An efficient epileptic seizure detection based on tunable Q-wavelet transform and DCVAE-stacked Bi-LSTM model using electroencephalogram

The recording and measurement of the electric events of the brain utilizing an electroencephalogram (EEG) has turned out to be a prominent equipment among the plethora of devices used for the analysis of neural ailments in the previous few years, especially for seizure detection. It is condemnatory to ascertain if the state of an epilepsy patient’s brain is symptomatic of a plausible seizure beginning; hence, EEG signals aid with the provision of significant information regarding the epileptogenic networks which is meant to be scrutinized and comprehended prior the commencement of therapeutic proceedings. And so relevant alarm or treatment could be provided in time. Desirable seizure prognosis depends on the ability to appropriately distinguish the preictal phase from the interictal phase of ictal in EEG. The pre-programmed seizure indication system can efficiently aid physicians to diagnose and proctor epilepsy which also decreases their amount of work. Various different remarkable studies are found to give great outcomes in the complication two-class seizure diagnosis, but the majority of the feature extraction techniques are contingent on manual techniques. This work nominates a start to end seizure indication structure that is automatic and whose methodology is grounded on deep learning. To achieve this, EEG signals from CHB-MIT Scalp EEG database. The Deep convolutional variational autoencoder (DCVAE) was utilized to study the expressed seizure features from EEG data and The tunable Q-factor wavelet transform (TQWT) was utilized for pre-processing. Later, these vigorous EEG characteristics relevant to seizures are provided as the input to the stacked bi-directional long short-term memory (SB-LSTM) model for the automatic detection of epileptic seizures. One of the best models, achieved an accuracy of 99.6%, sensitivity of 98.85%, and specificity values of 99.8%, respectively.