Enhanced Classification of Epileptogenic and Nonepileptogenic EEG Signals using ANN-FDM

EEG signals, which show the electrical activity of the brain, are considered to be the most significant physiological signals in the diagnosis of epilepsy. Manual EEG signal interpretation, however, is labour-intensive, time-consuming, and prone to error. Automatic identification of epileptic focus using computer technology and signal processing has been developed as an adjunct diagnostic tool to improve the accuracy of epilepsy diagnosis and the success rate of epilepsy surgery. As a result, it is crucial to create a system that can swiftly tell apart ordinary EEG data from generalised and focal epileptic seizures. This study introduces a new technique for identifying the brain areas affected by focal and non-focal epilepsy, in order to help patients suffering from epilepsy resulting from abnormal brain electrical activity. The method employs a deep neural network and involves recording EEG signals from the brain, followed by decomposing them into sub-bands using the Fourier Decomposition Method. Next, features are calculated from each sub-band and a very effective features are analyzed further. The final classification of the EEG signals into focal and non-focal epilepsy is then performed. The system achieved the highest classification accuracy of $ 93.0\%$ with the lowest training and validation loss of $ 4.3\%$ and $ 13.3\%$, respectively, when using 10 sub-bands. The lowest accuracy was $\mathbf{9 0 . 0 \%}$ with training and validation loss of $ 2.9\%$ and $ 26.7\%$, respectively, when using 20 sub-bands. Using the minimum of 5 sub-bands resulted in a classification accuracy of $ 92.0\%$. The results obtained imply that the suggested approach may, in actual applications, prove to be a helpful tool for physicians in determining the cause of epilepsy.