Abstracts

Spectral Entropy of Regional Maximum Provides Evidence of Seizure Precursors in Patients with MTLE

Abstract number : 1.125
Submission category :
Year : 2001
Submission ID : 2756
Source : www.aesnet.org
Presentation date : 12/1/2001 12:00:00 AM
Published date : Dec 1, 2001, 06:00 AM

Authors :
G. Vachtsevanos, Ph.D., ECE, Georgia Institute of Technology, Atlanta, GA; M. D[ssquote]Alessandro, MSEE, ECE, Georgia Institute of Technology, Atlanta, GA; R. Esteller, Ph.D., Dpto. Tecnologia Industrial, Universidad Simon Bolivar, Caracas, Edo. Miranda,

RATIONALE: Under normal conditions, IEEG signals of patients with MTLE appear to exhibit random behavior; however, damaged brain regions may exhibit persistent abnormal behavior. This behavior is not obvious through visual observation of the raw IEEG signal, but may be evident when analyzing features that exploit this abnormality.
METHODS: The IEEG recordings obtained from three patients with refractory MTLE were investigated by analyzing the spectral entropy (SE) of the IEEG regional maximum to find evidence of prediction indicators. 15 minute segments of all monopolar IEEG data spanning 10 minutes before and 5 minutes after the electrographic onset were clipped. Baseline data were clipped at least 3 hours before or after seizures, and also included 15 minute segments. A total of 13 seizures and 19 baselines were analyzed. A spatial mean remover that determines the mean across all channels and removes this common trend from the data, was used; and to remove the line noise, a 60 Hz notch filter was used. A sliding window of 7.5 seconds with 0.5 seconds of overlap was selected for this analysis, and the spectral entropy (SE) was extracted, then the data normalized with respect to the minimum SE.
RESULTS: It is clear that a seizure results in the synchronization of the disordered background activity exhibited in normal IEEG signals. In addition, the SE feature indicates that an increased order is present in the focal region not only during the seizure, but constantly. This finding may provide a means for understanding the most effective method for analyzing the IEEG signals for prediction indicators. Using a simple threshold detector, 10 out of 13 seizures provided prediction indicators during the 10 minutes preceding UEO, while the baselines remained well above the threshold.
CONCLUSIONS: The main objective of entropic feature analysis is to quantify regularity and order in the signal. The algorithm presented in this research exploits any potential order not only during the seizure, but prior to electrographic onset, thereby providing indicators that a seizure is imminent. There is evidence that a region other than the focal region may provide indicators that may assist in predicting seizures.
Support: IntelliMedix; American Epilepsy Society; Epilepsy Foundation; Whittaker Foundation.
Disclosure: Salary - Drs. Echauz and Esteller are salaried employees of IntelliMedix. Equity - Drs Echauz, Esteller, Litt, and Vachtsevanos are cofounders of IntelliMedix.