Hybrid approach for the detection of epileptic seizure using electroencephalography input

• Published in: International journal of information technology (Singapore. Online) Vol. 16; no. 1; pp. 569 - 575
• Main Authors: Basha, Niha Kamal, Surendiran, B., Benzikar, Amutha, Joyal, S.
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.01.2024; Springer Nature B.V
• Subjects: Algorithms; Artificial Intelligence; Availability; Bioelectricity; Computer Imaging; Computer Science; Convulsions & seizures; Datasets; Decision analysis; Deep learning; Electroencephalography; Epilepsy; Image Processing and Computer Vision; Kurtosis; Machine Learning; Original Research; Pattern Recognition and Graphics; Software Engineering; Vision; Rhythmic frequency; Ictal subject; Deep learning model; Statistical feature; Epilepsy; Electroencephalography; Confusion matrix
• ISSN: 2511-2104; 2511-2112
• DOI: 10.1007/s41870-023-01657-1

In the early days, it was difficult to study bio-electric signals, but now a days these problems have been solved by many hardware devices which are available at low cost. Even then there is a need for technical improvements to process bio-electric signals. Enormous effort has been taken for the detection of epileptic seizure from EEG signal. Considerable evidence available on Bonn university EEG dataset for epileptic seizure detection. However, deep learning algorithms have not applied often on Bonn university data like other machine learning algorithms for the detection of epileptic seizure due to the less availability of data. This work adopted machine learning and deep learning models for the detection of epileptic seizure. The data consists of a hundred subjects with sampling rate as 173.61 Hz of ‘S’ (ictal) and ‘Z’ (normal) dataset. The results show how the choice of healthy and ictic subjects through high order functional math decision like variance (STD), Power, Skewness, and Kurtosis values are analyzed to extract epileptic features. The performance of classifiers has been evaluated based on the evaluation metrics within which CGRU-SVM outperforms all other models with 97.54% accuracy. The features are extracted based on the statistical measures to detect epileptic seizure. Further, the proposed work provides more evidence on the epileptic seizure features, and it shows the new possibility for using deep learning models for epileptic seizure detection. In future, an analysis must be performed with dynamic sets of input and different analysis techniques.