Epileptic Seizure Prediction Based on Synchroextracting Transform and Sparse Representation
Article
Article Title | Epileptic Seizure Prediction Based on Synchroextracting Transform and Sparse Representation |
---|---|
ERA Journal ID | 210567 |
Article Category | Article |
Authors | Ra, Jee Sook, Li, Tianning and Li, Yan |
Journal Title | IEEE Access |
Journal Citation | 12, pp. 187684-187695 |
Number of Pages | 12 |
Year | 2024 |
Publisher | IEEE (Institute of Electrical and Electronics Engineers) |
Place of Publication | United States |
ISSN | 2169-3536 |
Digital Object Identifier (DOI) | https://doi.org/10.1109/ACCESS.2024.3514859 |
Web Address (URL) | https://ieeexplore.ieee.org/document/10788676/ |
Abstract | Feature extraction is crucial in machine learning and EEG analysis, where raw data often contains excess information. The prominence of machine learning has led to the development of numerous feature extraction methods over the past decade. This paper introduces an efficient feature extraction method that demonstrates superior experimental results. We employed the Synchroextracting Transform (SET) and Sparse Representation (SR) for enhanced feature extraction in epileptic EEG analysis. SET is a recently developed signal transformation technique, and SR effectively extracts information from multi-dimensional data. Our goal is to enhance time-frequency (TF) resolution using the SET-SR method, which offers a TF representation more concentrated with energy than traditional TF analysis methods. SR decomposes SET multi-dimensional sub-signals to accurately predict epileptic seizures. The significance of this feature extraction method was evaluated using a k-Nearest Neighbor (k-NN) algorithm, a traditional machine learning technique. Applying the SET-SR with the k-NN, we achieved an average accuracy of 99.48% on the CHB-MIT database and 100% accuracy on the Bonn University database in classifying pre-seizure signals. The SET-SR effectively detects pre-seizure signals, showing promise for developing an efficient patient-specific seizure prediction algorithm based on EEG data. Our findings demonstrate that enhanced feature extraction can reliably identify pre-seizure signals with high precision, even when using classical machine learning methods like k-NN. This research underscores the importance of feature extraction in EEG signal analysis and suggests that diverse classification methods can be employed for real-time seizure prediction while maintaining high accuracy. |
Keywords | EEG analysis; sparse representation (SR); epileptic seizure prediction; kNN; synchroextracting transform (SET) |
Related Output | |
Is part of | Epileptic seizure prediction with machine learning on EEG data |
Contains Sensitive Content | Does not contain sensitive content |
ANZSRC Field of Research 2020 | 460299. Artificial intelligence not elsewhere classified |
400399. Biomedical engineering not elsewhere classified | |
Public Notes | This article is part of a UniSQ Thesis by publication. See Related Output. |
Byline Affiliations | School of Mathematics, Physics and Computing |
https://research.usq.edu.au/item/zv1vw/epileptic-seizure-prediction-based-on-synchroextracting-transform-and-sparse-representation
Download files
Published Version
14
total views2
total downloads3
views this month2
downloads this month