Electroencephalogram Data-Based Analysis of Paroxysmal Slow Wave Events Patterns in Brain Pathologies

Slowing of brain activity observed in electroencephalography (EEG) recordings is normal under resting conditions such as sleep. However, a recent series of studies described a new pattern of cortical slowing in patients with epilepsy and Alzheimer's disease, consisting of transient paroxysmal slowing of the network. These paroxysmal slow wave events (PSWEs) were defined with median power frequency (MPF) less than 6 Hz and duration longer than 5 s. In this research, we are using clinical EEG recordings from the Temple University and Bonn University databases. We aim to: (1) Characterize the temporal and spatial characteristics of PSWEs in patients with epilepsy; (2) Identify PSWEs features that will assist in the diagnosis of epilepsy, specifically drug-resistant epilepsy; (3) Identify the sensitivity and specificity of selected combination of features that will help in differentiating between patients with epilepsy and other brain disorders (e.g. Alzheimer's disease, mood disorders). To this end, we trained machine learning models using the Temple University dataset, achieving a classification accuracy of 78.26% in distinguishing between epilepsy and non-epilepsy patients. Moreover, by training the models on the Bonn University database, we achieved an accuracy of 91.67% in classifying drug-resistant epilepsy versus seizure-free groups.Clinical Relevance- PSWEs serve as a potential biomarker for early epilepsy diagnosis and risk assessment, aiding in distinguishing isolated seizures from chronic epilepsy. Their association with neurodegenerative and cognitive disorders further highlights their clinical significance in neurological disease monitoring.