An approach for ideal detection of Epileptic Seizures using CAD techniques - DWT, LDA and Machine Learning algorithms

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

Neurological disorder, epilepsy, can be detected more precisely using the appropriate analysis method from its most reliable and convenient diagnosis method - Electroencephalography (EEG). This paper examines a number of formerly proposed seizure detection methods and, finally, proposes a superior method for seizure detection using EEG time series data. In the proposed method, we have acquired the use of discrete wavelet transform (DWT) for signal decomposition followed by the implementation of universal thresholding in each sub-band to eliminate the non-significant coefficients on the basis of hard threshold function. The features were extracted from the significant coefficients of DWT using linear discriminant analysis (LDA). This framework was analyzed using machine learning (ML) classifiers. Classification algorithms - random forest (RF), support vector machine (SVM), naive bayes (NB) and K-nearest neighbor (KNN) is implied on all the 15 different possible combinations developed from the dataset to examine the performance result obtained from each classifier in detecting epilepsy. The entire model was employed on publicly available EEG time series dataset available from the University of Bonn for a comparative analysis of the proposed studies to date. Cent percent classification results were accomplished by this model, which is better than any other model till date. As a result, it can be inferred that this model has the potential to be a more reliable method for seizure detection, as well as a potential supplementary method at the clinical level.