Role of High Frequency Electroencephalographic Oscillations during sleep in pediatric epileptic patients

Epilepsy is one of the most common neurological disorders in childhood and is characterized by recurrent seizures resulting from abnormal neuronal activity. Electroencephalography (EEG) plays a central role in its diagnosis and monitoring, particularly during sleep, when epileptiform activity can become more evident. Among advanced quantitative EEG measures, high-frequency oscillations (HFOs) have gained increasing attention as potential biomarkers of epileptic activity. However, their relationship with other properties of the EEG signal, such as signal complexity, remains less explored in pediatric patients. The aim of this thesis was to investigate the role of HFOs during sleep in pediatric patients with epilepsy, with particular attention to their relationship with quantitative changes in EEG complexity. A systematic review was conducted to examine the main quantitative EEG biomarkers investigated during sleep in pediatric epilepsy and to identify current methodological approaches and research gaps. The review highlighted growing interest in HFOs, entropy-based measures, functional connectivity, and other quantitative EEG approaches, together with considerable methodological heterogeneity across studies. Based on this background, an experimental analysis was performed using an open dataset of overnight EEG recordings from 30 pediatric patients with epilepsy. The analysis focused on N3 sleep, during which HFO occurrence is particularly relevant. The continuous EEG recordings were preprocessed and divided into non-overlapping 4-s windows. Fourteen complexity measures, selected from the literature to characterize different aspects of EEG dynamics, were then computed and analyzed in relation to HFO occurrence. Seven of the fourteen measures showed significant differences between HFO and No-HFO activity using the Wilcoxon signed-rank test with false discovery rate (FDR) correction for multiple comparisons.. Katz Fractal Dimension increased during HFO activity, whereas Petrosian Fractal Dimension, Permutation Entropy, Hjorth Complexity, Shannon Entropy, Tsallis Entropy, and Rényi Entropy decreased. A subsequent temporal analysis examined distant Baseline, Pre-HFO, HFO, and Post-HFO conditions using the Friedman test, followed by post-hoc Wilcoxon signed-rank tests with FDR correction. The most consistent changes were observed during HFO activity itself, while no significant differences were detected between Pre-HFO and distant Baseline or between Post-HFO and distant Baseline. Finally, the fourteen complexity measures were evaluated for subject-independent classification. Three supervised models were tested using leave-one-subject-out (LOSO) validation. Classification of HFO activity versus distant Baseline showed modest performance, with a maximum balanced accuracy of approximately 53.7%. Pre-HFO activity could not be reliably distinguished from distant Baseline, with a balanced accuracy of approximately 49.3% and an ROC-AUC close to 0.50. Overall, the results show that HFO occurrence during N3 sleep is associated with measurable changes in specific aspects of EEG complexity. However, these changes were mainly observed during HFO activity and showed limited generalization across subjects. No robust complexity pattern preceding HFO occurrence was identified with the measures and temporal resolution considered in this study. These findings suggest that EEG complexity can provide complementary information for the characterization of HFO-related activity, while its integration with other quantitative EEG features may be necessary to obtain more robust and generalizable markers in pediatric epilepsy.