Custom software code: Inter-regional recurrence dynamics of brain reveal the transition toward the preictal state and enable seizure prediction
Description
This repository contains source code relevant of the manuscript: Inter-regional recurrence dynamics of brain reveal the transition toward the preictal state and enable seizure prediction. Azam Ghanaei and Abbas Erfanian Summary: This study introduces a novel approach for seizure prediction in freely moving rats by utilizing nonlinear dynamical analysis of neural signals recorded from the right hippocampus (CA1), left anterior thalamic nuclei (ANT), and motor cortex (M1). The dynamic interactions among these three brain regions during epileptic activity are characterized using Cross-Recurrence Plot (CRP) analysis, a bivariate extension of the Recurrence Plot (RP) method that enables the analysis of dynamic interactions between two systems. The CRPs constructed from artifact-free intracranial electroencephalographic (iEEG) signals are subsequently used as inputs to a three-dimensional convolutional neural network (3D-CNN) for seizure prediction. The performance is compared with that of a 3D-CNN using RP-based features as input. The results indicate that CRP-based features are more effective in detecting the transition from the interictal to the preictal state in a real-time setting and in predicting seizure onset. For more details and code usage, please refer to the code description file.