Derivation of an intrinsic brain activity biomarker for the earliest prediction of cognitive decline

Early identification of cognitive decline in the elderly is essential for timely and effective intervention. Current neuroimaging and clinical biomarkers have largely focused on patients with mild cognitive impairment (MCI) and dementias, and though effective, have not been demonstrated to be sensitive to the earliest signs of functional abnormalities. The electroencephalogram (EEG), with a millisecond time base, allows a precise exploration of synaptic dysfunction even at the initial stages of the damage. This study introduces a novel non-invasive EEG-based biomarker to predict cognitive decline of participants showing only subjective cognitive impairment (SCI). Baseline EEG recordings from 88 SCI participants who had yearly assessment of cognitive changes and staging for 5–7 years were included for study. Quantitative EEG (qEEG) features were calculated from eyes closed resting surface EEG, including those reflecting broad band power spectra, connectivity, and complexity. Machine learning (ML) classifiers were trained on the qEEG features to estimate the likelihood of future progression to MCI or dementia. A feature selection pipeline optimized predictive accuracy of the ML algorithms while reducing the number of features. Prediction performance for the biomarker was over 80% in accuracy, with an area under the curve of 0.90. Independent validation on two external cohorts confirmed the biomarker’s robustness. Dominant contributors to the final locked models aligned with existing literature on neurodegenerative disorders. Features contributing most were those reflecting disruption in neuronal transmission (phase lag and asymmetry) with abnormalities in the alpha and theta frequency bands. Changes in measures of connectivity in the subjective cognitive impairment (SCI) population provide evidence of changes in neuronal transmission within frontal networks. These findings suggest that these EEG-based brain activity biomarkers are reflective of the earliest signs of brain dysfunction. The qEEG features contributing to the biomarkers reflect the underlying physiological mechanisms related to neurodegeneration. Furthermore, the non-invasive and cost-effective nature of EEG makes it a promising tool for clinical implementation, allowing for early risk assessment, disease monitoring, and intervention planning.