Unveiling the Dynamics of Epileptic Seizures Through Nonlinear EEG Analysis

Abstract

Epileptic seizures are sudden disruptions in brain activity that can have life-threatening consequences due to their unpredictable nature. Analyzing electroencephalogram (EEG) signals using nonlinear methods offers valuable insights into the brain’s behavior during different seizure phases: interictal (between seizures), preictal (preceding a seizure), and ictal (during a seizure). This study analyses the nonlinear characteristics of EEG signals collected from 10 epileptic patients. We investigate their dynamical differences as the seizure progresses through the interictal, preictal and ictal stages using nonlinear measures such as Hurst exponent, Lyapunov exponent, Sample entropy and Higuchi Fractal Dimension. Furthermore, the characteristics of the EEG signals are visualized by the Phase space portrait, Power spectral density, Recurrence Plot and quantified by means of Recurrence Quantification Analysis measures such as Determinism (DET), Average diagonal length (????????????????) and Recurrence Time Entropy (RTE). The results show that ictal EEG signals have lower values for entropy and fractal dimension than their interictal and preictal EEG signals, indicating the reduced complexity in the brain during the onset of the seizure. Similarly, the Hurst exponent was found to be greater than 0.5 for all three stages, with a rising trend in values as the brain passes from the interictal to the ictal stages. This demonstrates the persistent behavior and greater predictability of EEG signals in epileptic patients. The Lyapunov exponent revealed the existence of chaos in the brain. The long diagonal structures characterizing periodic behavior of the seizure EEG signals were identified from the Recurrence plot. We observed that the seizure EEG signals have high values for DET and ???????????????? and low values for RTE compared to interictal and preictal EEG signals, indicating that the seizure EEG signals were highly deterministic, recurrent and less complex. The phase portrait and the spectral analysis also justify these observations. As a result, this study suggests that during these three stages, the underlying dynamics of the epileptic brain is not only a chaotic complex dynamical system, but it is also highly deterministic in the sense that prediction of seizure activity is possible in the short term. These findings shed light on the insights into the dynamics of brain activity and may serve as a theoretical foundation for further research on the clinical diagnosis and for developing empirical models that will aid in better prediction of seizures in patients with epilepsy.