PROJECT SUMMARY/ABSTRACT Epilepsy is a devastating neurological disorder affecting 65 million people worldwide and more than 3 million people in the United States. Seizures impairing consciousness severely affect quality of life of people with epilepsy. However, the dysfunction in brain networks associated with these seizures is not fully understood. Previous studies explored limited set of features that can partially explain the underlying network dysfunction. Additionally, none of these studies investigated the association between ictal and interictal dysfunctional network connectivity patterns. Recently, we studied EEG characteristics of seizures impairing consciousness and developed a promising machine learning approach to predict impaired consciousness in absence epilepsy. The current proposal extends more broadly to other seizure disorders and interictal cognitive deficits. My central hypothesis is that seizures impairing consciousness are associated with both transient and chronic dysfunction in the same networks and that the characteristics of the transient dysfunction can be leveraged to develop a clinical tool to predict impaired consciousness during seizures based on scalp EEG without the need for behavioral testing. To address this hypothesis, a large EEG dataset of seizures impairing and sparing consciousness (impaired and spared seizures), as well as interictal recordings will be created. The spatiotemporal and spectral characteristics of behaviorally impaired and spared seizures will be investigated (Aim 1). A clinical tool based on conventional machine learning and deep learning methods will be developed to predict impaired consciousness based on pre-ictal, ictal, and post-ictal EEG and transfer learning will be used to enable model generalization across hospital settings (Aim 2). To characterize the relation between ictal and interictal connectivity dysfunction, we will investigate functional and effective connectivity patterns in relation to impairment during and between seizures (Aim 3). It is anticipated that spatiotemporal and spectral analyses will reveal statistically significant differences between impaired and spared seizures during pre-ictal, ictal, and post-ictal periods (Aim 1). We expect the developed clinical tool will be capable of predicting impairment in consciousness regardless of seizure type and origin and will perform efficiently across different hospital settings (Aim 2). Furthermore, the predictive models are expected to identify innovative information about evolution and stability of neural representations underlying impaired and spared seizures (Aim 2). Finally, we hypothesize that recurrent transient dysfunction in connectivity due to impaired seizures will be associated with chronic connectivity dysfunction in the same networks (Aim 3). A detailed understanding of transient and chronic network dysfunction associated with seizures may lead to development of novel biomarkers and targets for therapeutic neu...