Title: Automated epileptic seizure classification using adaptive fast Fourier transform with non-uniform sampling and improved deep belief network

Authors: Najmusseher; P.K. Nizar Banu; D.C. Janardhan

Addresses: Department of Computer Science, CHRIST (Deemed to be University), Central Campus, Bangalore, 560029, India ' Department of Computer Science, CHRIST (Deemed to be University), Central Campus, Bangalore, 560029, India ' Bangalore Medical College and Research Institute, Government of Karnataka Bangalore Medical College and Research Institute, Government of Karnataka, Bangalore-560002, India

Abstract: In automated brain-computer interaction (BCI), EEG signals are essential. This research uses AI to detect epileptic seizures, employing data from the BONN dataset (UCI), CHB-MIT dataset (physionet server), and Bangalore EEG Epilepsy Dataset (BEED). The goal is to develop an automated system for accurate seizure detection using adaptive fast Fourier transform with non-uniform sampling (AIFFT-NS) and an improved deep belief network (IDBN) model to enhance classification accuracy. The AIFFT-NS model serves as a channel for transforming spectro-temporal data. Using various EEG datasets, a number of extensive experiments are carried out, resulting in the validation of the efficacy of the proposed approach. High accuracy metrics, with 96.16% for the BEED dataset, 99.41% for the BONN dataset, and 96.31% for the CHB-MIT dataset, represent the evidentiary outcomes. This study emphasises the critical function of AI-facilitated spectro-temporal EEG analysis within the domain of medical diagnostics, going beyond the realm of automated seizure onset classification.

Keywords: EEG signals; brain-computer interaction; BCI; artificial intelligence; epileptic seizure onset classification; improved deep belief network; IDBN; fast Fourier transformation; extreme gradient boosting.

DOI: 10.1504/IJIEI.2024.142429

International Journal of Intelligent Engineering Informatics, 2024 Vol.12 No.4, pp.460 - 512

Received: 15 Jan 2024
Accepted: 28 May 2024
Published online: 30 Oct 2024 *

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