Automated Classification and Diagnosis of Focal and Non-Focal EEG Signals using a Hybrid Classification Approach

As a result of the non-linear in the categorizing technique, the initial section of the study, which sorts EEG data into focal and non-focal, is not suitable for severe evaluation. A combined classification system is therefore suggested for the purpose of automated epilepsy illness diagnosis and level of severity evaluation. Stages for breakdown, extraction of features, and categorization make up the suggested approach. After applying DTWCT to the EEG signals, we scale them utilizing a signal's scaling coefficient and use the wavelet transform to break them down into their component bands. Then, DTCWT takes these deconstructed sub band correlations and uses them to extract parameters like average, variance, skewing, average deviation, kurtosis, covariance, or period, and coefficients of correlation. The next step is to use a Neural Network (NN) categorization strategy to sort the EEG data into two categories: focal and non-focal. "Mild" and "Severe" are the additional diagnostic categories used for the categorized focused EEG data. A larger quantity of learned signals is needed by NN in order to perform the diagnostic procedure. If fuzzy limitations might improve the NN classifier's diagnostic accuracy, they aren't there. In comparison to Shearlet Transform, Discrete Wavelet Transform (DWT), and Contourlet Transform, the proposed method that makes use of DTCWT maximizes accuracy to 100%. This establishes the hybrid categorization approach as the gold standard for automated seizure detection.