Deep Learning Approaches For Epilepsy Seizures Detection

The most prevalent neurological illness is epilepsy, and a precise seizure forecast could make patients feel less uncertain and helpless. Electroencephalograms can be used to diagnose conditions involving the brain, such as epilepsy (EEG). However, manual analysis of EEG data is a method that is known to have relatively low inter-rater agreement and needs highly skilled doctors (IRA). Additionally, manual interpretation is a time-consuming, resource-intensive, and expensive procedure due to the amount of data and the rate at which new data is being made available. Additionally, every institution struggles to find qualified neurophysiologists with the necessary skills to analyse EEG findings. In contrast, by speeding up diagnosis and lowering human error, automated EEG data processing has the potential to enhance patient care. Utilizing technological advancements like deep learning is crucial and can lead to excellent outcomes. We use the TUH Abnormal EEG Corpus to provide the data that Deep Learning requires. Three models are described and trained on the corpus, with each model building on top of the others. The best one achieves a classification accuracy rate of 83.8%. The model makes use of cutting-edge deep learning techniques such densely connected convolutional neural networks, recurrent units, inception modules, and one-dimensional convolution (for speech technology).