He, Zhisen and Yang, Jing and Alroobaea, Roobaea and Por, Lip Yee (2024) SeizureLSTM: An optimal attention-based trans-LSTM network for epileptic seizure detection using optimal weighted feature integration. Biomedical Signal Processing and Control, 96 (B). p. 106603. ISSN 1746-8094, DOI https://doi.org/10.1016/j.bspc.2024.106603.
Full text not available from this repository.Abstract
Epileptic seizures are a neurological disorder of the brain and a dangerous disease that can cause death. Rapid diagnosis is required to help clinicians treat patients. For diagnosis, the EEG plays a pivotal role in grasping brain activity via electrodes on the scalp, providing substantial information for analyzing the disorder. Due to the nature of non-linear and non-stationary processes, developing an epilepsy detection model is critical. Previous models had limitations such as increased time consumption, increased risk of manual error, and ineffective results. To overcome these, an intelligent learning classifier is introduced with a heuristic strategy. Initially, the raw EEG signals are gathered from data sources. Further, it undergoes decomposition, where Tunable Q Wavelet Transform decomposes the raw EEG signals into different frequency bands. Consequently, the decomposed signal is used to extract the informative signal features, which are obtained by the 1DCNN and spectral features. To get the optimal values, the parameters in 1DCNN are tuned by the FC-RDA. After obtaining the two different features, they are concatenated with each other by estimating the optimal weights; thus, it generates the weighted fused features. In turn, the weight optimization is done by FC-RDA algorithm. Finally, the weighted features are fed as input to the new model as OAT-LSTM in which the hyperparameters will be optimally selected by using the FC-RDA algorithm. The model's performance is computed with different measures and compared with classical models. Extensive results elucidate that the proposed detection model has the potential to rapidly diagnose the seizure disorder.
| Item Type: | Article |
|---|---|
| Funders: | UNSPECIFIED |
| Uncontrolled Keywords: | EEG signal; Epileptic seizure detection; Adaptive one-dimensional convolutional neural; network; Weighted fused features; Fitness count-based red deer algorithm; Optimal attention-based trans-long short-term; memory |
| Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science |
| Divisions: | Faculty of Computer Science & Information Technology > Department of Computer System & Technology |
| Depositing User: | Ms. Juhaida Abd Rahim |
| Date Deposited: | 05 Feb 2025 07:56 |
| Last Modified: | 05 Feb 2025 07:56 |
| URI: | http://eprints.um.edu.my/id/eprint/47565 |
Actions (login required)
 |
View Item |
