An automated identification system using tunable-Q wavelet transform and centered correntropy features achieved 98.3% classification accuracy for seizure vs seizure-free EEG signals.
Does the proposed TQWT framework with CCE feature extraction accurately classify seizure EEG signals?
The proposed TQWT framework with CCE feature extraction provides high accuracy for classifying seizure EEG signals.
In the present work, EEG signals of different classes are analysed in tunable-Q wavelet transform (TQWT) framework. The TQWT decomposes the EEG signals into subbands and arrange them into decreasing order of frequency. The nonlinearity of the EEG signals is assessed by computing the centered correntropy (CCE) from the obtained subbands, which is further used as a feature for classifying the different categories of EEG signals. In this work, EEG signals are categorised in two different classification problems. First category is seizure free and seizure (NF-S) classes, and the other one is the normal, seizure free and seizure (ZO-NF-S) classes. Features obtained from the EEG signals of these classes are fed to the input of three different classifiers namely, random forest classifier (RF), multilayer perceptron (MLP) classifier, and logistic regression (LR) classifier. For NF-S classes, we achieved 98.3% classification accuracy with RF classifier for signal length of 1000 samples. The obtained accuracy of classification is 98.2% for ZO-NF-S classes using MLP classifier when features are extracted from signal length of 1000 samples.
Reddy et al. (Sun,) conducted a other in Seizure. Tunable-Q wavelet transform (TQWT) and machine learning classifiers was evaluated on Classification accuracy. An automated identification system using tunable-Q wavelet transform and centered correntropy features achieved 98.3% classification accuracy for seizure vs seizure-free EEG signals.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: