Binary Classification of Consciousness Using Cerebral Blood Flow and EEG Features

Area of Science:

• Neuroscience
• Critical Care Medicine
• Biomedical Engineering

Background:

• Assessing consciousness in neurocritical care is challenging due to sedation and treatment effects.
• Electroencephalography (EEG) is commonly used but provides limited insight into the neurovascular unit.
• There is a need for improved methods to evaluate consciousness in severe brain injury.

Purpose of the Study:

• To evaluate if combining cerebral blood flow (CBF) features with EEG enhances the binary classification of consciousness.
• To determine the effectiveness of multimodal neuromonitoring in neurocritical care settings.
• To explore new tools for real-time neurovascular monitoring.

Main Methods:

• Retrospective analysis of 35 adult patients with severe brain injury undergoing multimodal neuromonitoring.
• Segmentation of signals into 30-min windows and analysis of CBF low-frequency bands and EEG band powers.
• Utilized a random forest (RF) model with K-fold cross-validation, excluding highly correlated features (r > 0.8).

Main Results:

• Multimodal feature combinations significantly improved consciousness classification compared to EEG alone.
• The optimal combination (EEG ADR, total EEG power, CBF Band V) achieved a ROC-AUC of 0.86 and 82% accuracy.
• This multimodal model showed up to a 69% improvement over EEG-only models and performed well on noninvasive optical blood flow data.

Conclusions:

• Combining EEG and CBF metrics, especially low-frequency oscillations in perfusion, enhances consciousness classification in critically ill patients.
• This multimodal approach may support future bedside tools for real-time neurovascular monitoring.
• Findings can aid in treatment and rehabilitation decision-making for patients with severe brain injury.