Constructing a Consciousness Meter Based on the Combination of Non-Linear Measurements and Genetic Algorithm-Based...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Clinical Neurology

Background:

• Evaluating consciousness is crucial in neuroscience and clinical settings.
• Current methods lack a systematic framework for quantifying altered consciousness.
• Differentiating states like coma, anesthesia, MCS, and wakefulness remains challenging.

Purpose of the Study:

• To build a systematic framework for quantifying consciousness.
• To differentiate between coma, general anesthesia, minimally conscious state (MCS), and normal wakefulness (NWC).
• To develop EEG-based indices for both level and content of consciousness.

Main Methods:

• Analyzed EEG data from frontal channels in four groups: coma, MCS, general anesthesia, and NWC.
• Utilized non-linear methods (permutation entropy, sample entropy, Lempel-Ziv complexity, detrended fluctuation analysis) and relative power.
• Employed a genetic algorithm-SVM for classification and multivariable linear regression for index construction.

Main Results:

• Permutation entropy (PE) significantly differentiated all four states.
• Sample entropy (SampEn) and permutation Lempel-Ziv complexity (PLZC) best differentiated altered consciousness content.
• Relative power (RP) and PE best differentiated consciousness levels.
• A multi-dimensional index integrating PE, PLZC, SampEn, and DFA achieved 92.3% classification accuracy using GA-SVM.

Conclusions:

• Multi-dimensional EEG measurements, particularly PE, SampEn, PLZC, and DFA, show promise.
• Combining these measures with GA-SVM offers a robust framework for quantifying consciousness.
• This approach advances the ability to systematically measure altered states of consciousness.