We developed EEG monitoring techniques designed to detect non-convulsive status epilpeticus (NCSE) in non-responsive patients in the lCU and emergency room. This patient population typically does not receive brain monitoring, even though they are at risk for NCSE.

Our NCSE Detector is intended monitor brain state, by analysis of EEG, and provide warnings that the patient's brain is in a dangerous state of ongoing epileptic seizure activity that may require prompt treatment to avoid permanent damage.

To accomplish this, we developed a hybrid system based on a combination of two Artificial Neural Networks (ANN) and rule-based algorithms, resulting in a NCSE yes/no decision for each 1-minute epoch.

In our test database of ICU and emergency room EEG records (10 patients, 241 hours), we correctly classified NCSE 95% of the time, and non-NCSE 96% of the time, as compared to expert human scoring.

The NCSE Detector also creates a trend of paroxysmal epileptiform activity (PEA), The PEA trend tracks the percent time PEA in a sliding 1-minute window. We believe the PEA index can provide useful complementary real time information about brain state that is absent from other global EEG measures.

Project details:

Contact: Georgiy R. Minasyan (gminasyan@chattenassociates.com)

Publications:

Minasyan G.R., Chatten J.B., Harner R.N. "Detection of epileptiform activity in unresponsive patients using ANN" Neural Networks, 2009. International Joint Conference on Neural Networks, Atlanta, 2009

Other:

Minasyan, G.R., Chatten, J., Harner, R.N., "Automatic Detection of Non-Convulsive Status Epilepticus", American Epilepsy Society, Abst. 1.123, 2010 Annual Meeting