Introduction
| Author: | James R. McIntosh |
|---|---|
| Contact: | j.mcintosh@columbia.edu |
Purpose
- Code base to test machine learning approaches for EEG rhythm phase prediction.
- Related to arxiv manuscript: McIntosh, J. R., & Sajda, P. (2019). Estimation of phase in EEG rhythms for real-time applications. arXiv preprint arXiv:1910.08784.
Installation and getting started
- After cloning the repository, the following commands will install the necessary libraries:
conda env create -f py_eegepe.yml- To get started, an across subject analysis as described in the manuscript can be run.
conda activate py_eegepe
cd examples/
python gen_figure_supp_as.pyBefore the gen_figure_supp_as.py python scrip will work, the data has to be correctly configured in two locations:
- gen_figure_supp_as.py : datadir variable (the root of the dataset folder), dataset variable (the name of the dataset folder)
- data_specific.py : specifier, specifier_proc and _sublist which act to configure the location of data within the project directory. As well as preprocessor which handles the data specific pre-processing to be carried out (see examples for existing datasets within data_specific.py)
Examples directory has been configured to operate with the Child Mind Institute - healthy brain network data. After downloading this data must be processed by a matlab script (convert_mat_to_eeglab.m) to enable loading with MNE.
The path structure should look like this:
[datadir]/data/[subject_ID]/EEG/raw/mat_format/
And where resting data is present, the following will be added after running the matlab script:
[datadir]/data/[subject_ID]/EEG/raw/eeglab_format/
data_specific.py is currently setup to deal with this configuration, but this can be changed as is required.
gen_figure_supp_as.py as well as the other high-level functions in examples/ and manuscripts/ are configured to cycle through a series of experiments as configured in paradigm.py
X