ABSTRACT
Epilepsy is a communal neural ailment categorized by loss of consciousness and convulsions (seizure). Epilepsy happens erratically and randomly due to temporary electrical disturbance of the brain, which can be observed using an electroencephalogram (EEG). Manual detection of epilepsy is time consuming due to bulky recording and also needs an expert technician, who may be prone to human error. Automatic accurate detection and forecast of epilepsy help to overcome death hazards (SUDEP, which stands for sudden unexpected death in epilepsy) and also improves the quality of life for both patients as well as a caretaker. Selection and extraction of the specific EEG features, among others, time domain-like energy, entropy, kurtosis, and skewness; frequency domain-like power spectral density (PSD), intensity-weighted bandwidth, and spectral entropy are crucial parts due to noise and artifacts. Researchers are working on different seizure detection/prediction techniques and classification techniques to identify different phases of epilepsy, like pre-ictal, ictal, and interictal states. This chapter gives an in-depth look at classification techniques, from conventional methods to deep neural network-based techniques along with the shortcomings of each technique.
