Epilepsy is one of the most common chronic neurological disorders, affecting more than 50 million people worldwide. Although most patients are treated with antiepileptic drugs, about 30% fail to achieve adequate seizure control and are classified as drug-resistant. In these cases, although surgery may be a solution, it is not always feasible. Electroencephalogram (EEG) is a common technique for measuring brain activity, but it has limitations in spatial resolution and may include involuntary movement noise, suggesting the use of intracranial electrodes (EEG_i) to obtain better recordings.

EEG_i reflects the structure of neuronal rhythms interacting on different spatiotemporal scales, a phenomenon commonly referred to as frequency cross-coupling. It has been proposed that this type of coupling is a fundamental mechanism involved in cognitive functions, allowing coordination between neuronal groups. In addition, several studies have mentioned the possibility of identifying states predictive of epileptic seizures, which can be interpreted as deviations from the EEG_i reference signal, even when these changes are not easily detectable visually.

In this context, the present study uses mutual information analysis (MI) in multichannel EEG_i to measure the transmission of information between different neural rhythms in patients with epilepsy. MI, which measures statistical dependencies between two time series, is proposed as a potential biomarker to identify signal changes before an epileptic seizure.

As a result, it was observed that in the four patients studied, the preictal state showed high MI between beta-teta, alpha-teta and delta-teta rhythms, without this being attributable to significant cross-coupling between phases or amplitudes of the different bands.