Title :
An EEG feature-based diagnosis model for epilepsy
Author :
Luo, Kun ; Luo, Donghui
Author_Institution :
Dept. of Neurosurg. of the First Affiliated Hosp., Xinjiang Med. Univ., Urumchi, China
Abstract :
Electroencephalogram (EEG) is the most important clinical tool in evaluating patients with epilepsy. However, the EEG definite patterns correlated to various types of epilepsy are still unclear. In this paper, six features of EEG signal are extracted to construct an artificial neural network model of classifying controls and patients with epilepsy. The ROC-score (area under curve) of the model is 88.3%. SD of autocorrelation, Hurst indexes, and periodicity have a good capacity in identifying epilepsy.
Keywords :
diseases; electroencephalography; feature extraction; medical signal processing; neural nets; patient diagnosis; sensitivity analysis; EEG; Hurst indexes; ROC; area under curve; artificial neural network model; autocorrelation; diagnosis; electroencephalogram; epilepsy; feature extraction; periodicity; artificial neural network (ANN); electroencephalogram (EEG); features;
Conference_Titel :
Computer Application and System Modeling (ICCASM), 2010 International Conference on
Conference_Location :
Taiyuan
Print_ISBN :
978-1-4244-7235-2
Electronic_ISBN :
978-1-4244-7237-6
DOI :
10.1109/ICCASM.2010.5619259
