A framework for seizure detection using effective connectivity, graph theory, and multi-level modular network

Objective: An epileptic seizure has a characteristic EEG pattern, which allows for its automatic detection. Statistical dependence between different brain regions measures by functional brain connectivity (FBC). Specific directional effects cannot be considered by FBC, and thus effective brain connectivity (EBC) is used to measure causal dependencies between brain regions. Our main purpose is to provide a reliable automatic seizure detection approach. Methods: In this study, three new methods are provided. Multi-level modular network (MLMN) is proposed based on combining various EBC classification results at different frequencies. Another method named modular effective neural networks (MENN). This method combines the classification results of the three different EBCs at a specific frequency. Modular frequency neural networks (MFNN) is another method that combines the classification results of specific EBC at seven different frequencies. Results: The mean accuracies of MFNN are 97.14 %, 98.53 %, and 97.91 % using directed transfer function (DTF), directed coherence (DC), and generalized partial directed coherence (GPDC), respectively. By using MENN method, the highest mean accuracy is 98.34 %. Finally, MLMN has the highest mean accuracy, which is equal to 99.43. To the best of our knowledge, the proposed method is a new method that provides the highest accuracy in comparison to other studies - that used the MIT-CHB database. Conclusion and significance: The knowledge of structure-function relationships between different areas of the brain is necessary for characterizing the underlying dynamics. Hence, features based on EBC can provide a more reliable automatic seizure detection approach. (C) 2020 Elsevier Ltd. All rights reserved.