings were validated against BIS R2 4.1 / BISx 1.15.2.22 exports cross-checked with the official BIS PDF reports. Verify against your own device revision before relying on it. Raw EEG is stored as ADC counts; µV scaling needs the (not-exported) BIS R2 gain constant.

Install

pip install bisdecode

CLI

# decode one case (folder containing a DH*/ subfolder, or the DH folder itself)
bisdecode decode path/to/M-XXXX-... -o decoded

# batch every case under a root
bisdecode decode raw-data --all -o decoded

# anonymised export: per-case date-shift, output named by Case ID, key kept secret
bisdecode decode raw-data --all --anonymise -o decoded_anon --key-file KEY.csv

# one-row-per-case cohort summary from decoded trends
bisdecode cohort decoded -o cohort_summary.csv

Per case you get: *_trend.csv (per-second BIS / SR / SEF / median freq / total power / EMG / SQI / impedance / artifact / burst, per channel), *_spectra.csv (60-bin density spectral array), *_events.csv, *_messages.csv, *_meta.json, and *_eeg.npy (raw EEG, int16, 2ch @ 128 Hz).

Python API

import bisdecode
meta = bisdecode.decode_case("path/to/M-XXXX-...", "decoded")
trend = bisdecode.parse_spa("…/L…spa")          # -> pandas DataFrame
eeg   = bisdecode.parse_eeg("…/L….r2a")          # -> ndarray (n, 2) int16
row   = bisdecode.summarize("decoded/X/X_trend.csv", channel="Ch1")

Anonymisation

--anonymise applies an independent random per-case offset (magnitude in [--shift-min-days, --shift-days], random sign) to every timestamp — including those embedded in message text — so dates and clock-times move while all within-case intervals are preserved. Outputs are named by the monitor Case ID. The key file maps source_base → offset_seconds; keep it secret (it reverses the shift: real = shifted − offset). Re-runs reuse the key, so offsets are stable.

License

MIT. See LICENSE.