Algorithm Benchmarking
Compare multiple motor unit decomposition algorithms including CBSS, SCD, Upperbound, and AE methods with standardized evaluation metrics and report card generation.
Compare multiple motor unit decomposition algorithms including CBSS, SCD, Upperbound, and AE methods with standardized evaluation metrics and report card generation.
Generate simulated EMG datasets using NeuroMotion biomechanical simulation and experimental datasets in standardized BIDS and Croissant formats.
Access datasets through Harvard Dataverse with persistent DOIs, following FAIR principles for Findable, Accessible, Interoperable, and Reusable data.
Docker and Singularity support for reproducible algorithm execution, ensuring consistent results across different computing environments.
Comprehensive evaluation metrics including silhouette scores, spike train comparisons, and automated report card generation for algorithm comparison.
Interactive Jupyter notebooks providing step-by-step guides for dataset loading, algorithm usage, and result evaluation.
All datasets are hosted on Harvard Dataverse and follow FAIR principles with persistent DOIs:
Convolutional Blind Source Separation for motor unit decomposition.
Swarm Contrastive Decomposition - GitHub
Upper bound performance estimation for algorithm comparison.
Autoencoder-based decomposition methods.
# Clone the repository
git clone https://github.com/dfarinagroup/muniverse.git
cd muniverse
# Create virtual environment
python -m venv .venv
source .venv/bin/activate # On Windows: .venv\Scripts\activate
# Install package
pip install -e .
# Install development dependencies (optional)
pip install -e ".[dev]"from muniverse.datasets import load_dataset
from muniverse.algorithms import decompose_recording
# Load a dataset
dataset = load_dataset("neuromotion-test", output_dir="./data")
# Run decomposition
results, metadata = decompose_recording(
data="path/to/emg_data.edf",
method="cbss"
)If you use MUniverse in your research, please cite:
@software{muniverse2024,
title = {MUniverse: Benchmarking Motor Unit Decomposition Algorithms},
author = {Mamidanna, Pranav and Klotz, Thomas and Halatsis, Dimitrios and Mendez Guerra, Irene},
year = {2024},
url = {https://github.com/dfarinagroup/muniverse},
license = {GPL-3.0}
}MUniverse is licensed under the GNU General Public License v3.0 (GPL-3.0).
Important Licensing Notes:
See the LICENSE file for full details.