Epilepsy is a chronic neurological disorder that affects approximately 50 million people worldwide. These intractable seizures postures are a serious risk of injury, restrict the self-sufficiency and mobility of a person. While medication and surgery can, to some degree, relieve the symptoms, these treatments fail to help all patients. Population wide, approximately 1% is suffering from epilepsy. Current treatment is inefficient on about 30% among those people [1].
The sudden and apparently unpredictable nature of epileptic seizures is one of the most disabling aspects of epilepsy. Therefore, there has been a growing research interest, especially within the last 30 years, in seizure detection and prediction from EEG recordings. It is believed that developing a method capable of predicting the occurrence of seizures from the electroencephalogram (EEG) of epilepsy patients will open new therapeutic possibilities. Since the 1970s, studies on the predictability of seizures have advanced from preliminary reports of seizure precursors to controlled studies implementing prediction algorithms to continuous daily EEG recordings. Most of the seizure detection algorithms are patient specific; that is, they are applied to the patient for which training data are extracted. This is intuitively meaningful as each patient has a different nature for its EEG signals. On the other hand, efforts have also been made to develop seizure prediction algorithms. Although most of the studies published in the 1990s and around the turn of millennium yielded rather encouraging results, more recent evaluations could not reproduce their findings, thus raising a debate about the validity and reliability of those studies.
Considering the nature of epilepsy, there are many types of seizures. This can sometimes be a challenging task to address, especially when one considers that some of the epileptic syndromes are difficult to be characterized and therefore are being classified as particular category [2]. In addition, understanding of the underlying mechanisms leading to seizures and the origin of a seizure in each case is still under investigation.
Up to now, several investigations based on nonlinear time series analysis have been carried out on intracranial and surface EEG that provided data with promising results. It has been claimed that seizures can be predicted at least 20 minutes beforehand [3], maybe up to 1 hour and 30 minutes before the onset of temporal lobe epilepsy.
There are numerous techniques and algorithms for analysis and classification of bio signals, 1 or 2- dimensional, in time or frequency distribution. The majority of algorithms follow the structure of selecting and applying feature extraction methods on the EEG signal and using classification methods to conclude the diagnosis. The seizure detection method of epilepsy can be made on a single or multi-channel way. Single-channel seizure detection requires selecting the channel receiving the strongest EEG signal collected from the closest point to the seizure spot. Some algorithms create models for normal and abnormal EEG signals of the patients and use these models in the training process. EEG feature extraction has been studied from early time and lots of advanced techniques and transformations have been proposed for accurate and fast EEG feature extraction. For example, discrete Fourier transform (DFT) and discrete wavelet transform (DWT) have found popularity in seizure detection and prediction applications also the power spectral features, higher order spectral methods, nonlinear transformations such as Lyapunov exponents have been used as appropriate sources for feature extraction. The classifying methods which have been proposed during the last decade include Fuzzy Logic methods, Artificial Neural Network, Hidden Markov Model, Support Vector Machines, Cluster analysis, with each approach exhibiting its own advantages and disadvantages. Tzallas et al. presented a classification of EEG seizure detection methods into pattern recognition, morphological analysis, parametric, decomposition, clustering and data mining methods [4]. The combination of signal processing with the electronic devices serves as a primary root for the development of various biomedical applications. Ideally, a seizure prediction method/algorithm has to guess an impending epileptic seizure by raising an alarm in advance of the seizure onset [5]. A perfect prediction method will be able to show the exact point in time when a seizure will occur.
[1] D. J. Thurman, E. Beghi, C. E. Begley, A. T. Berg, J. R. Buchhalter, D. Ding, D. C. Hesdor_er, W. A. Hauser, L. Kazis, R. Kobau, et al., “Standards for epidemiologic studies and surveillance of epilepsy,” Epilepsia, vol. 52, no. s7, pp. 2{26, 2011. [2] P. Aswathappa and M. Ambar, “Separation of background activity of epileptic eeg using artificial wavelet transforms techniques,” 2014. [3] Moghim, Negin, and David W. Corne. “Predicting Epileptic Seizures in Advance.” PloS one 9.6 (2014): e99334. [4] T. N. Alotaiby, S. A. Alshebeili, T. Alshawi, I. Ahmad, and F. E. A. El-Samie, \Eeg seizure detection and prediction algorithms: a survey,” EURASIP Journal on Advances in Signal Processing, vol. 2014, no. 1, p. 183, 2014. [5] K. Lehnertz and C. E. Elger, “Can epileptic seizures be predicted? evidence from nonlinear time series analysis of brain electrical activity,” Physical review letters, vol. 80, no. 22, p. 5019, 1998.
