2020/12/01 by Khansa Rasheed, Junaid Qadir, Rasheed, Khansa +7 · 4 citations
Computer Science · Medicine · Neuroscience · #Blind Source Separation Techniques #EEG and Brain-Computer Interfaces #Epilepsy research and treatment #FOS: Computer and information sciences #Machine Learning (cs.LG)
paper · pdf · doi:10.48550/arxiv.2012.00430
openalex publication_date 2020/12/01 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Prediction of seizure before they occur is vital for bringing normalcy to the\nlives of patients. Researchers employed machine learning methods using\nhand-crafted features for seizure prediction. However, ML methods are too\ncomplicated to select the best ML model or best features. Deep Learning methods\nare beneficial in the sense of automatic feature extraction. One of the\nroadblocks for accurate seizure prediction is scarcity of epileptic seizure\ndata. This paper addresses this problem by proposing a deep convolutional\ngenerative adversarial network to generate synthetic EEG samples. We use two\nmethods to validate synthesized data namely, one-class SVM and a new proposal\nwhich we refer to as convolutional epileptic seizure predictor (CESP). Another\nobjective of our study is to evaluate performance of well-known deep learning\nmodels (e.g., VGG16, VGG19, ResNet50, and Inceptionv3) by training models on\naugmented data using transfer learning with average time of 10 min between true\nprediction and seizure onset. Our results show that CESP model achieves\nsensitivity of 78.11% and 88.21%, and FPR of 0.27/h and 0.14/h for training on\nsynthesized and testing on real Epilepsyecosystem and CHB-MIT datasets,\nrespectively. Effective results of CESP trained on synthesized data shows that\nsynthetic data acquired the correlation between features and labels very well.\nWe also show that employment of idea of transfer learning and data augmentation\nin patient-specific manner provides highest accuracy with sensitivity of 90.03%\nand 0.03 FPR/h which was achieved using Inceptionv3, and that augmenting data\nwith samples generated from DCGAN increased prediction results of our CESP\nmodel and Inceptionv3 by 4-5% as compared to state-of-the-art traditional\naugmentation techniques. Finally, we note that prediction results of CESP\nachieved by using augmented data are better than chance level for both\ndatasets.\n