Machine learning approach for epileptic seizure detection using the tunable-Q wavelet transform based time–frequency features

• Published in: International journal of information technology (Singapore. Online) 2022-12, Vol.14 (7), p.3495-3505
• Main Authors: Pattnaik, Sasweta, Rout, Nirmal, Sabut, Sukanta
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; Computer Imaging; Computer Science; Image Processing and Computer Vision; Machine Learning; Original Research; Pattern Recognition and Graphics; Software Engineering; Vision
• ISSN: 2511-2104; 2511-2112
• DOI: 10.1007/s41870-022-00877-1

Epilepsy is a prevalent neurological disorder among numerous neurons degenerative diseases after brain stroke. During a seizure event, there are bursts of electrical activity in the cerebral cortex of brain that disturbs the normal activity of patients. This abnormal electrical activity of the brain can be assessed by electroencephalogram (EEG) signals. In this work, the wavelet transform is used to detect epileptic seizures on EEG signals and classified with machine learning methods for seizure and non-seizure events. The recorded EEG signals were collected from the CHB-MIT scalp EEG dataset and total 48 events were considered for analysis. The EEG signal is decomposed into different sub-bands using the tuneable Q-wavelet transform (TQWT) and time–frequency based feature like entropy, temporal measures were extracted to make a large data set in order to detect the epilepsy events accurately. The data set is processed further for classification of epilepsy using support vector machine (SVM) classifier and random forest (RF) classifier. It was observed that the RF classifier achieved better classification performance in terms of sensitivity of 91.5% and accuracy of 93% than the SVM classifier which were 89.2 and 90.4% only. Hence the proposed system based on TQWT and RF classifier is a better choice for accurate detection of epilepsy in clinical practices.