Quantifying Consciousness for Alzheimer's Disease Diagnosis through Electroencephalogram Processing

Electroencephalograms (EEG) could serve as a biomarker for Alzheimer's disease (AD), leveraging its indicative features. However, the theoretical implications of EEG in the clinical diagnostic routine for AD remain underexplored. This study introduces an approach to analyze EEG signals within the computational framework proposed by integrated information theory, aiming to process EEG signals from a consciousness perspective for AD diagnosis. Initially, we quantified consciousness as an indicator, termed the Φ value, by analyzing a public EEG dataset comprising recordings from healthy elderly individuals and people with AD. Subsequently, we trained machine learning models using Φ values combined with additional information to classify the two population groups. Our best classification accuracy surpassed that of previous benchmark research. This study underscores the promising potential of EEG signals in measuring consciousness for diagnosing AD. As future endeavors, we outline three key aspects for exploration, including frequency domain analysis, electrode combination complexities, and temporal resolution.