Spatio-temporal evaluation of epileptic intracranial EEG based on entropy and synchronization: A phase transition idea. (August 2022)

• Record Type: Journal Article
• Authors: Zhong, Lisha; He, Shuling; Yi, Fangji; Li, Xi; Wei, Linran; Zeng, Chen; Huang, Zhiwei; Li, Zhangyong
• Is Part Of: Biomedical signal processing and control. Volume 77(2022)
• Publication Date: 2022-08
• Subjects: Phase transition -- iEEG -- Sample entropy -- Pearson correlation -- Spatio-temporal evaluation -- Seizure prediction; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.103689
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Epileptic phase transition model was constructed by combining sample entropy and Pearson correlation coefficients from the temporal and spatial features respectively. Results showed that seizures would be detected by detecting the critical point with an accuracy of 82.9%, sensitivity of 78.1% and FPR of 0.12/h. Spatial–temporal evolution of epileptic iEEG can be visualized through the two-dimensions plane. Abstract: The sudden onset of epilepsy resembles the critical phenomenon of a phase transition. The transition from normal to seizure undergoes the critical preictal state termed epileptic phase transition (EPT) in this paper. This EPT model contains two essential features, entropy and synchronization, to measure the fluctuations and correlations of intracranial electroencephalogram (iEEG) signals, respectively. The results show that the sample entropy is significantly greater in the critical state than that in the interictal state. Spatial connectivity of iEEG channels enhances rapidly at critical points to form a pathway either horizontally or vertically. These phenomena are similar to the percolation associated with critical transitions. This simple percolation measure achieved an accuracy of 82.95%, sensitivity of 79.55% and FPR of 0.136/h for seizure prediction with 30- minute early warning on canine data. EPT model also suggests a dynamical visualization scheme for the temporal and spatial evaluation of brain activity to help elucidate the mechanism of seizure.