Background
Based on the symptoms experienced during the episode and the Electroencephalograph (EEG) recording made during the inter-ictal phase, the doctor makes the epileptic seizure type diagnosis. The fundamental issue, however, is that patients frequently struggle to explain their symptoms in the absence of an observer and identify traces in inter-ictal EEG patterns.
Aims
This study examines electroencephalographic (EEG) signals from epileptic seizures in order to diagnose seizures in pre-ictal, ictal, and inter-ictal stages and to categorize them into seven groups.
Methods
For the investigation, a licensed dataset from Temple University Hospital was used. Seven different seizure types are pre-processed from the seizure corpus and divided into pre-ictal, ictal, and inter-ictal stages. K-Nearest Neighbor (KNN), Random Forest, and other machine and deep learning techniques were used to perform the multi-class categorization.
Result
With 20 channels and an 80 − 20 train-test ratio, multiclass classification of seven different types of epileptic seizures was accomplished. For the pre-ictal, ictal, and inter-ictal stages, weighted KNN achieved accuracy levels of 94.7%, 94.7%, 69.0% during training and 94.46%, 94.46%, and 71.11% during testing.
Conclusion
Seven epileptic seizure type classification using machine learning techniques carried out with MATLAB software and weighted KNN shows better accuracy comparatively.
Electroencephalography (EEG) is essential for tracking brain activity and identifying seizure effects. However, epileptic behaviour can only be detected after a specialist has carefully analysed all EEG recordings along with a proper history of the patient. A skilled physician is required for the right epilepsy diagnosis and therapy. But most of the time, patients visit the clinician in the interictal stage with no proper history documented. Therefore, it was essential to the automatic prediction of stages of seizure. K nearest neighbours (KNN) and random forest (RF) models using raw EEG signals, preictal, ictal, postictal, and interictal stages were identified in this study. The possibility of these characteristics is explored by examining how well time-domain signals work in the prediction of epileptic stages using intracranial EEG datasets from Freiburg Hospital (FH), Children's Hospital Boston-Massachusetts Institute of Technology (CHB-MIT), and Temple University Hospital (TUHEEG). To test the viability of this approach, two different types of simulations were carried out on three binary classifications (interictal vs. preictal, interictal vs. ictal, preictal vs. postictal, and interictal vs. postictal), and one four-class problem (interictal vs. preictal vs. ictal vs. postictal) was performed for each model. The average accuracy when using time-domain signals in the FH database was 90.5% and 75.0%; CHB-MIT was 92.87% and 75.9%; and TUHEEG was 94.46% and 76.8%, respectively, for the KNN and RF models.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.