2018/01/16 by Ihsan Ullah, Ullah, Ihsan, Muhammad M. Hussain +5 · 3 citations
Computer Science · Medicine · Neuroscience · #Blind Source Separation Techniques #Computer Vision and Pattern Recognition (cs.CV) #EEG and Brain-Computer Interfaces #Epilepsy research and treatment #FOS: Computer and information sciences
paper · pdf · doi:10.48550/arxiv.1801.05412
openalex publication_date 2018/01/16 · openalex created_date 2022/09/28 · openalex updated_date 2026/07/28
Epilepsy is a neurological disorder and for its detection, encephalography\n(EEG) is a commonly used clinical approach. Manual inspection of EEG brain\nsignals is a time-consuming and laborious process, which puts heavy burden on\nneurologists and affects their performance. Several automatic techniques have\nbeen proposed using traditional approaches to assist neurologists in detecting\nbinary epilepsy scenarios e.g. seizure vs. non-seizure or normal vs. ictal.\nThese methods do not perform well when classifying ternary case e.g. ictal vs.\nnormal vs. inter-ictal; the maximum accuracy for this case by the\nstate-of-the-art-methods is 97+-1%. To overcome this problem, we propose a\nsystem based on deep learning, which is an ensemble of pyramidal\none-dimensional convolutional neural network (P-1D-CNN) models. In a CNN model,\nthe bottleneck is the large number of learnable parameters. P-1D-CNN works on\nthe concept of refinement approach and it results in 60% fewer parameters\ncompared to traditional CNN models. Further to overcome the limitations of\nsmall amount of data, we proposed augmentation schemes for learning P-1D-CNN\nmodel. In almost all the cases concerning epilepsy detection, the proposed\nsystem gives an accuracy of 99.1+-0.9% on the University of Bonn dataset.\n