Prediction of children age using patterns of auditory event-related potentials

Many studies showed a relationship between au-ditory event-related potentials (ERP) and brain development. Particular attention was paid to N1, the ERP component, which reflects auditory detection and discrimination, and N2, which in-dicates attention allocation and phonological analysis. This study applied logistic regression algorithm (LRA) with regularization to classify children and adolescents to three age groups - 4–6, 7–12 and 12–18 years by individual amplitudes and latencies of P1, P2, N1 and N2, components of auditory ERPs. The Kruskal-Wallis H-test indicated significant group differences in amplitudes of P1 with p < 6.67e-
⁵
, N1 with
$p$
< 2.34e-
⁷
, N2 with p < 5.95e-
⁵
and latencies of P1 with p < 1.14e-
⁵
and P2 with p < 2.57e-
³
• The classification accuracy between all three age groups was 0.63, macro-average and micro-average area under the curve (AUC) - 0.79 and 0.70. Classification accuracy of one age group against the rest was 0.71, 0.45 and 0.75 with A|UC - 0.91, 0.50 and 0.76 for 4–6, 7–12 and 12–18 age group, respectively. The most informative feature for the correct age group recognition, which was reflected in the largest positive change of odds ratio, was amplitude of N1. The obtained results confirm association of the N1 amplitude with brain maturation and demonstrate sensitivity of LRA in detection of individual age using patterns of auditory ERPs.