Aperiodic Methods Project
Contents
Aperiodic Methods Project¶
Welcome to the Aperiodic Methods project!
This website hosts an online version of the code and figures for the Aperiodic Methods project, which surveys and compares available methods for measuring aperiodic activity in neural data.
This project is led by Tom Donoghue, with contributions from members and affiliates (past and present) of the Voytek lab.
Overview¶
To examine these methods, we use simulated data with known aperiodic and periodic properties in order to evaluate aperiodic methods and how they perform on neural data.
This broadly is organized into the following sections:
Introduction: introducing the main ideas and approaches of this project
This section introduces the list of included methods, examines their prevalence in the literature, and introduces the data simulations
Simulations: simulating time series and power spectra that mimic neural data
This section overview the approaches for simulating data, and how we explore across different parameters
Time Domain Method Evaluations: evaluating time domain methods for measuring aperiodic activity
This section introduces a series of time domain methods, and evaluates their performance on simulated data
Frequency Domain Method Evaluations: evaluating frequency domain methods for measuring aperiodic activity
This section introduces a series of frequency domain methods, and evaluates their performance on simulated data
Method Comparisons: comparing different methods head-to-head
This section compares methods directly to each other, on simulated data
Data Applications: examining method performance and comparing between methods on real data
This section applies the methods to EEG and iEEG datasets, and compares the methods
Available materials are created as Jupyter Notebooks, and are intended to be executed and explored in a hands-on manner. There is a download link at the top left of the page, that can be used to download each page as a notebook.
Methods¶
For this project, ‘aperiodic methods’ are broadly construed
Methods included in this project:
AutoCorrelation Measures
Fluctuation Measures (e.g. Hurst Exponent, DFA)
Fractal Dimension Measures (e.g. Higuchi’s Fractal Dimension)
Complexity Measures (e.g.: fractal dimension)
Information Theory Metrics (e.g. entropy)
Spectral Domain Methods (e.g. spectral parameterization, IRASA)
Reference¶
This project is described in the following preprint:
Donoghue T, Hammonds R, Eric Lybrand, Waschke L, Gao R, & Voytek B. Evaluating and
Comparing Measures of Aperiodic Neural Activity. bioRxiv. DOI: 10.1101/2024.09.15.613114
Direct Link: https://doi.org/10.1101/2024.09.15.613114
Source¶
The materials hosted here are developed and managed in the project repository.
The source code for creating this website is available in the site repository.