Title: EEG-based epileptic seizure state detection using deep learning

Authors: Vibha Patel; Dharmendra Bhatti; Amit Ganatra; Jaishree Tailor

Addresses: Department of Computer Engineering, Chhotubhai Gopalbhai Patel Institute of Technology, Uka Tarsadia University, Bardoli - 394601, Gujarat, India ' Uka Tarsadia University, Bardoli – 394601, Gujarat, India ' Parul University, Waghodia, Vadodara – 391760, Gujarat, India ' Shrimad Rajchandra Institute of Management and Computer Application (SRIMCA), Uka Tarsadia University, Bardoli – 394601, Gujarat, India

Abstract: Artificial intelligence-assisted diagnostics are booming with advanced computing power and technology. An automated approach to detect the seizure state from EEG recordings is highly desirable as the manual approach is tedious, time-consuming, and prone to errors. Our work proposes a hybrid deep learning architecture for automated seizure state detection from long-term patient-specific EEG. The architecture uses one-dimensional convolutional neural network (1D-CNN) and stacked long short-term memory (LSTM) networks. An open-source epilepsy dataset, CHB-MIT, is used in this work for experiments. The synthetic minority oversampling technique (SMOTE) is used for handling class imbalance issues. Our proposed approach achieves an average of 90% accuracy, sensitivity, and specificity with an AUC value of 0.96 and an FPR of 0.10. This performance is remarkable, considering varying EEG channels, channel montages, and EEG durations. Our work facilitates seizure detection devices for faster and more precise decision-making for epilepsy treatment.

Keywords: artificial intelligence; machine learning; deep learning; epileptic seizure detection; convolutional neural network; CNN; long short-term memory network; LSTM.

DOI: 10.1504/IJMIC.2024.135541

International Journal of Modelling, Identification and Control, 2024 Vol.44 No.1, pp.57 - 66

Received: 29 Aug 2022
Received in revised form: 18 Oct 2022
Accepted: 24 Oct 2022
Published online: 18 Dec 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article