invertmeeg - A high-level Python library for M/EEG inverse solutions

This package provides 96 inverse solvers for M/EEG source imaging, integrating with the mne-python framework. It covers minimum norm methods, beamformers, Bayesian approaches, sparse recovery, subspace methods, and deep learning models in a unified API.

Read the documentation here.

Installation

The recommended way to install invertmeeg is using uv:

uv add invertmeeg

Python 3.14 or newer is required. The core installation does not include neural-network, GUI, or notebook dependencies.

Alternatively, you can use pip:

pip install invertmeeg

Extras

• Neural Networks: uv add "invertmeeg[ann]" (or pip install "invertmeeg[ann]")
• Notebooks: uv add "invertmeeg[notebooks]" (or pip install "invertmeeg[notebooks]")
• Visualization: uv add "invertmeeg[viz]" (or pip install "invertmeeg[viz]")

Development

If you are contributing to invertmeeg, we use uv and make to manage the development environment and code quality.

Setup

make install

This will sync all dependencies (including dev tools) and install the pre-commit hooks.

Manual uv setup, install all packages except ann

uv sync --locked --extra viz --extra docs --extra notebooks

Useful Commands

• make lint: Run Ruff and Mypy checks.
• make format: Auto-format code with Ruff.
• make test: Run the fast test suite with pytest.
• make test-all: Run the complete suite, including accuracy and orientation regressions.
• make check: Run all linting and tests.
• make clean: Remove temporary cache files (pycache, .mypy_cache, etc.).

Quick Start

from invert import Solver

# fwd = ...   (mne.Forward object)
# evoked = ... (mne.Evoked object)

solver = Solver("MNE")
solver.make_inverse_operator(fwd)
stc = solver.apply_inverse_operator(evoked)
stc.plot()

Features

• 96 inverse solvers accessible through a single Solver("solver_id") interface
• Interactive benchmark leaderboard -- compare solver performance across datasets, metrics, and categories
• Automatic regularization (GCV, L-curve, product methods)
• Returns standard mne.SourceEstimate objects
• Simulation utilities for benchmarking

Solver Categories

Full Algorithm List

Minimum Norm

LORETA

Other Minimum-Norm-like Algorithms

Bayesian

Beamformers

Dipole Fitting

Structured Sparsity

Artificial Neural Networks

Matching Pursuit / Compressive Sensing

MUSIC/RAP/Subspace

Basis Functions

Other

Licensing

This project is dual-licensed:

• Open source: GPLv3.
• Commercial alternative: if you want to use invertmeeg in a proprietary/closed-source product without GPLv3 obligations, a separate commercial license is available (contact below).

For Researchers & Open Source Users (GPLv3)

We love science. This library is free to use for research, education, thesis projects, and open source projects under the terms of GPLv3.

• Attribution required: please credit invertmeeg (and ideally cite this repository) in any publications, reports, or released artifacts that use it.
• Distribution note: if you distribute software that includes/links this library, you need to comply with GPLv3 (including providing corresponding source under GPLv3).

For Industry & Commercial Entities

If you are a company, a startup, or an individual intending to use this library in a proprietary/closed-source product, please contact me to obtain a separate commercial license.

Contact: lukas.hecker.job@gmail.com

Citation

If you use this package and publish results, please cite as:

@Misc{invertmeeg2022,
  author =   {{Lukas Hecker}},
  title =    {{invertmeeg}: A high-level Python library for M/EEG inverse solutions.},
  howpublished = {\url{https://github.com/LukeTheHecker/invertmeeg}},
  year = {since 2022}
}

Send feedback or feature requests to lukas.hecker.job@gmail.com.