Interpreting deep learning models for epileptic seizure detection on EEG signals

作者：

Highlights：

• Shallow Neural Networks achieve an F1 of 0.873 on seizure detection from raw EEGs

• DL models and human experts find discriminant phenomena at the same time scale

• The size of the first layer kernels is determinant in the explainability of the model

• The power in the alpha band is the most relevant feature for interictal classification

摘要

•Shallow Neural Networks achieve an F1 of 0.873 on seizure detection from raw EEGs•DL models and human experts find discriminant phenomena at the same time scale•The size of the first layer kernels is determinant in the explainability of the model•The power in the alpha band is the most relevant feature for interictal classification

论文关键词：Epilepsy,EEG Seizure detection,Interpretable deep learning,Convolutional neural networks

论文评审过程：Received 19 December 2020, Revised 27 April 2021, Accepted 29 April 2021, Available online 1 May 2021, Version of Record 12 May 2021.

论文官网地址：https://doi.org/10.1016/j.artmed.2021.102084