AUTOMATED REAL-TIME SEIZURE DETECTION ALGORITHM
Abstract number :
2.153
Submission category :
Year :
2003
Submission ID :
3923
Source :
www.aesnet.org
Presentation date :
12/6/2003 12:00:00 AM
Published date :
Dec 1, 2003, 06:00 AM
Authors :
Wanpracha Chaovalitwongse, James C. Sackellares, Deng-Shan Shiau, Paul R. Carney, Panos M. Pardalos, Leonidas Iasemidis Industrial and Systems Engineering, University of Florida, Gainesville, FL; Neurology, University of Florida, Gainesville, FL; Neurosci
Epilepsy consists of more than 40 clinical syndromes affecting 1% of population. The hallmark of epilepsy is recurrent seizures. The importance of automated real-time detection and quantitative analysis of epileptic seizures has been recognized by physicians and researchers. Consequently, the challenge is to develop such a seizure detection algorithm that addresses the nonstationary and high complexity of seizures and computes in real time with high sensitivity and specificity. In this study, we first employ Teager[rsquo]s formulation to estimate the signal energy of electroencephalograms (EEG[rsquo]s) recorded in temporal lobe epilepsy. This energy measure and the proposed new detection technique constitute an automated real-time seizure detector.
Continuous 26-channel long-term intracranial EEG recordings previously obtained in 4 patients with medically intractable partial seizures were used to test the automated real-time seizure detector. Patient 1 had 15 seizures in a 10-day recording; patient 2 had 8 seizures in 6 days; patient 3 had 7 seizures in 12 days; and patient 4 had 5 seizures in 1 day. The seizure detector involved the following steps: (1) calculate the average energy of EEG, based on Teager[rsquo]s algorithm, over all electrodes for each sequential 10.24 second epoch, (2) use the average energy level during the ictal period of the first seizure as a preset threshold, (3) a detection is declared when the current average energy is higher than the mean value of the average energy level of the first seizure and average energy level of previous 10 points (approximately 1.7 minutes), (4) the detection was correct if a seizure occurred in 2 minutes before or after the detection.
The sensitivity of the automated real-time seizure detector in patients 1, 2, 3, and 4 was 92.86%, 100.00%, 83.3% and 100.00%, respectively. The average false detection rate (false per hour) was 0.09, 0.06, 0.11, and 0.74% for patients 1-4 respectively.
Based on a simple algorithm derived that enables on-the-fly calculation of the energy, this automated real-time seizure detector can detect seizures with performance characteristics that could have practical clinical utility. The detector could be incorporated into clinical EEG monitoring system or, incorporated into an implantable chip and used to activate timed physiological or pharmacological interventions.
[Supported by: the NIH, NSF, VA, and Whitaker research grants]