Using deep learning methods for seizure detection in newborns by using EEG signals

Epileptic seizure is a clinical condition that occurs as a result of the sudden and irregular electrical discharge seen in all or part of the brain. Neonatal seizures show different clinical symptoms than those in childhood and adulthood. Failure to recognize seizures in time and the resulting lack of treatment can result in death in cases. It is of vital importance to detect seizures urgently and initiate treatment in newborn babies. For these reasons, there is a need for computer aided systems that assist experts in the decision-making process in the diagnosis of seizures. Deep learning is a class of machine learning algorithms that uses multiple layers to progressively extract higher-level features from raw input. Most deep learning methods use neural network architectures. Convolutional neural networks (CNN) eliminate the need for manual feature extraction in traditional machine learning methods, so you don't need to define the features used to classify images. For this reason, although deep learning went through a sluggish process for a while due to hardware deficiencies and insufficient data set, it has become popular today with the elimination of the deficiencies and has recently contributed greatly to increase the quality of health services. In this thesis, signal analysis was performed in all EEG channels by using many classification algorithms from machine learning algorithms for seizure detection in newborns. Then, spectrograms were obtained by converting 1-dimensional signal data to 2-dimensional images on the C4-P4 channel, which provides the highest performance using pre-trained deep learning architectures. The classification was made in both 1D and 2D on the data set. In the engineering field of the study, many methods have been tried for the success of the classification in the diagnosis of epileptic seizures and the algorithm successes have been tested for the most effective classification. In the field of medicine, the effect of deep learning methods on the EEG signals of newborns on the diagnosis of seizures has been analyzed. As a result of this study conducted with the proposed method, it is seen that high performance is obtained faster and with less cost.