Robust Prior Stage Epileptic Seizure Diagnosis System using Resnet and Backpropagation Techniques

The human Brain generates the Electroencephalogram (EEG) signals, which contains physiological information. This EEG oriented the human brain activities, making it utilized for epileptic seizure identification and diagnosis. The present epilepsy visual techniques take a large amount of time for inspection of EEG data. The human proficient is identified epilepsy, but slow diagnosis happens. Therefore patient may often inconsistence stage; it sometimes causes deaths. In this investigation, an advanced early-stage epileptic seizure identification and classification framework has developed. Which includes two steps, at first step EEG signal has preprocessed with ResNet deep learning mechanism. This technique calculated the abnormal signal identification at alpha, beta, gamma and delta waveforms. These observations find out the patient early-stage epileptic seizure detection. The ResNet learning model trains the suspicious raises in the EEG signal, but clear classification required. Thus moving to the second stage, in this backpropagation (BP) auto stack encoder is classifying epilepsy efficiently. The dataset is a real-time clinical EEG database collected from practical and meaningful epilepsy patients. For testing and training, CHB-MIT datasets are selected for the proposed framework, and proposed ResNet and BP achieved a classification accuracy of 99.83% and throughput 99.72%.