BrainO¶

BrainO is an open-source EEG analysis engine for researchers. It imports recordings in any common format into de-identified BIDS datasets, runs a validated analysis pipeline (filtering, ICA, bad channels, epoching, ERPs, spectra, time-frequency and group statistics), and records everything it does, so that every analysis can be exported as a script that re-runs it and checks its results.

It is built on MNE-Python, MNE-BIDS, PyPREP, autoreject and mne-icalabel, and is the engine of a larger platform with a local AI analysis agent and a desktop app (see the roadmap).

Status: alpha (v0.1)

The EEG engine, its command line and its Python API are usable; the AI agent, the desktop app and the 3D viewer are in development. Interfaces may still change between 0.x releases.

What it does¶

Import EDF/BDF, BrainVision, FIF, EEGLAB, Neuroscan CNT, EGI MFF, GDF, Curry, Nihon Kohden and Persyst recordings into BIDS, removing names, dates and other personal details on the way. More
Analyse with typed tools: one function per step, with a JSON schema for its parameters. The same tools are used from Python, the command line and (later) the AI agent. More
Decide what the results rest on: steps that drop data (bad channels, ICA components) only propose; a person approves, edits or rejects, and the decision is recorded.
Pipelines declared in YAML run over a whole dataset, skipping steps whose inputs and settings have not changed. More
Reproduce: every tool call is recorded with input and output hashes and library versions; any run or pipeline exports as a Python script that re-runs it and reports whether each result is the same. More
Validated against ERP CORE (Kappenman et al., 2021): the P3 and N170 pipelines reproduce the published effects, participant by participant (r = 0.95 and 0.97). More

Install¶

pip install "braino[eeg]"

Python 3.11 or newer. The eeg extra installs the scientific stack; without it, the core (tools, provenance, pipelines specs, command line) still installs and imports.

Next: Getting started.