Skip to content

nih-megcore/MegNet

Repository files navigation

MEGNET

megnet-tests

This repository is a fork of the code listed below in the original code reference. This repository adds an automated processing wrapper and python package installation around the original codebase. The current codebase utilizes mne python to preprocess the data, generate the infomax ICA components (n=20), circular topography maps, and timeseries outputs. The architecture of neural net has been preserved, however, the weights have been reset to uniform distribution and retrained using repository data from MEGIN, CTF, 4D, and KIT systems.

Install

conda create -n megnet 'mne>=1.6' 'python<3.12'
conda activate megnet
pip install MEGnet-neuro

Post install initalization (downloading huggingface model weights)

megnet_init

Original Code Repository

https://github.com/DeepLearningForPrecisionHealthLab/MegNET_2020
Manuscript available: https://pubmed.ncbi.nlm.nih.gov/34274419/
DOI: https://doi.org/10.1016/j.neuroimage.2021.118402
PMID: 34274419

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •