neuro.lab

Topics: eeg, neuroimaging, meg, mri, tauri, rust

Neuro Lab is a local-first desktop workspace for EEG, MEG, and MRI research data. It has no server dependency: the React interface runs inside Tauri and calls typed Rust commands directly. All computation and persistence stay on the local filesystem.

• Inspect, validate, and process EEG, MEG, and MRI recordings in one workspace.
• Reproducible JSON pipeline chains for filtering, artifact rejection, epoching, and more.
• Versioned datasets, experiments, analysis runs, and artifacts with provenance.
• neuro CLI for scripting: inspect, validate, convert, and export from the shell.
• Runs fully offline; data never leaves your machine.

Table of Contents

• neuro.lab
  • Table of Contents
  • Features
  • Repository layout
  • Command line
  • Get Started
    • Prerequisites
    • Desktop app
    • Checks
  • Landing site
  • License

Features

• Local workspace with projects, experiments, datasets, versioned artifacts, provenance records, and resumable chunked uploads.
• Background analysis runs with progress reporting and cancellation.
• EEG: NPY/EDF/BrainVision input, EDF+ annotations and .vmrk markers as events, filtering, PSD, bad-channel detection, re-referencing (average/CAR/mastoid/median/robust/Laplacian), baseline correction, event-locked epoching with ERP, artifact rejection, FastICA.
• MEG: FIF/CTF/KIT/BIDS/OPM input, CTF compensation grades and FIF projectors/compensation applied to data, KIT trigger events and .elp/.hsp digitization sidecars, BIDS multi-run concatenation and per-run channel reconciliation, epoching/evoked responses, synthetic gradiometers, ICA artifact rejection, SSP, a documented low-order SSS approximation for gross interference rejection, and NPY/CSV/EDF/FIF/BIDS export (the FIF writer is validated against MNE).
• MRI: NIfTI and DICOM input (native 8/16-bit, multi-frame, RLE, JPEG baseline/extended, series directories), multiplanar slices, MIP, statistics, histogram, single operations and multi-step pipeline chains, bias-field correction, heuristic skull stripping, and rigid/affine registration with NCC or mutual-information metrics.
• Dataset search/sort, file-set validation, and artifact export.

Repository layout

crates/
  neuro-core/       Shared identifiers, provenance, checksums, run state
  neuro-eeg/        EEG readers (NPY, EDF+/EDF, BrainVision), pipelines
                    (filter, PSD, bad channels, re-reference, baseline,
                    epoching/ERP, ICA/FastICA with topographies,
                    SSP), EDF export
  neuro-meg/        MEG readers (FIF, CTF .ds, KIT .sqd/.con, BIDS, OPM),
                    compensation/projectors, events, pipelines
                    (filter, PSD, epoching/evoked, synthetic gradiometers,
                    ICA, SSP, low-order SSS), NPY/CSV/EDF/FIF/BIDS export
  neuro-mri/        NIfTI + DICOM input, slices, projections, statistics,
                    threshold/smooth/mask/segment/resample/z-score ops,
                    bias-field correction, heuristic skull stripping,
                    rigid/affine registration (NCC or mutual information)
  neuro-store/      Local workspace persistence: projects, datasets,
                    dataset versions, experiments, runs, artifacts,
                    upload sessions
  neuro-cli/        `neuro` command-line interface: inspect/validate EEG,
                    MEG, and MRI files, generate synthetic EEG, export EDF,
                    convert DICOM to NIfTI — JSON output for scripting
packages/desktop/   Tauri + React + TypeScript frontend
packages/landing/   Static landing page (no build step)
reference/          Imported source repositories retained for migration reference

Command line

cargo install --path crates/neuro-cli   # or: cargo run -p neuro-cli -- <cmd>

neuro inspect-eeg recording.edf --pretty
neuro validate-meg /path/to/meg/ds
neuro convert-dicom scan.dcm -o volume.nii
neuro --help                            # all subcommands

Output is JSON on stdout; diagnostics go to stderr.

Get Started

Prerequisites

• Node.js (see .nvmrc) and pnpm
• Rust stable toolchain

Desktop app

Prebuilt bundles are attached to the latest GitHub release:

scripts/install.sh        # download a release bundle, or build from source

Or from a local checkout:

pnpm install
pnpm tauri:dev            # development
pnpm tauri:build          # production bundles under target/release/bundle

Checks

cargo fmt --all -- --check
cargo clippy --workspace --all-targets -- -D warnings
cargo test --workspace

pnpm typecheck            # from packages/desktop
pnpm lint
pnpm test
pnpm build
pnpm test:e2e

Landing site

The static marketing site lives in packages/landing (no build step, deployed to GitHub Pages): https://hieudoanm.github.io/neuro.lab/

License

This project is licensed under the MIT License.