Title: A heuristic aided optimal deep temporal convolution layer with an attention-based long short-term memory framework for epileptic seizures detection

Authors: Vedavati Bhandari; Manjaiah Doddaghatta Huchaiah

Addresses: Department of Computer Science, Mangalore University, Karnataka, 574199, India ' Department of Computer Science, Mangalore University, Karnataka, 574199, India

Abstract: The EEG signals required for modelling our system are gathered. The most crucial features are mined from the obtained input EEG signal. These features are now given for the optimal weighted feature selection phase. Here the weights are tuned, and the features are optimally selected utilising the newly suggested algorithm, called the fused Archimedes with differential evolution optimisation algorithm, which integrates Archimedes optimisation algorithm and differential evolution technique. The tuned weights are combined with the optimally selected features to obtain the weighted fused features. These weighted fused features are fed to the optimal deep temporal convolution layer with an attention-based long short-term memory unit for epilepsy detection and classification. The same FA-DEOA algorithm is utilised by this classifier to tune the elements in it. The efficient detection of the epileptic seizure by the suggested model is confirmed by conducting experiments and validating it against other state-of-the-art models.

Keywords: epilepsy; electroencephalogram; seizure detection; spatial feature extraction; temporal feature extraction; optimal deep temporal convolution layer with attention-based long short-term memory; fused Archimedes with differential evolution optimisation algorithm.

DOI: 10.1504/IJBRA.2026.154827

International Journal of Bioinformatics Research and Applications, 2026 Vol.22 No.3, pp.266 - 293

Received: 14 Jul 2023
Accepted: 04 Dec 2023
Published online: 15 Jul 2026 *

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