Classification and Diagnosis of EEG Signals for Epilepsy Detection Using a Proposed CNN-FCM Approach

An innovative structure for EEG analyzing signals designation, CNN-FCM, is introduced. To address the limitations of traditional CNN’s simulations the suggested approach employs parallel structures to reduce detector time, making it ideal for categorization of focal or non-focal EEG signals. Improving EEG information, decomposing signals via the EMD technique, and extracting features from inherent properties for categorization are all part of the approach. In turn, CNN-FCM uses many convolutional neural network layers to combine this information, improving the precision of classification above what it was before. When the system was evaluated in two public reference EEG datasets, the CHB-MIT and the Bern-Barcelona, it achieved gains of up to 99.99 percent in pattern classification capacity parameters including sensitivity, specificity, and accuracy. The CNN-FCM approach outperforms both the conventional CNN approach and models generated from k-means clustering, particularly when dealing with unbalanced sets of information. The suggested structure was also used to conduct a diagnostic check for the severity of the focused signals, and it produced a $100 \%$ right rate of categorization involving both mild and moderate signals. We see the fact that the CNN-FCM model with EMD is a rapid and dependable way to classify EEG signals, and that it requires much less time to identify signals compared the conventional approaches.