TY - GEN
T1 - Comparing two time-scale and time-frequency based methods in newborns' EEG seizure detection
AU - Zarjam, Pega
AU - Mesbah, Mostefa
AU - Boashash, Boualem
PY - 2007
Y1 - 2007
N2 - In this research, two different approaches for detecting seizure patterns in newborns' Electroencephalogram (EEG) signals are compared. The first proposed approach is a time-frequency (TF) based method, in which, the discrimination between seizure and non-seizure states is based on the TF distance between the consequent segments in the EEG signal. Three different TF measures and three different reduced time-frequency distributions (TFD) are used in this study. The second proposed approach is a discrete wavelet transform (DWT) based method, in which, the detection scheme is based on observing the changing behavior of few statistical quantities of the wavelet coefficients (WCs) of the EEGs at various scales. These statistics form a feature set which is fed into an artificial neural network (ANN) classifier to organize the EEG signals into seizure and non-seizure activities. The proposed methods are tested on the EEG data acquired from three neonates with ages under two weeks. The empirical results validate the suitability of the two proposed methods in automated newborns' seizure detection. The results present an average seizure detection rate (SDR) of 96% and false alarm rate (FAR) of 5% using Kullback-Leibler measure which outperforms the other two distance measures and the DWT based method.
AB - In this research, two different approaches for detecting seizure patterns in newborns' Electroencephalogram (EEG) signals are compared. The first proposed approach is a time-frequency (TF) based method, in which, the discrimination between seizure and non-seizure states is based on the TF distance between the consequent segments in the EEG signal. Three different TF measures and three different reduced time-frequency distributions (TFD) are used in this study. The second proposed approach is a discrete wavelet transform (DWT) based method, in which, the detection scheme is based on observing the changing behavior of few statistical quantities of the wavelet coefficients (WCs) of the EEGs at various scales. These statistics form a feature set which is fed into an artificial neural network (ANN) classifier to organize the EEG signals into seizure and non-seizure activities. The proposed methods are tested on the EEG data acquired from three neonates with ages under two weeks. The empirical results validate the suitability of the two proposed methods in automated newborns' seizure detection. The results present an average seizure detection rate (SDR) of 96% and false alarm rate (FAR) of 5% using Kullback-Leibler measure which outperforms the other two distance measures and the DWT based method.
KW - Discrete wavelet transform
KW - EEG
KW - Reduced interference distributions
KW - Seizure
KW - Time-scale/frequency
UR - http://www.scopus.com/inward/record.url?scp=60349091035&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=60349091035&partnerID=8YFLogxK
U2 - 10.1109/ICSPC.2007.4728635
DO - 10.1109/ICSPC.2007.4728635
M3 - Conference contribution
AN - SCOPUS:60349091035
SN - 9781424412365
T3 - ICSPC 2007 Proceedings - 2007 IEEE International Conference on Signal Processing and Communications
SP - 1579
EP - 1582
BT - ICSPC 2007 Proceedings - 2007 IEEE International Conference on Signal Processing and Communications
T2 - 2007 IEEE International Conference on Signal Processing and Communications, ICSPC 2007
Y2 - 14 November 2007 through 27 November 2007
ER -