Epileptic seizure classification using feed forward neural network based on parametric features

Citation data:

International Journal of Pharmaceutical Research, ISSN: 0975-2366, Vol: 10, Issue: 4, Page: 189-196

Publication Year:
2018

No metrics available.

DOI:
10.31838/ijpr/2018.10.04.046
Author(s):
T. Rajendran; K. P. Sridhar
Tags:
Pharmacology, Toxicology and Pharmaceutics
review description
Globally, epilepsy is a severe neural disorder occurring among 0.6-0.8% of the population. The formation of pattern-change from normal to disturbed factors that all gets triggered at once is called seizure. Many researchers introduced different techniques, but the problem of detecting epileptic seizures remains unsolved. This paper presents a new technique for detection of epileptic seizure-based Electroencephalogram (EEG) signals. The detection scheme adapts the non-invasive measure of the brain’s electrical activity by placing the electrodes on the scalp. The collection of such electrical activity and diagnosing is a complex task because the brain is composed of numerous classes with numerous overlying features. Feature extraction based on parametric and non-parametric method is employed to extract the features vectors from EEG signals. The extracted features are forwarded to machine learning algorithms. Feed Forward Neural Network (FFNN) is implemented to detect the epileptic seizure. The performance results are evaluated by comparison of previous Modified Back Propagation Neural Network, Multilayer perceptron neural network, combined neural network, Probabilistic Neural Network and FFNN methods with respect to feature extraction in terms of accuracy, specificity and sensitivity. The FFNN has the higher classification accuracy as 97.23% demonstrates that it has a great potentiality of the real-time epileptic seizure detection.