Identification of Epileptic Seizures Utilising a Computationally Powerful Spiking Neuron

Epilepsy is considered a complex neurological condition that causes the central nervous system to malfunction, which results in seizures. Therefore, in-depth identification of the condition with high accuracy is required, and Electroencephalogram (EEG) plays a vital role. To classify seizures from the EEG signals, a computationally efficient Spiking Neural Networks (SNN) classifier inspired by WOLIF that utilises the error-function adjusted by the Grey Wolf Optimizer (GWO) and derived from spiking neurons with Leaky Integrate and Fire (LIF) is employed in this paper. It works on the principle of SNN, which is computationally very efficient and does not need hidden layer(s) even though the data is highly non-linear, such as EEG signals. The features are extracted from the raw EEG signals utilising the Discrete Wavelet Transform (DWT). Then, the information is encoded using the population coding technique since SNN works with temporal information. This study makes use of the epileptic seizures dataset from Bonn University. Four binary classifications between healthy and seizure classes from five data sets were done. Unlike WOLIF, \(\alpha\) $\alpha$ -kernel is used as the synapse model rather than the double-decaying kernel function, improving the overall accuracy. The results shown by the proposed approach are very appealing if both computational efficiency and accuracy are considered the metrics for analysis, which is very important.