FBCPM: A Filter Bank Connectome-Based Predictive Modeling Framework for EEG Signals

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• The human brain connectome is vital for cognitive functions.
• Connectome-based prediction modeling (CPM) predicts behavior, but EEG application is limited by data complexity.
• Existing methods struggle with EEG frequency information diversity.

Purpose of the Study:

• To develop a novel filter bank CPM (FBCPM) framework for individual prediction using narrowband EEG functional connectivity (FC).
• To address limitations in applying CPM to complex EEG frequency data.
• To enhance individual-level prediction accuracy in cognitive neuroscience.

Main Methods:

• Developed a filter bank CPM (FBCPM) framework utilizing narrowband EEG functional connectivity (FC).
• Evaluated FBCPM on four independent datasets (280 healthy subjects, 392 EEG recordings) during the psychomotor vigilance test (PVT).
• Predicted mean reaction time (RT) and time-on-task (TOT) effects, validating against benchmark approaches and across diverse datasets.

Main Results:

• FBCPM achieved notable prediction accuracy for mean RT, outperforming four benchmark methods.
• The framework demonstrated robustness across hyper-parameters and generalizability to independent datasets with varying settings.
• FBCPM showed satisfactory performance for predicting time-on-task effects (ΔRT, TOTslope).

Conclusions:

• FBCPM offers a significant methodological advancement for EEG-based individual prediction.
• The framework leverages narrowband EEG FC for improved cognitive performance prediction.
• Findings support FBCPM's potential for practical applications in cognitive neuroscience.