Development of an Epileptic Seizure Monitoring Information System Based on Intelligent Algorithms

Abstract

Epilepsy is a common neurological disorder that needs a direct interference to control the seizures properly. However, the detection of seizure often requires the presence of experienced neurologists, which is not always accessible. This challenge highlights the need to develop an automated system for seizure detection. In response, the main aims and objectives of this article are to propose and develop an algorithm for efficiently detecting instant notifications based on machine learning algorithms using EEG signals. Filtering technique is applied to the captured data to eliminate noise from the data before doing feature extraction, power spectral density (PSD) and spectral entropy. These features are then computed to train several classifiers like support vector machines (SVM), K-nearest neighbors (KNN), multi-layer perceptron (MLP), random forest (RF), and gradient boosting (GB). KNN turns out to be the most successful classifiers of all. In this model, precision was shown to be at 0. 94, recall of 0. 93, and F1-score of 0. 93 percentage accuracy for seizure identification which, in turn, gives a total of 95% for total accuracy. The contribution and novelty of this article is that it can classify the situations of epileptic seizure monitoring into three classes (pre-seizure), (seizure), and (non-seizure), the alert messages are sent through the GSM using an Arduino microcontroller that connects to a SIM900 module to send alert message during the seizure event.