Epilepsy Unveiled using a Multi-Classifier Approach integrated in EAS system for Enhanced Detection Using EEG Signal Analysis.

Abstract

The study focuses on the advantages of using non-invasive approaches to detect epileptic seizures, aiming to prevent injuries from sudden falls and improve medical diagnosis and management. It presents an EEG Acquisition System (EAS) that integrates a multi-level amplifier with four machine learning algorithms—K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Multilayer Perceptron Neural Networks (MLPNN), and Convolutional Neural Network (CNN)—for signal classification. These methods are chosen for their accuracy and rapid processing, making them ideal for real-time use. The research analyzed data from 310 participants to detect epileptic patterns under various conditions. The results show that the SVM model achieved the highest accuracy at 96.37%±2.68 in the small data, followed by the CNN with 95.25%±2.25 in real-time detection, MLPNN at 89.87%, and KNN with 86.37% . This demonstrates CNN's strong performance in seizure detection, complemented by a detailed comparison of sensitivity and specificity across all models to assess their predictive efficiency. The system tells the scientist if there is an epilepsy , dementia ,or a normal brain activity, if the system detects the abnormal activities, it could map it on the brain map with a precision of 98%.