Automated Epileptic Seizure Detection from EEG signals using Deep CNN model

This is the authors' abstract. We don't add key points for this paper.

Epilepsy is a condition that affects the brain and is characterized by irregular disruptions in theregular functioning of the brain. Epileptic seizures affect about one percent of the humanpopulation. At some point in their lives, about ten percent of the population in the United Statesexperiences at least one seizure. The process of automated seizure detection from EEG signals forepilepsy using Convolutional Neural Networks (CNN) entails the training of a deep learning modelto discern seizure-associated patterns within brainwave data. CNN receives EEG signals, whichrepresent electrical activity in the brain, and it acquires the ability to differentiate between statesof normalcy and seizures. Utilizing CNN's capability to autonomously extract and acquireknowledge of features from unprocessed EEG data enhances the precision of detection. Theimplementation of automation improves the efficiency and dependability of seizure diagnosis, whichmay facilitate prompt medical intervention. This study introduces a new and innovative deepConvolutional Neural Network model with back-propagation methods. The model is designed toclassify EEG signals into three specific classes: pre-ictal, normal, and seizure. The experiment isconducted on a publicly accessible benchmark database from Bonn University. The model'sperformance was assessed by measuring its sensitivity, specificity, and accuracy by 10-fold crossvalidation.The acquired experimental results include an accuracy of 97.33%, sensitivity of 96.00%,and specificity of 98%. These results are then compared to those of other existing literature.