Seizure Detection of West Syndrome Based on Hybrid Optimization of Entropies With Multisensor EEG Data

West syndrome is a persistent disease with a common, age-dependent, severe neurological disorder that poses a substantial risk to intellectual and motor deficits in children. Research on intelligent seizure detection methods for West syndrome can help doctors make effective judgments. In this article, we propose an infantile spasm seizure detection algorithm based on hybrid optimization of entropy of entropy (EoE) and Shannon entropy (SE). Different from traditional EoE and SE, fusion features are extracted from the optimal division of EoE (ODEoE) and optimal SE (OSE) by considering the epoch length and the number of subepochs in an epoch jointly. In addition, multivariate entropy parameterization via enhanced multistrategy sparrow algorithm (MEPEM-SSA) is employed for choosing the ideal parameters. Finally, the data from Zhejiang University School of Medicine Children’s Hospital (ZUCH) dataset, the CHB-MIT dataset, and the Siena scalp electroencephalogram (EEG) dataset are extracted using EEG sensors, and a tenfold cross-validation evaluation is performed on these EEG data from West syndrome patients. The experimental results on the ZUCH dataset show that the mean values of precision, specificity, sensitivity, and
${F}1$
score of the algorithm are 98.08%, 97.93%, 98.46%, and 98.09% respectively. On the CHB-MIT dataset, the mean values of precision, specificity, sensitivity, and
${F}1$
score reach 98.76%, 99.27%, 97.94%, and 98.59%, respectively. On the Siena Scalp EEG dataset, the mean values of precision, specificity, sensitivity, and
${F}1$
score reach 98.33%, 98.04%, 98.39%, and 98.20%, respectively.