A normative model-based assessment framework for large-scale, multi-site EEG data

Area of Science:

• Neuroscience
• Biomedical Engineering
• Psychology

Background:

• Electroencephalography (EEG) offers objective assessment, surpassing subjective methods.
• Current EEG evaluations often oversimplify continuous neural dynamics into discrete states.
• Large-scale, multi-site EEG studies face challenges like batch effects and montage inconsistencies.

Purpose of the Study:

• To develop a robust framework for harmonizing large-scale, multi-site EEG data.
• To improve the reliability and efficiency of attention assessment using EEG.
• To address limitations of existing EEG evaluation methods by accommodating continuous neural dynamics.

Main Methods:

• Constructed normative models using EEG features from 1212 young individuals.
• Employed quantile ranks, elastic net regression, and support vector regression for attention modeling.
• Assessed test-retest reliability and generalizability of the developed framework.

Main Results:

• Identified significant statistical differences in attention performance between top and bottom participant groups (q < 0.05).
• EEG features showed distinct patterns related to accuracy and reaction time in distributed and focused attention tasks.
• The normative model demonstrated superior predictive performance, enhanced stability, and interpretability with high reliability (ICC > 0.9).

Conclusions:

• A normative model-based framework was proposed to harmonize multi-site EEG data.
• The framework enables efficient and reliable attention assessment.
• This approach shows potential for broader applications in EEG-based research and clinical practice.