Seizure Detection from the Time-Frequency Based Multichannel Newborn EEG Signal through the Application of Advanced Noise Filtering and Classification Methods

Abstract

Worldwide survey from health department indicates that approximately 50 million people are currently affected with epilepsy, which is caused due to seizure. Among the top four common neurological diseases in the United States after migraine, stoke and Alzheimer‟s disease is epilepsy. Internationally, a vague count of average epilepsy patient‟s each year is 2.4 million. Electroencephalogram (EEG) monitors the electrical activity inside our brain, which is due to the movement of neurons. It is used for the in time detection of various diseases in neonatal and adults, such as a seizure. EEG displays the signals received by our brain from all body parts. Any sort of seizure that is likely to occur in our body or brain can be seen through EEG. As only time or frequency analysis is not sufficient to clearly depict the non-stationery electrical activity. Time-frequency (TF) analysis is helpful for the dynamic property of EEG signals. The signal is affected by different artefacts, which produce false detections. Distinct research has been carried out in this field. Various methods have been tested for extracting features of the EEG signal; also classifiers, such as Neural Networks and support vector machine (SVM), were applied for the detection purpose. TF representation provides a wealth of information about the underlying EEG in temporal as well as spectral domains. This work will use novel image-processing methods and machine learning procedures for the feature extraction stages to improve the accuracy (in terms of both sensitivity and specificity) of existing methods. The understanding and assessment about epilepsy is still a long way ahead. Epilepsy awareness and its care among the masses are below a considerate level. This work will assist the doctors in the field of neurology to improve the timely detection of seizures.