Optimal Selection of Intrinsic Mode Functions Applied to Seizure Detection

Context: Epilepsy is a severe chronic neurological disorder with considerable incidence due to recurrent seizures. These seizures can be detected and diagnosed noninvasively using an electroencephalogram. Empirical mode decomposition has shown excellent results in identifying epileptic crises.Method: This study addressed a significant gap by proposing a novel approach for the automated selection of the most relevant intrinsic mode functions (IMFs) using empirical mode decomposition and discrimination metrics such as the Minkowski distance, the mean square error, cross-correlation, and the entropy function. The main objective was to address the challenge of determining the optimal number of IMFs required to accurately reconstruct brain activity signals.Results:The results were promising, as they facilitated the identification of IMFs that contained the most relevant information, marking a significant advancement in the field. To validate these findings, standard methods including the correlation coefficient, the p-value, and the Wasserstein distance were employed. Additionally, an EEGLAB-based brain mapping was conducted, adding robustness and credibility to the results obtained. Conclusions: Our method is a fundamental tool that enhances epileptic seizure identification from EEG signals, with significant clinical implications in the diagnosis and treatment of epilepsy.