A data-driven process for the development of an eyes-closed EEG normative database

A normative database constitutes a representative sample of a neurologically and clinically healthy population. The practical utility of a normative EEG database is to evaluate the clinical status of a subject whose EEG patterns statistically diverge from average population patterns. These normative data are daily used in clinical practice and in the evaluation of therapeutical interventions. The main obstacle of all normative databases developed to date is inter-individual variability. Such difficulty has been addressed by stratifying the population by age and then using regression in the EEG groups to bound variability, which is always an approximation. This paper describes the first data-driven EEG normative database that explicitly deals with EEG variability by stratifying the population based on their EEG patterns. The database has been constructed for 84 subjects in eyes-closed condition and has been validated by cross validation, leading to a global specificity of 100%.