Classification of 1D and 2D EEG Signals for Seizure Detection in the Newborn Using Convolutional Neural Networks

Abstract

Unlike adults, neonates do not always show clinical symptoms during seizures. Therefore, uncontrolled seizures lead to severe brain damage. Timely recognition of seizures plays a crucial role for neonates. In this study, a deep transfer learning approach was proposed for automatic detection of seizures on the C4-P4 channel using electroencephalography (EEG) signals from neonates. The EEG signals were used in 1D and 2D dimensions to ensure performance, robust functionality, and a clinically acceptable level of detection accuracy. The pre-trained deep learning models Alexnet, ResNet, GoogleNet and VggNet were used in the study. Spectrograms were obtained by converting 1-dimensional signal data into 2- dimensional images, and then classification was performed for both the 1D and 2D datasets. For 1D classification, the highest performance was obtained by VggNet architecture with 91.67%, while 2D classification was obtained by AlexNet and ResNet architecture with 95.83%. The use of spectrograms significantly improved classification performance and made neonatal seizure detection and decision-making more clinically reliable.