UnitRefine is a machine-learning-based toolbox designed to streamline spike sorting curation by reducing the need for manual intervention.
With a focus on accessibility and broad community adoption, UnitRefine offers:
- Seamless integration with SpikeInterface.
- Pre-trained machine learning models for effortless application.
- The flexibility to train custom models using your own curated datasets and metrics.
- Easy sharing of trained models via Hugging Face Hub, fostering collaboration and reproducibility.
- A manually curated dataset, labeled by 7 experts across 11 Neuropixels 1.0 recordings in mice, is also available.
- Each recording was annotated by 2 to 5 people, with an agreement rate of 80% among the curators.
- Pre-trained Models: Apply ready-to-use classifiers for noise removal and unit refinement.
- Custom Training: Train models on your own data to meet specific experimental needs.
- Integration: Fully integrated with SpikeInterface for a smooth user experience.
- Collaboration: Share or download models from the Hugging Face Hub, enabling community-driven advancements.
Ensure that SpikeInterface is installed in your environment. Installation instructions can be found here.
To get started with UnitRefine, refer to the automated curation tutorials available in the SpikeInterface documentation:
Automated Curation Tutorials
Additionally, this repository includes Jupyter Notebooks with detailed step-by-step tutorials on how to:
- Apply pre-trained models.
- Train your own classifiers.
This repository contains two scripts, model_based_curation.py and train_manual_curation.py, that provide a detailed explanation of how certain features work when integrated with the SpikeInterface library.
- These scripts cannot be used independently. They are designed for understanding the inner workings of SpikeInterface-related functionalities.
- For seamless integration and practical use, please install and use the official SpikeInterface repository.
- These scripts rely on features already available in the SpikeInterface library.
I would like to express my sincere gratitude to the following individuals for their invaluable contributions to this project:
-
Code Refactoring and Integration in SpikeInterface:
Chris Halcrow, Jake Swann, Robyn Greene, Sangeetha Nandakumar (IBOTS) -
Model Curators:
Nilufar Lahiji, Sacha Abou Rachid, Severin Graff, Luca Koenig, Natalia Babushkina, Simon Musall -
Advisors:
Alessio Buccino, Matthias Hennig, Simon Musall
If you find UnitRefine useful in your research, please cite the following DOI: https://doi.org/10.6084/m9.figshare.28282841.v2.
We will be releasing a pre-print soon. In the meantime, please use the above DOI for referencing.
We encourage feedback, contributions, and collaboration from the community to improve UnitRefine. Feel free to open issues or submit pull requests to enhance the toolbox further.