General and patient-specific seizure classification using deep neural networks

Abstract Seizure prediction algorithms have been central in the field of data analysis for improvement epileptic patients’ lives. The most recent advancements which include use deep neural networks to present an optimized, accurate seizure system. This work puts forth learning methods automate process detection with electroencephalogram (EEG) signals as input; both a patient-specific and general approach are followed. EEG time structure series motivating sequence such temporal convolutional (TCNNs), long short-term memory networks. We then compare this methodology other prior pre-implemented structures, including our previous using machine approaches support vector random under-sampling boost. Moreover, used evaluate performance best algorithms. Area under curve (AUC) is select performing algorithm account imbalanced dataset. presented TCNN model showed results than that with, AUC 0.73, while ML had classification 0.75.