Electroencephalography (EEG) is a powerful tool for assessing brain complexity (BC), offering valuable insights into neurological processes in both health and disease. This work builds on previous studies that used recurrence entropy to evaluate BC across different neurophysiological conditions, such as rest and cycling movement [1]. In addition to this method, we implement machine learning techniques to classify brain states based on recurrence microstate patterns.

Entropy, in the realm of information theory, quantifies the information content embedded within a time series. In neuroscience, this concept is crucial for understanding the complexity of EEG signals. Recurrence microstates, representing distinct patterns within time series data, allow us to examine neural dynamics in more detail. By segmenting each EEG recording and performing recurrence analysis, we capture the periodicity and complexity of brain signals efficiently, even with smaller data samples.

EEG data were recorded from 24 healthy participants using a bimodal assembly of eight electrodes: F3-Fz, F4-Fz, C3-Cz, C4-Cz, P3-Pz, P4-Pz, O1-A2, and O2-A1. Recordings were taken in both resting and cycling conditions, with eyes open and closed. Recurrence entropy was computed for each EEG series, revealing higher entropy in resting states and with eyes open compared to cycling and eyes closed.

The classification of brain states using recurrence analysis data resulted in higher accuracy compared to using raw EEG time series alone. This suggests that recurrence analysis reveals additional information not evident in the original EEG data, highlighting its value in uncovering underlying neural dynamics.

The integration of these techniques has the potential to shed light on the neural mechanisms underlying motor control and sensory processing.

[1] Ferré, Iara Beatriz Silva, et al. "Cycling reduces the entropy of neuronal activity in the human adult cortex." bioRxiv (2024): 2024-01.