Semi-dilated convolutional neural networks for epileptic seizure prediction

Epileptic Seizure Detection Using Rhythmicity Spectrogram and Cross-Patient Test Set

“…containing 844 hour of multi-channel scalp-EEG [15]. 23 pediatric patient data has been recorded who suffer from intractable seizures.…”

“…It 1 Palak Handa and 2 Nidhi Goel are with Department of Electronics and Communication Engineering, Indira Gandhi Delhi Technical University for Women (IGDTUW), New Delhi. palakhanda97@gmail.com and nidhi.iitr1@gmail.com has been used as an input image to various deep learning methods in several medical fields such as depth of anesthesia (DOA) [3], brain coma [5], human imagination [6], epilepsy and seizures [8], [9], [10], [11], [12], etc. Deep learning methods have shown remarkable diagnostic accuracy towards understanding complex patterns in medical AI due to their automated complex feature extraction efficacy [8], [9], [10].…”

“…Recent studies carried out using deep learning such as convolution neural networks (CNN), LSTM, Residual CNN, Deep neural networks, and transfer learning architectures etc. are trained on raw dataset instead of statistically developed features [8], [9], [10]. Such techniques have achieved higher accuracy levels but also helps in automating the entire pipeline making the pipeline 'end to end'.…”

“…Mandhouj et al [8] performed epileptic seizure classification of spectrograms developed from University of Bonn EEG signals. Hussein et al [9] generated scalograms from CHB MIT scalp EEG signals and performed seizure prediction using semi-dilated convolutional network (SDCN). An eight class classification of different seizures such as tonicclonic, clonic, and non-specific seizures etc.…”

“…Hussein et al [15] converted the scalp and invasive EEG signals to a scalogram using continuous wavelet transform and proposed a semi-dilated convolutional network for seizure prediction task. Another work analysed the ROC curve of converted EEG signals and sub-bands to images using STFT and naive bayes classifier [16].…”