Reconfiguration of Brain Network Between Resting State and P300 Task

Fali Li, Chanlin Yi, Yuanyuan Liao, Yuanling Jiang, Yajing Si, Limeng Song, Tao Zhang, Dezhong Yao, Yangsong Zhang, Zehong Cao, Peng XuIEEE Trans. Cogn. Dev. Syst 2021 · OpenReview

TL;DR

This study investigates how brain networks reconfigure from a resting state to an oddball task (used to elicit the P300 brainwave). They found that during the task, delta/theta brainwave connectivity increases while alpha band connectivity (associated with the default mode network) decreases. These network changes correlate with P300 amplitudes and can be used to classify individuals, suggesting potential for optimizing P300-based brain-computer interfaces.

What Problem Does This Paper Address?

The P300 event-related potential (ERP) is a well-studied brainwave crucial for cognitive functions like attention and decision-making, and it's widely used in brain-computer interfaces (BCIs). While previous research has looked at how brain activity (like power spectra) changes between resting states and P300 tasks, it's been unclear how the functional connectivity between different brain regions reconfigures. Understanding this dynamic network shift is important because individual differences in P300 amplitude are significant and can affect BCI performance. This paper aims to bridge this gap by examining how brain networks, specifically in different frequency bands, adapt from a passive resting state to an active oddball task, and how these adaptations relate to P300 characteristics.

Methods

The researchers employed electroencephalography (EEG) to record brain activity. Their approach involved:

  • Resting-state and Oddball Task EEG: Participants underwent a 4-minute eyes-closed resting-state recording followed by three runs of a visual oddball task. In the oddball task, participants were presented with frequent "standard" stimuli and infrequent "target" stimuli and asked to mentally count the targets.
  • Functional Network Construction: EEG data was processed and analyzed in different frequency bands (delta/theta: 1-8 Hz, alpha: 8-13 Hz). They used coherence as a measure of linear interaction between pairs of electrodes (nodes in the network). This resulted in weighted adjacency matrices representing functional connectivity for each subject under both resting and task conditions.
  • Network Reconfiguration Analysis: They compared the functional networks between the resting and task states to identify significantly enhanced or suppressed connections.
  • Time-Frequency Analysis: Wavelet analysis was used to examine the dynamic changes in brain activity across different frequencies, particularly focusing on the Pz electrode, known to be a generator of P300.
  • Correlation and Classification: They analyzed the correlation between P300 amplitudes and the degree of network reconfiguration. Furthermore, they used machine learning (linear discriminant analysis and support vector machine) to classify subjects into high and low P300 amplitude groups based on their reconfigured network strengths.

Datasets

The study utilized EEG data collected from 24 healthy postgraduate students. No new public benchmarks were introduced.

Key Insights and Ideas

  • Brain Network Reconfiguration is Key: The study highlights that the brain doesn't just activate specific areas for tasks but dynamically reconfigures its entire network architecture.
  • Frequency-Specific Shifts: The reconfiguration isn't uniform across all brainwave frequencies. Delta/theta bands show increased connectivity, suggesting enhanced information processing, while the alpha band (linked to the default mode network) shows decreased connectivity, indicating a shift away from internal thought processes towards external task engagement.
  • Default Mode Network Suppression: The observed decrease in alpha band connectivity is interpreted as a suppression of the default mode network (DMN), which is typically active during resting states and internal cognition. This suppression is necessary to allocate resources for the task.
  • P300 Amplitude Predictability: The degree of network reconfiguration, particularly the increase in delta/theta connectivity and decrease in alpha connectivity, is directly correlated with the amplitude of the P300 response. This suggests that individuals with more efficient network reconfigurations are better at processing target stimuli.
  • Potential for BCI Optimization: The ability to classify individuals into high/low P300 amplitude groups based on network reconfiguration patterns opens doors for pre-screening or optimizing BCI systems that rely on P300 detection.

Key Results

  • Enhanced Delta/Theta Connectivity: Significantly increased functional connectivity was observed in the 1-8 Hz (delta/theta) frequency band when transitioning from resting state to the P300 task.
  • Decreased Alpha Connectivity: A significant decrease in connectivity was found in the 8-13 Hz (alpha) frequency band, indicative of a suppressed default mode network-like topology.
  • Correlation with P300 Amplitude: Increased coupling strengths in the 1-8 Hz band positively correlated with P300 amplitudes (r = 0.532), while decreased coupling strengths in the 8-13 Hz band showed a negative correlation (r = -0.364, though not passing Bonferroni correction).
  • Classification Accuracy: A classifier could differentiate between high and low P300 amplitude groups with an accuracy of 77.78% using the reconfigured coupling strengths as features.

Limitations and Future Work

  • Outlier Removal: The study acknowledges the presence of outlier subjects and excluded the top 10% based on Mahalanobis distance to ensure robust analysis. The impact of these exclusions on the generalizability of findings could be further explored.
  • Coherence as a Measure: Coherence measures linear relationships. Future work could explore non-linear connectivity measures to capture more complex brain interactions.
  • Single Electrode Focus for TFD: Time-frequency analysis was primarily conducted on a single electrode (Pz). Expanding this to a network-wide time-frequency analysis could provide a more comprehensive picture.
  • BCI Application: While the study suggests potential for P300-BCI optimization, further research is needed to translate these findings into practical BCI applications, including real-time implementation and validation with diverse user populations.

Summary updated

Abstract

Previous studies have explored the power spectra from the resting-state condition to the oddball task, but whether significant difference in the brain network exists is still unclear. This article aims to address how the brain reconfigures its architecture from a resting-state condition (i.e., baseline) to the P300 task in the visual oddball task. In this article, electroencephalograms (EEGs) were collected from 24 subjects, who were required to only mentally count the number of target stimulus; afterward, EEG networks constructed in different bands were compared between baseline and task to evaluate the reconfiguration of functional connectivity. Compared with the baseline, our results showed the significantly enhanced delta/theta functional connectivity and decreased alpha default mode network in the progress of brain reconfiguration to the task. Furthermore, the reconfigured coupling strengths were found to relate to P300 amplitudes, which were then regarded as features to train a classifier to differentiate the brain states and the high and low P300 groups with an accuracy of 100% and 77.78%, respectively. The findings of this article help us understand the updates in functional connectivity from resting state to the oddball task, and the reconfigured network structure has the potential for the selection of good subjects for P300-based brain-computer interface.
BibTeX
@article{li2021reconfiguration,
  author = {Fali Li and Chanlin Yi and Yuanyuan Liao and Yuanling Jiang and Yajing Si and Limeng Song and Tao Zhang and Dezhong Yao and Yangsong Zhang and Zehong Cao and Peng Xu},
  title = {{Reconfiguration of Brain Network Between Resting State and P300 Task}},
  journal = {IEEE Trans. Cogn. Dev. Syst.},
  year = {2021},
  url = {https://openreview.net/forum?id=BcydGphtBgk},
}
Figure