Intracranial EEG and Mutual Information: Identifying the Preictal State

Refractory epilepsy represents a significant clinical challenge due to the resistance of certain patients to pharmacological treatments capable of suppressing the seizures inherent to the disease. Although surgical removal of the affected areas may be a solution, this option is not always feasible, forcing patients to live with a limited quality of life. The use of intracranial electrodes iEEG allows obtaining high resolution signals both spatially and temporally, which have allowed the identification of three main brain states: basal, postictal and preictal. Early detection of the preictal state is crucial, as it can predict a seizure minutes before it occurs. In this study, mutual information (MI) was used to analyze statistical dependencies, both linear and nonlinear, between multichannel iEEG time series. both linear and nonlinear, between multichannel time series of $iEEG$. MI was applied to measure the transmission of information between different neuronal rhythms in status epilepticus, acting as an objective biomarker with potential for early diagnosis of seizures. Finally, MI calculations between different frequency bands were used to train neural networks that were able to predict the preictal state with an accuracy of 81%. This approach represents a promising tool to improve the quality of life of patients with refractory epilepsy through early seizure prediction.