Cycling & EEG Entropy

Cycling reduces the entropy of neuronal activity in the human adult cortex

Research Pipeline
EEG
R
Reproducible Research
An R analysis pipeline examining recurrence-matrix entropy in scalp EEG signals recorded during cycling and rest across four behavioural conditions in healthy adults.
Published

April 1, 2024

About

This repository contains all data and analysis code for a study examining how physical activity modulates the complexity of cortical neural signals, using recurrence-matrix entropy (S_Max) computed from scalp EEG.

The analysis pipeline is fully implemented in R and renders a self-contained HTML report. Raw EEG and pre-computed entropy data are archived on OSF.

Ferré IBS, Corso G, dos Santos Lima GZ, Lopes SR, Leocadio-Miguel MA, França LGS, de Lima Prado T, Araújo JF (2024). Cycling reduces the entropy of neuronal activity in the human adult cortex. PLOS ONE, 19(10): e0298703. doi:10.1371/journal.pone.0298703

Study Design

24 healthy adults (13 women, 11 men; mean age ~21 years) performed four behavioural conditions — resting eyes open, resting eyes closed, cycling eyes open, and cycling eyes closed — on a horizontal stationary bicycle. EEG was recorded from eight bimodal electrode pairs covering frontal, central, parietal, and occipital scalp regions. Recurrence-matrix entropy (S_Max) was computed for each 300-sample window across the 2-minute recordings, producing 100 time points per condition per electrode.

Key Findings

Cycling was associated with significantly reduced cortical entropy across all eight electrode sites compared to rest (all p ≤ 0.042), suggesting that physical activity organises neural dynamics toward lower complexity. A further entropy reduction was observed when cycling with eyes closed, most prominently at right occipital (p = 0.045) and left parietal (p = 0.006) sites.

The eyes-open vs eyes-closed comparison at rest confirmed the expected entropy reduction in the occipital region (t = −2.657, p = 0.015) — consistent with well-established alpha-rhythm physiology and serving as a validation of the method.

Statistical Approach

All models are generalised linear mixed models (GLMMs) with participant as a random effect, tested using 10,000-repetition two-sided permutation tests via the ptestR package:

# Main effect: cycling vs resting
med_entr ~ State + (1|sub)

# Interaction: State × Eyes
med_entr ~ State * Eyes + (1|sub)

Analysis Pipeline

The pre-computed entropy data in data/ are the output of circadia-bio/maxEntropy, a Julia script that computes S_Max from the raw EEG recordings. The R analysis then:

  1. Loads and reshapes all 32 entropy files (8 electrodes × 4 conditions) via src/legacy.R
  2. Fits permutation GLMMs per electrode in translating.Rmd
  3. Generates topographic scalp maps via topoplots.Rmd

The environment can be reproduced via Docker (docker-compose up) or locally in R ≥ 4.3.