Resting-state is not enough: alpha and mu rhythms change shape across development, but lack diagnostic sensitivity

This repository contains all the necessary code to reproduce the analysis and figures of the following manuscript:

• Bender, Voytek, & Schaworonkow: Resting-state is not enough: alpha and mu rhythms change shape across development, but lack diagnostic sensitivity. bioRxiv (2023). doi: https://doi.org/10.1101/2023.10.13.562301.

Dataset

The results are based on an openly available developmental dataset of resting-state electroencephalography (EEG) recordings from over 2500 children (ages 5-18) from the Child Mind Institute published by Alexander and colleagues (2017):

Alexander et al.: An open resource for transdiagnostic research in pediatric mental health and learning disorders Scientific Data (2017). doi: 10.1038/sdata.2017.181.

The dataset is a part of the Healthy Brain Network, an ongoing initiative focused on creating a transdiagnostic biobank of data from over 10,000 children and adolescents (ages 5-21). The whole dataset contains EEG, magnetic reasonance imaging (MRI), and phenotypic data, any of which can be downloaded here. Some participants did not complete all assessments, so the number of participants varies depending on the particular type of data one wishes to analyze.

To reproduce our results for Figures 1-3, the EEG data for each of the 10 releases should be downloaded and placed in the data/download folder. These data are distributed under the Creative Commons, Attribution Non-Commercial Share Alike License.

To reproduce our results for Figure 4, the phenotypic data containing participant diagnoses must be downloaded. Details on how to access this phenotypic data are provided here. The diagnostic information should be contained in a CSV file, consisting of subject and diagnosis columns, with each row containing the subject ID and diagnosis for one participant. The path to this CSV file should be set as the DIAGNOSES_CSV variable in the code/params.py file.

Requirements

The provided Python 3 scripts require the following packages:

• numpy and scipy for numerical computation
• pandas for storage and manipulation of tabular data
• mne for reading and storing EEG data
• fooof for parameterizing neural power spectra
• bycycle for computing waveform shape features from neural time series
• matplotlib, pyvista, scikit_image, and seaborn for visualizing data and generating figures
• pinguoin and statannotations for calculating and visualizing statistics

The particular versions used to perform the analysis and create the figures for the manuscript are specified in the requirements.txt file. To install all these packages in one's local Python environment, simplify enter pip install -r requirements.txt into the command line from the root of this GitHub repository.

Processing Pipeline

Alpha and mu rhythms are mixed in sensor-space EEG due to volume conduction. In our analysis, we isolated these alpha-band rhythms from resting-state EEG using spatio-spectral decomposition (Nikulin et al., 2011) and template matching:

Our processing pipeline can be run from start to finish by running the code/proc_*.py files in order. The output files sources.csv and waveform_feats.csv are provided in the results folder if one wishes to avoid downloading the data and processing the entire pipeline. All parameters used to generate these output files are contained in the core/params.py file.

To see how the results and figures change with different choices of parameters, change the parameters in the core/params.py file and rerun the necessary code/proc_*.py files.

Figures

To generate figures from the manuscript, simply run the corresponding code/fig*.py file(s). Figures will be saved in the figures folder.